Skeptics Guide to the Universe Forums

General Discussions => Tech Talk => Topic started by: Obsessed With Gaming on December 09, 2018, 03:27:50 PM

Title: Will A.I replace most if not all of our jobs?
Post by: Obsessed With Gaming on December 09, 2018, 03:27:50 PM
And what will people do if they don't have a job such as go out with friends, stay at home all day?

And how will people survive if they don't have money because A.I replaced all their jobs?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: John Albert on December 09, 2018, 03:57:06 PM
What makes you think A.I. will replace most or all of our jobs?

Got any evidence to back up this begged question?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on December 09, 2018, 05:37:42 PM
Ideally, we end up like Star Trek.  Society aligns with humanity's health and best interests.

Probably, we end up like Elysium.  Society doesn't align so, and access to goods and services is still mediated by either (1) having lots of money or (2) being employable. 

Got any evidence to back up this begged question?

Mass unemployment's plausible in the near-term. 

Article: Every study we could find on what automation will do to jobs, in one chart (https://www.technologyreview.com/s/610005/every-study-we-could-find-on-what-automation-will-do-to-jobs-in-one-chart/)
From: MIT Technology Review
Date: 2018 JAN 15

Quote
...

Here’s the problem: the findings cited emanate from a wide array of studies released by companies, think tanks, and research institutions. And their prognostications are all over the map. [...]

...

(https://i.imgur.com/yiXyBtQ.png)

...

Of course, not all statistics are created equal. The most commonly cited numbers are from three places: a 2013 Oxford study (not listed in the table) that said 47 percent of US jobs are at high risk of automation in the next few decades, an OECD study suggesting that 9 percent of jobs in the organization’s 21 member countries are automatable, and a McKinsey report from last year that said 400 million to 800 million jobs worldwide could be automated by 2030. [...]

...

With regards to the long-term, it's reasonable to surmise that eventually we'll run out of things we do better than machines and computer.  At that point, culture and politics will decide what happens.

And what will people do if they don't have a job such as go out with friends, stay at home all day?

And how will people survive if they don't have money because A.I replaced all their jobs?

Wrt near-term mass-unemployment, I'd hope for Basic Income and 20 hour work weeks.  Share the benefits from productivity gains and split jobs up between more people.

Wrt long-term total obviation of labor, I think this would be tantamount to post-scarcity.  "Fully Automated Luxury Communism," is my favorite 'comedy option' descriptor.

In the latter scenario, we'd still constrained by resources and science/engineering.  Because of that, I expect everyone would have access to a baseline but would still somehow compete for access to above-baseline goods and services. 

Also, people'd still want work (of some kind) for all the other reasons they work. People would work for some combination of social standing, purpose, fulfillment, identity, avoiding atrophy, etc.  Most people acclimate to baseline then want a better diet, better residence, etc.

I think all you'd really do is remove, "not dying in the next few weeks from starvation or exposure," from the list of motivations people have for working. 

Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on December 09, 2018, 05:54:54 PM
A.I. changes the nature and character of work. It displaces some workers while creating jobs for others. A more relevant issue is that A.I. is an extension into the modern world of industrial automation, which transfers wealth from laborers to owners. When I was growing up in the 1950’s we were told that automation would relieve people of the need to work. We were told that people would work 5 or 6 hours a week (because machines would do the rest) and we would all have massive amounts of free time.

What happened was very different. As labor became more productive and less labor produced more wealth, the profits of that wealth all went to the owners, not to the workers. Wages did not rise with productivity, but rather workers had to work the same number of hours as before to make ends meet.

There was a blip in the U.S. after WWII because we were one of very few industrial economies undamaged by the war, and workers made great advances. But once the rest of the world repaired itself, and emerging nations became competitive, we’ve seen the trend go back to normal: the rich get richer and the poor get poorer. Now it takes two wage-earners to support a family because all this increased productivity goes into the pockets of the owners.

A.I. will continue the trend and unskilled labor will earn less and less, while the middle class shrinks as the need for skilled and educated labor decreases.

Eventually the U.S. will look like a third-world country. Either that or general discontent will devolve into rioting and lead to either a military dictatorship or the collapse of the economy.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on December 09, 2018, 06:39:44 PM
Yeah, I've seen that process first-hand while working as a production machinist.

The shop bought machine which could:The work output of that machine?  In two days, it could do what previously would've taken multiple people and multiple machines a week to do. 

And where did the benefits accrue?  That shop had no profit sharing.  They were stingy as hell with raises. And machinists are all the more interchangeable with off-the-street newbies as the machines improve so their bargaining position just gets worse and worse.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on December 09, 2018, 07:43:47 PM
This seems relevant to this conversation.

https://www.youtube.com/watch?v=7Pq-S557XQU
Title: Re: Will A.I replace most if not all of our jobs?
Post by: 2397 on December 09, 2018, 11:12:39 PM
That is the point of technology, to be able to do more with less, as well as to do things we couldn't do before.

I'm okay with doing away with as many mundane jobs as possible. If they don't need to be done by humans, why should they? The key issue being maintaining or further developing democracy and ensuring that people hold the power, rather than only a minority of individuals with money and wealth.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Tassie Dave on December 10, 2018, 01:03:24 AM
Anyone who thinks businesses are using technology to replace us in mundane jobs for our own benefit is delusional.
They are doing it for their bottom line and their shareholders.

The industry I work in (Mining) is one of the main users of technology to replace humans.
There are whole mines in Australia ( and I assume the world) that have no human driven machinery. All trucks, loaders etc are computer controlled with 1 or 2 humans monitoring them from a control room.

Eventually every section of mining will be computer controlled.

I only have 12 years to retirement, so I'll be ok. But the next generation will not have work in this industry.

Title: Re: Will A.I replace most if not all of our jobs?
Post by: Calinthalus on December 10, 2018, 07:35:37 AM
As soon as it becomes more profitable for Walmart to replace drivers with self-driving semi-trucks that will happen overnight; followed by every other trucking company.  In the short term, this will only be used for long-haul trucking, but eventually truck driving will not be a thing people are hired to do.  Right now there are 3.5 million truck drivers in the U.S. alone.  This is just one of thousands of jobs that AI/tech will lower the number of people needed.  Sure, we still need repairmen for these machines, or people to help the machines do their jobs, or people to do some specific work; but the overall number of jobs are going to go down while the population still climbs.


Without some UBI scheme I don't see any happy path forward.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Noisy Rhysling on December 10, 2018, 08:56:48 AM
And how much disposable income will the lumpen proletariat have?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Rai on December 10, 2018, 09:36:51 AM
I wouldn't think so. Capitalism will just find a way to create more bullshit jobs (http://) to keep us employed to be exploited, unfree and unhappy.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on December 10, 2018, 09:53:38 AM
And how much disposable income will the lumpen proletariat have?

They don’t have any disposable income now. It all goes to food and housing and often doesn’t even cover that so they go into debt. Effectively, many have negative disposable income.

At some point they either riot in the street or elect a racist maniac as president. Oh, wait, they’ve already done that.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: stands2reason on December 10, 2018, 10:14:34 AM
The unemployment rate is basically at full employment (with caveats) and the fertility rate is dropping fast. (MISSION DUCKING ACCOMPLISHED)

The only way workers have a competitive advantage is if they are in shortage: in other words, lower the fertility rate. Thankfully this is already happening, and a trend that will hopefully continue. Remember, serfdom basically ended in Europe after the Plague wiped out about 1/3 of the population.

Perhaps the better question is, will AI be able to replace missing laborers fast enough? Over the next couple of decades there will be a very much needed population crash as the Baby Boomer die-off finishes up, and Gen-X starts retiring & dying.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: moj on December 10, 2018, 10:24:32 AM
I would love to see a digital age of leisure. Universal basic income and bots to do all the jobs people don't chose to do. I'm afraid it won't be that idealic, but its what I would like to see. I don't think automation will ever get rid of all human jobs, but a very large number have to the potential to be done but robot or program. UBI will become very important when it happens but fear we won't get it going in time.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: wastrel on December 10, 2018, 11:29:19 AM
I keep glancing at this topic and the though gets put in my head that will.i.am is going to replace us all.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Rai on December 10, 2018, 12:19:02 PM
I keep glancing at this topic and the though gets put in my head that will.i.am is going to replace us all.

He is already a shoddy Wyclef Jean replacement
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on December 10, 2018, 02:41:22 PM
I would love to see a digital age of leisure

(https://i.imgur.com/crXTITz.jpg)
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Ah.hell on December 10, 2018, 03:34:20 PM
One thing I'm sure of is that most predictions will be wrong.

There a few things which are probably true.
A.  Computers and robots will continue to displace human labor, especially low skilled and repetitive work. What kinds of jobs will be at risk will continue to expand.
B.  Some amount of other work will be created that humans will be better suited to than machines(for now anyway).  Where the balance is?  IDK, I doubt anyone else does but there are no shortage of doom sayers nor utopian dreamers. 
C.  The machines will also create a lot of wealth and make things a lot cheaper which will benefit the rest of us along with the owners of said machines.  Again, where the balance ends up, IDK.  I think we'll all be better of but those in power and wealth will be much better of where the rest of us will be a little bit better off.   
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on December 10, 2018, 03:46:21 PM
I'm definitely of two minds about this.  I mean, my whole career is "automation", and I go more in the Steve Novella direction for predictions of AI as "expert systems" and not as "electronic people".

There are some things that an expert system is just straight up better at than a person.  I mean, a PLC is better than a man looking at a pressure gauge as far as alerting about and responding to an alarm condition.  It's just true, and not even the guy who lost his "staring at a gauge" job can argue about it. That guy over there with a pipewrench is safer if the PLC is monitoring the plant than if Staring Guy is monitoring the plant.  And it doesn't make sense to talk about that system "making decisions".  The decisions are all made by a room full of engineers working out control narratives and shut down keys.  The system just responds in the way it was designed by people to do - it is no more "deciding" to shut down a process when the vessel pressure is too high, than a car is "deciding" to stop when the brake is pressed.

Ultimately, it's too broad of a category to just be For or Against.  There are trivial examples where you have to be For, and there are extreme counter examples where you have to be Against.  A lot of the popular media worry about AI is based on movies and SF stories and not really rooted in reality.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Calinthalus on December 10, 2018, 03:49:27 PM
My fear of automation comes from living in the South (specifically Appalachia).  With mining shutting down, all jobs are either customer service, call center, or data entry.  There are a couple of factories still holding on, but when they get better robots they will pay less people.  Unless you can cut in in the health sector (which always does well) you don't have much hope.


Granted, I'm a code monkey writing data entry/OCR software to take peoples jobs away so maybe I'm part of the problem.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on December 10, 2018, 03:53:28 PM
I think it ends up being like a lot of modern problems - you can't look at any one instance and find an issue, it's only in aggregate that there starts to be something to worry about.

Like, it would be very difficult to take a look at any one specific case of automation that costs someone their job, and say "No, we should take out this expert system and put the human back in, even though the human will take longer/be less safe/have lower quality/etc."
Title: Re: Will A.I replace most if not all of our jobs?
Post by: stands2reason on December 10, 2018, 03:58:11 PM
Remember when society collapsed because diesel tractors got rid of those oh-so-rewarding farming jobs?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: SkeptiQueer on December 10, 2018, 04:25:32 PM
Remember when society collapsed because diesel tractors got rid of those oh-so-rewarding farming jobs?

If the tractors gained the ability to manufacture new tractors, plant seed, monitor the growth, optimize fertilization and irrigation, monitor market conditions, harvest the crop, transport the crop, stock the goods for sale, and conduct the sales, accounting, managerial work, optimize and write new code for the next generation, and recycle out of date tractors, then yeah that's a good analogy. Otherwise you're confusing AI for a work-saving machine.

Additionally, the mechanization of labor did lead to labor crises in the South as farm hands sought new work. What new work will we do when machines are also the ones designing new machines, writing code, and performing surgery?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on December 10, 2018, 04:59:19 PM
What new work will we do when machines are also the ones designing new machines, writing code, and performing surgery?

But the surgery one is a perfect example - what individual undergoing surgery is going to say "Nay, do not use this safer, faster, more accurate, better* expert system to save my life!  I demand instead a human surgeon who should have a job, and will perform, at best, similarly!"?

*Assuming such a machine demonstrably existed - if the machine is poor or experimental or inadequate, the example is not in effect.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Noisy Rhysling on December 10, 2018, 07:09:28 PM

Additionally, the mechanization of labor did lead to labor crises in the South as farm hands sought new work. What new work will we do when machines are also the ones designing new machines, writing code, and performing surgery?
The Y-donor stared at the south-end of a north-bound mule instead of going to school after the fifth grade. He said that when the farmers got tractors they farmed more land.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: SkeptiQueer on December 10, 2018, 07:28:12 PM
What new work will we do when machines are also the ones designing new machines, writing code, and performing surgery?

But the surgery one is a perfect example - what individual undergoing surgery is going to say "Nay, do not use this safer, faster, more accurate, better* expert system to save my life!  I demand instead a human surgeon who should have a job, and will perform, at best, similarly!"?

*Assuming such a machine demonstrably existed - if the machine is poor or experimental or inadequate, the example is not in effect.

That's the point I'm making. When farming became mechanized, farm laborers (and freed slaves)oced to cities to work in manufacturing and new tech jobs like typidt and programmer. Stands2reason is saying "well there will be new jobs just like when tractors replaced farm labor" but AI could well replace even formerly skilled labor. Even if we protect jobs, 10 paralegals could be replaced by 1 computer, 3 nurses and a surgical team by a doctor overseeing a machine. The hospital I worked at was replacing custodians with robotic floor cleaners and supply clerks with robotic supply drones to transport material.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Tassie Dave on December 11, 2018, 01:37:56 AM
It was for rioting the against the mechanisation of farming equipment (and the subsequent job losses) in the early part of the 19th century that got one of ancestors sent from England to prison in Tasmania (Van Diemen's Land)
https://en.wikipedia.org/wiki/Swing_Riots

The farm labourers of the time had no other work to go to. Just as those replaced by AI today are not always going to have some other career path.

I can see a not too distant future where goods will turn up at our houses that have had no human involved in their manufacture/growing/sales/packaging/transport/delivery etc



Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on December 11, 2018, 02:10:29 PM
I can see a not too distant future where goods will turn up at our houses that have had no human involved in their manufacture/growing/sales/packaging/transport/delivery etc

I think that was a Black Mirror episode.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on December 11, 2018, 11:17:34 PM
(https://www.smbc-comics.com/comics/1544547685-20181211.png)

http://smbc-comics.com/comic/human-jobs
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on December 12, 2018, 10:12:37 AM
What a stool.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on January 20, 2019, 08:26:01 PM
I suspect there may be more truth to that comic than we'd like to think. Although I'm sure sexbots will replace most of those jobs, too.

But yeah, actually, as someone who studied A.I. about ten years ago and still works in the field, I'm both amazed, excited, and terrified by the progress we've made in the last decade with just incremental technological increases in GPU power and small modifications to existing algorithms. I'm not worried about an A.I. destroying humanity any time soon, but I think there will be major political-economic disruption on a scale that will surpass that of the industrial revolution. This is the start of the first time human workers will become truly obsolete. Of course, some things will be holdouts (like academic work, programming, social work, and the arts), but over time those too will be replaced by superior A.I.s to an increasing extent (I thought maybe art would be a holdout, but look at a website like ostagram, or at the New Rembrandt, or listen to any of the various music producing A.I.s and you start to wonder). I worry that, unless we prepare for this in a major way, this will cause a massive divide between the "haves" who are the last remaining human workers and those who control the A.I.s, and the "have nots", who will be a mass of unemployed people with no source of income or means of survival. And while it may not happen this decade or the next, I think people might overestimate the long-term timescale of this. My guess is I'll see at least the beginnings of this in my lifetime.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on January 20, 2019, 09:16:07 PM
I think my job is safe. After all, you've got to call someone when your AI crashes, right?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on January 20, 2019, 10:13:08 PM
I think my job is safe. After all, you've got to call someone when your AI crashes, right?
For now, but I wouldn't expect that to last forever either (probably quite a while though). People are working on AIs that can create other AIs. And there's research going on in AIs to write more efficient programs. They're not there yet, but I suspect that neural net design will become increasingly automated. As will the diagnostics and maintenance.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Ah.hell on January 21, 2019, 09:11:08 AM
My job is safe as long as engineering write the building codes*, when we start letting programs or programs do that, Then I'll have to worry.


*Building codes tend to include provisions like, "must be signed and stamped by a licensed professional engineer"
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Quetzalcoatl on January 22, 2019, 03:33:30 PM
the poor get poorer.

That's not really true. If you look at the world during the past decades or so, extreme poverty has been massively reduced.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Quetzalcoatl on January 22, 2019, 03:46:32 PM
One thing I'm sure of is that most predictions will be wrong.

I think so too.

There a few things which are probably true.
A.  Computers and robots will continue to displace human labor, especially low skilled and repetitive work. What kinds of jobs will be at risk will continue to expand.
B.  Some amount of other work will be created that humans will be better suited to than machines(for now anyway).  Where the balance is?  IDK, I doubt anyone else does but there are no shortage of doom sayers nor utopian dreamers. 
C.  The machines will also create a lot of wealth and make things a lot cheaper which will benefit the rest of us along with the owners of said machines.  Again, where the balance ends up, IDK.  I think we'll all be better of but those in power and wealth will be much better of where the rest of us will be a little bit better off.

You are probably right. Over here, the trend has been for at least quite some years if not longer, in grocery stores, to have self-serving payouts. I.e, people scan their items by themselves in a machine, instead of going to a cashier the old-fashioned way. However, there still needs to be some human assistance. Some customers need help with the machines. Some items have age-limits, which require human staff to authorize the purchase. But this whole system requires a much smaller staff size compared to the old one with cashiers. Though it is not completely free of human staff.

Maybe this gives a clue to the nearby future? Machines and automation making workplaces requiring vastly smaller human staff forces? I'm not sure some limited forms of AI changes the equation in any substantial way.

Could the assistant human staff in the grocery store eventually be replaces by a robot? Probably. But that might be quite a bit away. A robot, as we know them at least, can't possibly handle all customer requests the way that a human could.

I'm sure similar systems exist in other countries. And of course traditional cashiers have far from disappeared, but the general prediction is that they eventually will, more or less. In Sweden it is very expensive to employ people, which further pushes this trend.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: brilligtove on January 22, 2019, 04:41:06 PM
I'm not sure if this has been mentioned already, but corporations are already big dumb AIs and have been for a loooooong time. They are not smart/conscious in any sense, but they do act like they are alive in many ways. Automation has been a continual process of replacing bits of the corporate organism with parts that are more efficient, more productive, or both. If we were talking about a biological organism it would be the evolution of better enzymes and organelles and such to do things like add abilities or reduce the energy needs of the organism.

Human jobs will continue to shift and change as automation takes over things that used to take a human. This only happens where it is less expensive or more productive to eliminate the human.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Calinthalus on January 22, 2019, 04:47:33 PM
I'm not sure if this has been mentioned already, but corporations are already big dumb AIs and have been for a loooooong time. They are not smart/conscious in any sense, but they do act like they are alive in many ways.
Hell, they already have Free Speech rights and the right to own Congressmen.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on January 22, 2019, 06:00:58 PM
Hell, they already have Free Speech rights and the right to own Congressmen.

The source of all the U.S.'s troubles, in a nutshell.

Well, maybe not all of them.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: John Albert on January 24, 2019, 07:45:09 PM
Hell, they already have Free Speech rights and the right to own Congressmen.

The source of all the U.S.'s troubles, in a nutshell.

Well, maybe not all of them.

Corporations = The People, and Money = speech.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on January 24, 2019, 10:10:35 PM
... Could the assistant human staff in the grocery store eventually be replaces by a robot? Probably. But that might be quite a bit away. A robot, as we know them at least, can't possibly handle all customer requests the way that a human could.

The way corporate managers think, if they ever got a “robot” that could handle 99% of customers’ needs, they’d probably fire the humans and tell the 1% of customers to bugger off, we don’t need your business.

Title: Re: Will A.I replace most if not all of our jobs?
Post by: 2397 on January 27, 2019, 01:25:52 AM
Quarter of U.S. jobs could be jeopardized by AI, research shows (https://www.cbsnews.com/news/quarter-us-jobs-could-be-jeopardized-ai-brookings-institution-artifcial-intelligence/)

Quote
Quarter of U.S. jobs could be jeopardized by AI, research shows

Updated on: January 24, 2019 / 5:55 AM / AP

Robots aren't replacing everyone, but a quarter of U.S. jobs will be severely disrupted as artificial intelligence accelerates the automation of existing work, according to a new Brookings Institution report.

The report, published Thursday, says roughly 36 million Americans hold jobs with "high exposure" to automation - meaning at least 70 percent of their tasks could soon be performed by machines using current technology. Among those most likely to be affected are cooks, waiters and others in food services; short-haul truck drivers; and clerical office workers.

"That population is going to need to upskill, reskill or change jobs fast," said Mark Muro, a senior fellow at Brookings and lead author of the report.

Muro said the timeline for the changes could be "a few years or it could be two decades." But it's likely that automation will happen more swiftly during the next economic downturn. Businesses are typically eager to implement cost-cutting technology as they lay off workers.

This seems to be the source: https://www.brookings.edu/research/automation-and-artificial-intelligence-how-machines-affect-people-and-places/
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 12, 2019, 12:38:27 PM
Found this (https://www.nytimes.com/2019/02/04/business/economy/productivity-inequality-wages.html?action=click&module=Top%20Stories&pgtype=Homepage) article kind of interesting (about how automation is changing the economical landscape). I'll highlight the last bit, because a lot of my colleagues take the stance that the AI revolution is like the industrial revolution (a view I don't agree with), and it echoes some of my own concerns:

Quote
By reducing prices and improving quality, technology was expected to raise demand, which would require more jobs. What’s more, economists thought, more productive workers would have higher incomes. This would create demand for new, unheard-of things that somebody would have to make.

To prove their case, economists pointed confidently to one of the greatest technological leaps of the last few hundred years, when the rural economy gave way to the industrial era.

In 1900, agriculture employed 12 million Americans. By 2014, tractors, combines and other equipment had flushed 10 million people out of the sector. But as farm labor declined, the industrial economy added jobs even faster. What happened? As the new farm machines boosted food production and made produce cheaper, demand for agricultural products grew. And farmers used their higher incomes to purchase newfangled industrial goods.

The new industries were highly productive and also subject to furious technological advancement. Weavers lost their jobs to automated looms; secretaries lost their jobs to Microsoft Windows. But each new spin of the technological wheel, from plastic toys to televisions to computers, yielded higher incomes for workers and more sophisticated products and services for them to buy.

Something different is going on in our current technological revolution. In a new study, David Autor of the Massachusetts Institute of Technology and Anna Salomons of Utrecht University found that over the last 40 years, jobs have fallen in every single industry that introduced technologies to enhance productivity.

The only reason employment didn’t fall across the entire economy is that other industries, with less productivity growth, picked up the slack. “The challenge is not the quantity of jobs,” they wrote. “The challenge is the quality of jobs available to low- and medium-skill workers.”
(emphasis mine)

EDIT: And from the abstract of the study (https://www.brookings.edu/wp-content/uploads/2018/03/1_autorsalomons.pdf):

Quote
Our estimates indicate that the labor share-displacing effects of productivity growth, which were essentially absent in the 1970s, have become more pronounced over time, and are most substantial in the 2000s. This finding is consistent with automation having become in recent decades less labor-augmenting and more labor-displacing.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: superdave on February 19, 2019, 09:05:33 AM
I am an engineer by training and a teacher by profession. 
There are A LOT of people working right now to replace teachers with computer guided curriculum.  It just doesn't seem to work and the big challenge is discipline and motivation.  If you have ever tried an online only course or used the duolingo app to learn a language, you understand the issue.  I do think there is room for AI supported teaching, especially for testing and diagnostic purposes.


Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 19, 2019, 10:15:49 AM
I am an engineer by training and a teacher by profession. 
There are A LOT of people working right now to replace teachers with computer guided curriculum.  It just doesn't seem to work and the big challenge is discipline and motivation.  If you have ever tried an online only course or used the duolingo app to learn a language, you understand the issue.  I do think there is room for AI supported teaching, especially for testing and diagnostic purposes.

When I first decided to try to learn Spanish, I tried several self-teaching methods, ending with a fancy set of CDs. Granted that this was not an interactive course on computer, but I finally realized that I needed an actual teacher to interact with. Until computers can interact with students the way a teacher can, computerized courses will not be as effective as a human teacher. And I don't believe they ever will. Some people can teach themselves some things from a book, but most people need a teacher for most subjects.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 19, 2019, 10:09:26 PM
Here's another article (https://www.abc.net.au/news/2019-02-20/choosing-a-career-these-jobs-wont-go-out-of-style/10828914).

According to the authors,

Quote
Non-routine jobs which need human problem solving, creativity, adaptability, flexibility, physical dexterity, and communication skills will be the jobs of the future.

So will jobs requiring physical proximity and interpersonal skills. Examples include engineering, design, construction, education, health services and care work.

Also,

Quote
Almost two-thirds of employment growth is projected to be in four sectors: health care and social assistance; construction; education and training; and professional, scientific and technical services. Jobs will exist for people with the skills to fill those jobs.

New jobs are projected to be created across a range of occupations. Aged and disabled care, registered nursing, child care, software and applications programming, and waiting are the top five areas of growth.

So I'm in the right industry then, being in a problem-solving, interpersonal-communicating, technical service job. I'm safe.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: superdave on February 20, 2019, 09:18:58 AM
The problem is that as time goes to infinity..the percent of jobs a computer will be able to do goes to 100.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 20, 2019, 08:29:13 PM
The problem is that as time goes to infinity..the percent of jobs a computer will be able to do goes to 100.

I don't think that there's much evidence for that.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on February 20, 2019, 08:35:48 PM
Do you think there's aspects of human consciousness which can't be replicated by software?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 20, 2019, 09:34:45 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do, but I don't think that's relevant to the question as I don't think intelligence is impossible without them. Anyway, there is no such thing as infinite time.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 21, 2019, 08:17:44 AM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on February 21, 2019, 09:37:29 AM
Saaame.

In general...Organization → Information

There exists information which creates life, creates consciousness.  Computers can do information. 

I see no reason to set a boundary between 'what a person is' and 'what a computer can do'
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 21, 2019, 11:46:37 AM
There are, I believe, some things that computers will never be able to do. But there will be a diminishing number of people who rely on services that require humans. And at some point, the Corporation that eventually owns everything will decide that that number is too small to be profitable to service, and will fire the last human workers because the profit they produce is smaller than the wages they are being paid.

Human workers will have become obsolete and unnecessary to the economy.

It won't be "intelligent" machines that take over. It will be the Corporation that decides human workers are unnecessary. At that point, if you don't own enough shares of the Corporation to live off your dividends, you will be left to starve. The last to go will be people in the underground economy servicing the private needs of stockholders, but eventually even they will be replaced.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on February 21, 2019, 12:19:11 PM
I think some people have too firm an opinion on this sort of stuff.  Singularity, AI taking over, etc.  I rate all of it as "plausible but not inevitable".
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on February 21, 2019, 12:22:44 PM
There are [...] some things that computers will never be able to do.

Said the (meat-based) computer.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 21, 2019, 12:36:15 PM
Saaame.

In general...
  • Biology is chemistry organized
  • Psychology is biology organized
Organization → Information

There exists information which creates life, creates consciousness.  Computers can do information. 

I see no reason to set a boundary between 'what a person is' and 'what a computer can do'

Well, unless you're not allowing for computer development in terms of parallelization and such (which I do think we will need more of), but that just seems silly given the developments in the GPU world especially in recent years.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 21, 2019, 04:07:33 PM
There are [...] some things that computers will never be able to do.

Said the (meat-based) computer.

The comparison of people, or of the human brain, to computers is entirely misguided. Brains work nothing like computers, and vice versa. There are a few tasks that both brains and computers can accomplish, but what our brains do is entirely, totally, completely different than what computers do. The processes are completely different, the strengths and weaknesses are completely different, and any notion of any similarity is purely in the realm of sci-fi/fantasy.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 21, 2019, 05:24:55 PM
There are [...] some things that computers will never be able to do.

Said the (meat-based) computer.

The comparison of people, or of the human brain, to computers is entirely misguided. Brains work nothing like computers, and vice versa. There are a few tasks that both brains and computers can accomplish, but what our brains do is entirely, totally, completely different than what computers do. The processes are completely different, the strengths and weaknesses are completely different, and any notion of any similarity is purely in the realm of sci-fi/fantasy.

Of course your current desktop works differently than a brain, but once we understand brains better, we will be able to make computers that work the same way as a brain.  Is there something about a brain that prevents us from modeling it in software and hardware other than that we just don't understand it enough right now?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on February 21, 2019, 05:56:10 PM
The comparison of people, or of the human brain, to computers is entirely misguided. Brains work nothing like computers, and vice versa. There are a few tasks that both brains and computers can accomplish, but what our brains do is entirely, totally, completely different than what computers do. The processes are completely different, the strengths and weaknesses are completely different, and any notion of any similarity is purely in the realm of sci-fi/fantasy.

Where I'd diverge from this reasoning is that a computer doesn't need to look today's computers.

For example, neural networks.  Neural Networks can be their own Turing Complete system and they're based off a rudimentary model of a neuron.  A while back, I read a thing about, "linear threshold units:"
(https://i.imgur.com/4KZi30m.png)

A researcher modeled neurons as those.  Integer inputs from the left.  The neuron sums inputs.  When a threshold is reached, (1) output (threshold) value to the right and (2) retain remainder in neuron.

The researcher showed that a sufficiently large number of those could form a Turing Complete system!  At the time, this was a big deal. 

For example, mechanical computers:
https://www.youtube.com/watch?v=s1i-dnAH9Y4
Recommended Snippet: 5:32 - 7:35

Even if the brain is some kind of completely batshit electrochemical mechanical computer, you could still replicate it through computers.  Even if you end up having to add biological components to do it, which I would not consider cheating.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: CarbShark on February 21, 2019, 06:58:45 PM
I actually worked a job as it was being replaced by advances in technology. Yuk.

For most jobs that are being automated the challenge is not getting the computers to think like humans, it's much simpler (yet still no small task).

The challenges are getting high quality sensory information in a useable way to the processor and having the processor correctly intemperate the data.

These do not require human intelligence.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 21, 2019, 07:36:52 PM
There are [...] some things that computers will never be able to do.

Said the (meat-based) computer.

The comparison of people, or of the human brain, to computers is entirely misguided. Brains work nothing like computers, and vice versa. There are a few tasks that both brains and computers can accomplish, but what our brains do is entirely, totally, completely different than what computers do. The processes are completely different, the strengths and weaknesses are completely different, and any notion of any similarity is purely in the realm of sci-fi/fantasy.

Of course your current desktop works differently than a brain, but once we understand brains better, we will be able to make computers that work the same way as a brain.  Is there something about a brain that prevents us from modeling it in software and hardware other than that we just don't understand it enough right now?

You are predicting that some day we will be able to make computers that work the same way. I dispute this. I do believe that we will be able to build computers capable of doing a wider array of tasks now performed by people. But there is absolutely nothing in common between computers and brains. The dream of a computer that works like a brain is a fantasy probably rooted in the fact that computers can do arithmatic, and do it much faster than a person, so the early ones were often called "mechanical brains."

My prediction is that we will be making significant modifications to human DNA and creating "designer" people a thousand years before we even understand the human brain well enough to attempt to design a computer capable of accurately emulating it.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 21, 2019, 09:33:01 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 23, 2019, 03:23:52 PM
any notion of any similarity is purely in the realm of sci-fi/fantasy.

This is entirely overstated. Artificial neural networks are an abstraction of biological neurons, sure, but in that abstract sense they were inspired by and perform roughly the same kinds of computations that our brains do, albeit in a completely different medium. That's why they're so successful in solving tasks that are typically thought of as "human", like computer vision, or learning language structures. And there are neuromorphic computers (https://en.wikipedia.org/wiki/Neuromorphic_engineering) that are strongly inspired by biological brains. None of that is sci-fi or fantasy.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 23, 2019, 03:25:44 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?

I didn't see anything in there to answer my question, so I don't see how it's relevant?

EDIT: To elaborate: I think that "there's aspects of human consciousness which can't be replicated by software" sounds like an extraordinary claim, regardless of whether you think they are needed for consciousness in general. I'd like to know what you think they are and why they can't be replicated.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 24, 2019, 09:17:48 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?

I didn't see anything in there to answer my question, so I don't see how it's relevant?

EDIT: To elaborate: I think that "there's aspects of human consciousness which can't be replicated by software" sounds like an extraordinary claim, regardless of whether you think they are needed for consciousness in general. I'd like to know what you think they are and why they can't be replicated.

Not that extraordinary. I'm referring to the biological aspects of consciousness - neurotransmitters, axons and dendrites, etc - structures that undoubtedly form part of our own consciousness, but which I don't think are necessary for all consciousnesses.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 25, 2019, 10:07:21 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?

I didn't see anything in there to answer my question, so I don't see how it's relevant?

EDIT: To elaborate: I think that "there's aspects of human consciousness which can't be replicated by software" sounds like an extraordinary claim, regardless of whether you think they are needed for consciousness in general. I'd like to know what you think they are and why they can't be replicated.

Not that extraordinary. I'm referring to the biological aspects of consciousness - neurotransmitters, axons and dendrites, etc - structures that undoubtedly form part of our own consciousness, but which I don't think are necessary for all consciousnesses.

Hm, maybe not literally, but you could simulate those things given enough computational power. At that point, does what difference is left really have any relationship to consciousness at all?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 25, 2019, 11:32:30 PM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?

I didn't see anything in there to answer my question, so I don't see how it's relevant?

EDIT: To elaborate: I think that "there's aspects of human consciousness which can't be replicated by software" sounds like an extraordinary claim, regardless of whether you think they are needed for consciousness in general. I'd like to know what you think they are and why they can't be replicated.

Not that extraordinary. I'm referring to the biological aspects of consciousness - neurotransmitters, axons and dendrites, etc - structures that undoubtedly form part of our own consciousness, but which I don't think are necessary for all consciousnesses.

Hm, maybe not literally, but you could simulate those things given enough computational power. At that point, does what difference is left really have any relationship to consciousness at all?

You could simulate them, yes. But you couldn't replicate them with nonbiological systems. And I don't think you need to.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 26, 2019, 07:11:10 AM
Do you think there's aspects of human consciousness which can't be replicated by software?

Yes, I do

I think that's an extraordinary claim. Could you elaborate?

You do realise that you deleted the relevant part of my quote, right?

I didn't see anything in there to answer my question, so I don't see how it's relevant?

EDIT: To elaborate: I think that "there's aspects of human consciousness which can't be replicated by software" sounds like an extraordinary claim, regardless of whether you think they are needed for consciousness in general. I'd like to know what you think they are and why they can't be replicated.

Not that extraordinary. I'm referring to the biological aspects of consciousness - neurotransmitters, axons and dendrites, etc - structures that undoubtedly form part of our own consciousness, but which I don't think are necessary for all consciousnesses.

Hm, maybe not literally, but you could simulate those things given enough computational power. At that point, does what difference is left really have any relationship to consciousness at all?

You could simulate them, yes. But you couldn't replicate them with nonbiological systems. And I don't think you need to.

I don't think we disagree, but my point is I think we can probably replicate any aspect of the brain that is relevant for consciousness or the mind. But I guess that's mostly arguing over semantics.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 26, 2019, 10:05:09 AM
I would say that even with all we know about brain function, we still do not know enough about it or about consciousness to be able to speculate on whether or not we could replicate it.

I personally do not believe that an electronic computer will ever be able to replicate or even simulate the actual workings of the brain. I do not doubt that someone could write a program that could convince a person that it was a person. But that's not the same thing at all. IOW, I do not believe that the Turing test is proof of consciousness, any more than winning a game of Go.

Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 26, 2019, 11:22:11 AM
I would say that even with all we know about brain function, we still do not know enough about it or about consciousness to be able to speculate on whether or not we could replicate it.

I personally do not believe that an electronic computer will ever be able to replicate or even simulate the actual workings of the brain. I do not doubt that someone could write a program that could convince a person that it was a person. But that's not the same thing at all. IOW, I do not believe that the Turing test is proof of consciousness, any more than winning a game of Go.

Do you think there is something going on in the brain that is not a physical process?  Something spiritual or supernatural?  If that is what you believe, then I see why you hold this opinion.  But, if you believe that the brain is a 100% physical process, then I just don't understand why you don't think we can simulate that in a computer once we understand it at an atomic level.  Perhaps you agree that we could, but that we will never understand the brain with sufficient fidelity?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: moj on February 26, 2019, 11:36:13 AM
I don't want to say never, but think it's much further away then many here are making it out to be to actually create a new consciousness. I think in the mean time AI will get better and take a lot of jobs and create a smaller amount of new jobs but is not going to be aware of itself or anything like that for sometime to come. I think it's possible to figure out and do, I just don't think we are there yet.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: CarbShark on February 26, 2019, 11:55:01 AM
I would say that even with all we know about brain function, we still do not know enough about it or about consciousness to be able to speculate on whether or not we could replicate it.

I personally do not believe that an electronic computer will ever be able to replicate or even simulate the actual workings of the brain. I do not doubt that someone could write a program that could convince a person that it was a person. But that's not the same thing at all. IOW, I do not believe that the Turing test is proof of consciousness, any more than winning a game of Go.

Do you think there is something going on in the brain that is not a physical process?  Something spiritual or supernatural?  If that is what you believe, then I see why you hold this opinion.  But, if you believe that the brain is a 100% physical process, then I just don't understand why you don't think we can simulate that in a computer once we understand it at an atomic level.  Perhaps you agree that we could, but that we will never understand the brain with sufficient fidelity?

There is in the brain an extremely complex interaction of electrical signals and chemical reactions ranging from extremely local to "global." There's interaction and physical responses to hormones and release of hormones which impact other parts of the brain. Some of these interactions occur at the molecular level. Some across small areas, some throughout the brain.  There's also an interplay of the unconscious/conscious brain, and instinctual aspects and theories about "lizard brain" and "mammalian" brain behaviors. And there are parts of the brain dedicated to fight or flight, and responding to stress, pain and pleasure. All of that combine to form HI (human intelligence) and consciousness, and there is no telling which, if any of those are superfluous or required. And there are huge gaps in our understanding of the physiology of thought and intelligence.

I actually don't believe we could ever replicate even a fraction of HI using computers and electricity alone, nor would we need to in order to create devices that act with intelligence. But even that is probably not really necessary. 

For something like driving, the actual AI needed to be able to drive a car in every situation is fairly simple and has been done. What's difficult is getting high quality information from the environment and accurately interpreting the data so the AI knows what's happening and can respond (all in an instant). Being able to determine the difference between a pedestrian that could dart out into traffic and a tree is not an intelligence issue. It's a data and data interpretation issue.

That's the same challenge that or roadblock to AI replacing a huge number of jobs.

Replacing novelists, artists, poets, scientists would be far more difficult than replacing drivers, plumbers, electricians, construction workers or laborers.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 26, 2019, 12:40:11 PM
I don't want to say never, but think it's much further away then many here are making it out to be to actually create a new consciousness. I think in the mean time AI will get better and take a lot of jobs and create a smaller amount of new jobs but is not going to be aware of itself or anything like that for sometime to come. I think it's possible to figure out and do, I just don't think we are there yet.

I think that what I was just suggesting is incredibly far off.  What I think will happen is that we'll try simulating mammalian brains over the next 50 years or so, learning more and more about the actual brains as we reverse engineer them to simulate them in software.  Then, we'll try a human brain based on what we've learned, and it will be all kinds of F'd up.  Then we will iterate that technology over 50 years until we have a pretty damn good grasp on how the human brain works and a pretty good computer simulation of one.  It may take 200 years instead of 100, but I think we'll get human level intelligence at some point out of a computer.

That said, I think this has very little relevance to the topic at hand.  I agree with everyone else that the AI that will become ubiquitous will not have general intelligence.  It will just be good at processing data and deciding what to do.

One thing I often ponder, though, is what affect climate change will have on all this future technology.  If we end up starving, then how important will the newest AI really be?
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 26, 2019, 01:17:56 PM
I would say that even with all we know about brain function, we still do not know enough about it or about consciousness to be able to speculate on whether or not we could replicate it.

I personally do not believe that an electronic computer will ever be able to replicate or even simulate the actual workings of the brain. I do not doubt that someone could write a program that could convince a person that it was a person. But that's not the same thing at all. IOW, I do not believe that the Turing test is proof of consciousness, any more than winning a game of Go.

Do you think there is something going on in the brain that is not a physical process?  Something spiritual or supernatural?  If that is what you believe, then I see why you hold this opinion.  But, if you believe that the brain is a 100% physical process, then I just don't understand why you don't think we can simulate that in a computer once we understand it at an atomic level.  Perhaps you agree that we could, but that we will never understand the brain with sufficient fidelity?

There is in the brain an extremely complex interaction of electrical signals and chemical reactions ranging from extremely local to "global." There's interaction and physical responses to hormones and release of hormones which impact other parts of the brain. Some of these interactions occur at the molecular level. Some across small areas, some throughout the brain.  There's also an interplay of the unconscious/conscious brain, and instinctual aspects and theories about "lizard brain" and "mammalian" brain behaviors. And there are parts of the brain dedicated to fight or flight, and responding to stress, pain and pleasure. All of that combine to form HI (human intelligence) and consciousness, and there is no telling which, if any of those are superfluous or required. And there are huge gaps in our understanding of the physiology of thought and intelligence.

I actually don't believe we could ever replicate even a fraction of HI using computers and electricity alone, nor would we need to in order to create devices that act with intelligence. But even that is probably not really necessary. 

For something like driving, the actual AI needed to be able to drive a car in every situation is fairly simple and has been done. What's difficult is getting high quality information from the environment and accurately interpreting the data so the AI knows what's happening and can respond (all in an instant). Being able to determine the difference between a pedestrian that could dart out into traffic and a tree is not an intelligence issue. It's a data and data interpretation issue.

That's the same challenge that or roadblock to AI replacing a huge number of jobs.

Replacing novelists, artists, poets, scientists would be far more difficult than replacing drivers, plumbers, electricians, construction workers or laborers.

I sometimes disagree with CarbShark, but in the above, he answers the question as well as I could.

I do not believe there is anything supernatural about the brain or human intelligence. But while computers deal in nothing but electrical ones and zeros and are strictly digital, relying on complex algorithms to simulate anything analog, the brain is a chemical system which is highly analog and at least partly chaotic. The transmission of (chemical) signals along a nerve cell may look approximately digital in the string of pulses, but at the synapses a soup of hundreds of transmitter chemicals washes over the next cell to trigger it in a process that at the very least involves chaos and random interactions.

To replicate this using digital technology you would need a computer the size of the galaxy if you used components available today, and then you'd have to shrink it down to the size of my living room in order for the components to communicate with each other fast enough, and since you would need to perform calculations in real time to simultaneously simulate the chaotic interactions at a hundred trillion synapses, your clock speed would probably have to exceed the plank constant.

Accurately simulating such a massively complex analog system in a digital machine would probably require you to break several of the laws of physics. You might be able to build a computer that could simulate a femtosecond of brain activity if you ran it for a hundred thousand years.

We will build DNA to create synthetic people before we replicate the human brain in a computer. And then we will argue whether they should have rights or not. And then, being stronger and smarter and healthier than us, they'll take over. Except I don't believe humanity will survive long enough.

So the short answer is: Not supernatural, just too complex to simulate in a digital computer without breaking the laws of physics.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on February 26, 2019, 01:30:16 PM
Meh, computers don't have to be digital.  They just are right now.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Soldier of FORTRAN on February 26, 2019, 01:32:27 PM
Yeah, this feels a little slippery.

Contemporary tech?  Nah.  Agreed on this point.  edit: ←Human brain.  There's simpler critters out there.

Some unspecified future tech which could be considered a computer?  I don't see why not.  The theoretical basis seems fine to me. 

I'm of the notion that there's no reason to suspect a boundary in general with regards to computers in general.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 26, 2019, 02:28:53 PM
(click to show/hide)

Ah, okay.  We mostly agree then.  We just disagree slightly about the details.  We both agree that with enough processing power, it could be done.  We just disagree if enough processing power will ever be available.  Fair enough!
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 26, 2019, 04:46:53 PM
Meh, computers don't have to be digital.  They just are right now.

Okay. So maybe in the future there will be a technology as yet undreamed of, utilizing principles as yet undiscovered, that will be able to mimic a human brain, and for purposes of linguistic backward compatibility we will call "computers."

And maybe somebody an honest politician will be born.

No. I'm saying that for any mechanical device to perform the necessary computations to simulate a human brain, it would need to violate the laws of physics to be able to simulate more than a millisecond of human thought in a lifetime of calculation.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: brilligtove on February 26, 2019, 05:32:00 PM
I don't want to say never, but think it's much further away then many here are making it out to be to actually create a new consciousness. I think in the mean time AI will get better and take a lot of jobs and create a smaller amount of new jobs but is not going to be aware of itself or anything like that for sometime to come. I think it's possible to figure out and do, I just don't think we are there yet.

I think that what I was just suggesting is incredibly far off.  What I think will happen is that we'll try simulating mammalian brains over the next 50 years or so, learning more and more about the actual brains as we reverse engineer them to simulate them in software.  Then, we'll try a human brain based on what we've learned, and it will be all kinds of F'd up.  Then we will iterate that technology over 50 years until we have a pretty damn good grasp on how the human brain works and a pretty good computer simulation of one.  It may take 200 years instead of 100, but I think we'll get human level intelligence at some point out of a computer.

That said, I think this has very little relevance to the topic at hand.  I agree with everyone else that the AI that will become ubiquitous will not have general intelligence.  It will just be good at processing data and deciding what to do.

One thing I often ponder, though, is what affect climate change will have on all this future technology.  If we end up starving, then how important will the newest AI really be?

Large scale projects to simulate human brains are underway today. They are well aware that they have a lot to learn, and discover new subtlties of neural architecture, synaptic responses, etc., all the time. It won't be a done deal tomorrow, but 20 years is not insanely optimistic.

Meh, computers don't have to be digital.  They just are right now.

Okay. So maybe in the future there will be a technology as yet undreamed of, utilizing principles as yet undiscovered, that will be able to mimic a human brain, and for purposes of linguistic backward compatibility we will call "computers."

And maybe somebody an honest politician will be born.

No. I'm saying that for any mechanical device to perform the necessary computations to simulate a human brain, it would need to violate the laws of physics to be able to simulate more than a millisecond of human thought in a lifetime of calculation.

Daniel, I don't think you understand how large numbers work in this context. The computational power of a human brain is enormous, yes, but that isn't really the issue.

In 2019 the most powerful single computer is 122.3 petaFLOPs. That's 122.3x1015 calculations per second. The Folding@HOME distributed system is in the same ballpark, at about 1/10th of an exoFLOP (which is 1018 calculations per second). The first exoFLOP supercomputer should be built in the early 2020s - likely no later than 2025, maybe as early as 2022.

By some estimates, an exoFLOP is around the computational power of a human brain.

Even if we assume that it will take 1000x as much computational power to simulate a brain than to be a brain - say, a zettaFLOP, or 1021FLOPs - the timeline is still sooner than you think. Historically, we've seen a factor of 10 increase in power every 3 years. That means zettaFLOP power should be here in around 2030. 2040 if you think brains are a million times more powerful and need a yottaFLOP.

The limiting factor, I think, will be understanding the architecture of the brain, not having a computer fast enough to simulate that architecture. There are still huge gaps in our understanding of how neurons work and how they connect, let alone how critical networks of neurons function. Or how other brain cells contribute to cognition. All that might take much longer than it takes to develop the waw power to crunch the numbers.

 
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 26, 2019, 09:00:50 PM
The human brain does calculations very slowly and poorly (inaccurately). That's not the point. The point is the chaotic interplay of chemicals in the synapses.  Computers can follow a rule where X number of 1s verses Y number of 0s produces output Z. But that's not what happens in a brain. To simulate a brain the computer will have to calculate an output based on a chaotic analog soup of hundreds of chemicals each at a different concentation. And it will have to do this calculation simultaneously for a hundred trillion synapses. It would be like trying to calculate with 99% accuracy whether or not it's going to rain on a given square meter of ground, a hundred trillion times per femtosecond and integrating all that together in real time.

You'd need to break the laws of physics to have a machine small enough to be useful (say, smaller than North America) that could do that.

OTOH you might be able to synthesize the DNA that would grow an actual brain. But would that be useful? We already have brains. If your goal is to accomplish tasks that are hard for people to perform, you don't want an artificial brain. You want a computer. The nerds may be trying to figure out how to simulate a brain, but the capitalists are just working on building computers tha can perform human jobs faster and cheaper and good enough to be profitable. They may call it A.I., but what they want is not artificial intelligence. What they want is acceptable performance from a machine that costs less than a human worker. It doesn't even have to do the job as well as a human. It just has to do it well enough, except for critical functions like driving a car or operating air traffic control where it does have to be better. Until the machine economy becomes mature enough that the capitalists don't need workers any more at all and order the machines to grind up the workers for compost. Skynet is not the danger. It's the owners of skynet that will order it to eliminate the surplus population. Though without really good A.I. the capitalists will still have need of one category of workers: sex workers. The rest of us will be expendible.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 26, 2019, 09:16:51 PM
I don't think we disagree, but my point is I think we can probably replicate any aspect of the brain that is relevant for consciousness or the mind. But I guess that's mostly arguing over semantics.

Yeah, adding the "relevant" modifier, we completely agree.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 27, 2019, 01:19:44 PM
The human brain does calculations very slowly and poorly (inaccurately). That's not the point. The point is the chaotic interplay of chemicals in the synapses.  Computers can follow a rule where X number of 1s verses Y number of 0s produces output Z. But that's not what happens in a brain. To simulate a brain the computer will have to calculate an output based on a chaotic analog soup of hundreds of chemicals each at a different concentation. And it will have to do this calculation simultaneously for a hundred trillion synapses. It would be like trying to calculate with 99% accuracy whether or not it's going to rain on a given square meter of ground, a hundred trillion times per femtosecond and integrating all that together in real time.

You'd need to break the laws of physics to have a machine small enough to be useful (say, smaller than North America) that could do that.

I am not sure whether I agree with this or not.  I think it may come down to the detailed level you describe here, but maybe we can find some short cuts and encapsulate some of the parts into more efficient parts.  For example, a neuron may have all that detail that you say, but we may be able to some day figure out all the inputs and outputs of a neuron and which inputs generate which outputs, and then we can simulate that logic straight up rather than simulating all the chemical and electrical stuff at an atomic level.  That would GREATLY simplify the brain simulation.  Then, it may be possible to figure out all the inputs and outputs of a small group of neurons, and we can do the same thing there.  If we are able to do ANY simplification like this at all, then it makes the whole endeavor a lot more possible than you think.

On the other hand, we night need to start with the full simulation like you are proposing.  Let's say they figure out how to do that with a computer the size of North America.  Fifty years after that, that same computing power could fit into a large building.  I agree with brilligtove that computing power is not going to be the limiting factor.  It will take far longer to figure out how brains work.  I imagine we will have a functioning rat brain within our lifetimes.  I can imagine a day when they take a real rat and replace its brain with something that can receive all the inputs and send all the outputs that is connected wirelessly to a giant computer somewhere that does all the brain simulation.  The rat will run around and eat and pee and do all the things a real rat would do.  Obviously, it is a HUGE leap from such a rat brain to a human brain, but one that I think is ultimately doable provided humanity doesn't snuff itself first.

OTOH you might be able to synthesize the DNA that would grow an actual brain. But would that be useful? We already have brains. If your goal is to accomplish tasks that are hard for people to perform, you don't want an artificial brain. You want a computer. The nerds may be trying to figure out how to simulate a brain, but the capitalists are just working on building computers tha can perform human jobs faster and cheaper and good enough to be profitable. They may call it A.I., but what they want is not artificial intelligence. What they want is acceptable performance from a machine that costs less than a human worker. It doesn't even have to do the job as well as a human. It just has to do it well enough, except for critical functions like driving a car or operating air traffic control where it does have to be better. Until the machine economy becomes mature enough that the capitalists don't need workers any more at all and order the machines to grind up the workers for compost. Skynet is not the danger. It's the owners of skynet that will order it to eliminate the surplus population. Though without really good A.I. the capitalists will still have need of one category of workers: sex workers. The rest of us will be expendible.

I agree with all this.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: amysrevenge on February 27, 2019, 01:32:32 PM
I think that's Steve's usual contention about AI.  AI of the golden age SciFi style is kind of useless.  AI as an "artificial human" is an intellectual exercise, but kind of a dumb one.  We already have people, we already know how to make more people, why have artificial people on top of that? 

What we will make are expert systems.  We could, for instance, create a system capable of planning, designing, procuring, building, operating, and maintaining a mass transit system for a city.  Why would it also need to be able to create art or cry at a sunset or learn about the stars or manage a baseball team or or or...?  General intelligence doesn't help.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 27, 2019, 02:26:17 PM
Quote
... a neuron may have all that detail that you say, but we may be able to some day figure out all the inputs and outputs of a neuron and which inputs generate which outputs, and then we can simulate that logic straight up rather than simulating all the chemical and electrical stuff at an atomic level.

The problem is that the output is not a straight-up function of the inputs. The output arises from a chaotic chemical soup. And chaotic systems are inherently impossible to model precisely.

And you cannot build a computer the size of a continent. And in spite of the past rate of advancement in processing power, there are limits imposed by physical laws. The only way we'll ever have artificial intelligence is by re-defining "intelligence."
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 27, 2019, 02:45:45 PM
Quote
... a neuron may have all that detail that you say, but we may be able to some day figure out all the inputs and outputs of a neuron and which inputs generate which outputs, and then we can simulate that logic straight up rather than simulating all the chemical and electrical stuff at an atomic level.

The problem is that the output is not a straight-up function of the inputs. The output arises from a chaotic chemical soup. And chaotic systems are inherently impossible to model precisely.

And you cannot build a computer the size of a continent. And in spite of the past rate of advancement in processing power, there are limits imposed by physical laws. The only way we'll ever have artificial intelligence is by re-defining "intelligence."

What is your take on these organizations?

https://en.wikipedia.org/wiki/Blue_Brain_Project    (https://bluebrain.epfl.ch/)
https://en.wikipedia.org/wiki/Human_Brain_Project    (https://www.humanbrainproject.eu/en/)
Title: Re: Will A.I replace most if not all of our jobs?
Post by: daniel1948 on February 27, 2019, 03:30:53 PM
Quote
... a neuron may have all that detail that you say, but we may be able to some day figure out all the inputs and outputs of a neuron and which inputs generate which outputs, and then we can simulate that logic straight up rather than simulating all the chemical and electrical stuff at an atomic level.

The problem is that the output is not a straight-up function of the inputs. The output arises from a chaotic chemical soup. And chaotic systems are inherently impossible to model precisely.

And you cannot build a computer the size of a continent. And in spite of the past rate of advancement in processing power, there are limits imposed by physical laws. The only way we'll ever have artificial intelligence is by re-defining "intelligence."

What is your take on these organizations?

https://en.wikipedia.org/wiki/Blue_Brain_Project    (https://bluebrain.epfl.ch/)
https://en.wikipedia.org/wiki/Human_Brain_Project    (https://www.humanbrainproject.eu/en/)


Basic research is always good. I expect them to learn a lot of useful stuff about the brain, and I expect that a great deal of good will come of it.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 27, 2019, 06:50:30 PM
I think we're getting too caught up in the idea that you'd need to model the brain atom-by-atom, so to speak. So far, everything we've gleamed from A.I. advances like neural networks suggest we don't actually need to model the complex chemical soup between neurons to create networks that process information in a very brain-like way. What seems to matter is the connection between neurons; what neurons they're connected to, and how strong they activate or inhibit each other (and how that changes as a function of the input), and over what time period (whether through direct synaptic connections or indirect stimulation due to hormone release elsewhere). Right now, our biggest neural networks are about the size of a frog's brain (https://phys.org/news/2018-06-ai-method-power-artificial-neural.html). I remember it was a bee's brain around 2000. Of course, predictions are difficult to make, especially about the future, but given the direction the field is headed in and the developments in hardware architecture and cloud computing, I'm pretty optimistic that I'll see ANNs that rival the human brain in terms of that number of connections in my lifetime. At that point I suspect that there's no substantial computational hurdles to creating a "conscious" or human-level intelligence kind of AI. But I'm guessing we won't have the neuroscientific understanding needed to actually create the right network architecture for something like that for some time. Although with the Human Connectome Project (https://en.wikipedia.org/wiki/Human_Connectome_Project) and other such endeavors, who knows.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: arthwollipot on February 27, 2019, 09:48:49 PM
I think that's Steve's usual contention about AI.  AI of the golden age SciFi style is kind of useless.  AI as an "artificial human" is an intellectual exercise, but kind of a dumb one.  We already have people, we already know how to make more people, why have artificial people on top of that?

The problem with this argument is that someone's going to do it just because they can.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: The Latinist on February 28, 2019, 09:15:05 AM
Every time I see the title of this thread I think it's about Will.i.am.

I think we're getting too caught up in the idea that you'd need to model the brain atom-by-atom, so to speak. So far, everything we've gleamed from A.I. advances like neural networks suggest we don't actually need to model the complex chemical soup between neurons to create networks that process information in a very brain-like way. What seems to matter is the connection between neurons; what neurons they're connected to, and how strong they activate or inhibit each other (and how that changes as a function of the input), and over what time period (whether through direct synaptic connections or indirect stimulation due to hormone release elsewhere).

Exactly this.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: wastrel on February 28, 2019, 12:25:17 PM
Every time I see the title of this thread I think it's about Will.i.am.

:D :D

I keep glancing at this topic and the though gets put in my head that will.i.am is going to replace us all.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on February 28, 2019, 01:35:38 PM
The rate at which we are learning about the brain is, in my opinion, about to change.  With the advent of AIs that can sift through big data, I think the rate at which we gain understanding of the brain is in the process of increasing significantly.  I may have to rethink what I posted a few posts ago about our knowledge of the brain being the limiting factor in creating artificial brains.  Maybe our knowledge of the brain and our ability to compute will increase together, and we will really start making significant progress.

I think one of the complications is going to be that the parts of the brain that we don't really need in a brain simulation (the parts that control balance, heartbeat, pain, etc) may be too intertwined with the parts needed for cognition.  We will either have to figure out how to include all the parts by simulating a body in software, or we'll have to figure out how to modify the brain so that it can cogitate without all those other parts.  It would be great if we could just make a bank of neurons available to a specialized AI to help it do computations and find patterns that it otherwise couldn't.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: brilligtove on February 28, 2019, 03:48:38 PM
The rate at which we are learning about the brain is, in my opinion, about to change.  With the advent of AIs that can sift through big data, I think the rate at which we gain understanding of the brain is in the process of increasing significantly.  I may have to rethink what I posted a few posts ago about our knowledge of the brain being the limiting factor in creating artificial brains.  Maybe our knowledge of the brain and our ability to compute will increase together, and we will really start making significant progress.

I think one of the complications is going to be that the parts of the brain that we don't really need in a brain simulation (the parts that control balance, heartbeat, pain, etc) may be too intertwined with the parts needed for cognition.  We will either have to figure out how to include all the parts by simulating a body in software, or we'll have to figure out how to modify the brain so that it can cogitate without all those other parts.  It would be great if we could just make a bank of neurons available to a specialized AI to help it do computations and find patterns that it otherwise couldn't.

My understanding is that brains need to be embodied, and can not really be considered in isolation from that embodiment. The nature of the body is not important: a car, a human, a network of IOT devices. What matters is that the brain is in constant contact with external events, and adapting to those external inputs.

For creatures like humans, this means a brain that runs a powerful predictive simulation of the physical and social external world, with various self-correcting mechanisms that tend to keep the predictions in line with what actually happens.[1] For a spider? Maybe  there is more value in reacting swiftly? The point is that the type of embodiment dictates what kind of brain functionality is useful to keep the whole system operating.

I'm looking forward to the AIs that are embodied in research laboratories with associated manufacturing capabilities. If one of those is set up to create better computational hardware things will get wild fast.


===
[1]I'm well aware that this is a gross simplification. Go with it for a minute.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on February 28, 2019, 05:46:09 PM
The rate at which we are learning about the brain is, in my opinion, about to change.  With the advent of AIs that can sift through big data, I think the rate at which we gain understanding of the brain is in the process of increasing significantly.  I may have to rethink what I posted a few posts ago about our knowledge of the brain being the limiting factor in creating artificial brains.  Maybe our knowledge of the brain and our ability to compute will increase together, and we will really start making significant progress.

I think one of the complications is going to be that the parts of the brain that we don't really need in a brain simulation (the parts that control balance, heartbeat, pain, etc) may be too intertwined with the parts needed for cognition.  We will either have to figure out how to include all the parts by simulating a body in software, or we'll have to figure out how to modify the brain so that it can cogitate without all those other parts.  It would be great if we could just make a bank of neurons available to a specialized AI to help it do computations and find patterns that it otherwise couldn't.

My understanding is that brains need to be embodied, and can not really be considered in isolation from that embodiment. The nature of the body is not important: a car, a human, a network of IOT devices. What matters is that the brain is in constant contact with external events, and adapting to those external inputs.

I think embodiment certainly gives you more avenues for learning; you can pick things up or move around them to look at them from different angles or you can interact with them to see what happens. And I think our language is ultimately grounded in reality in a pretty fundamental way (which is why I did my MSc thesis on language grounding). But I do think there may be other ways to general intelligence (although an embodied AI may be more relatable and recognizable to us). Ultimately learning in AIs is just doing really complicated statistics. I think an AI that learned from text or databases could maybe learn enough theory about the world that we might say it "understands" it and that it would feel "intelligent", albeit in a way that would probably be far more foreign than an embodied AI.
What I suspect it definitely needs is stimulation, i.e. some sort of input that it can learn from and reason about. You can be born paralyzed, but you can still learn about the world, and I imagine you can have a rich inner life even if you barely ever actively interact with the rest of the world. But for a hypothetical person who does not and has never possessed any senses, I have a hard time imagining that their mind would be anything but completely alien to the rest of us.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: brilligtove on March 01, 2019, 12:35:14 AM
The rate at which we are learning about the brain is, in my opinion, about to change.  With the advent of AIs that can sift through big data, I think the rate at which we gain understanding of the brain is in the process of increasing significantly.  I may have to rethink what I posted a few posts ago about our knowledge of the brain being the limiting factor in creating artificial brains.  Maybe our knowledge of the brain and our ability to compute will increase together, and we will really start making significant progress.

I think one of the complications is going to be that the parts of the brain that we don't really need in a brain simulation (the parts that control balance, heartbeat, pain, etc) may be too intertwined with the parts needed for cognition.  We will either have to figure out how to include all the parts by simulating a body in software, or we'll have to figure out how to modify the brain so that it can cogitate without all those other parts.  It would be great if we could just make a bank of neurons available to a specialized AI to help it do computations and find patterns that it otherwise couldn't.

My understanding is that brains need to be embodied, and can not really be considered in isolation from that embodiment. The nature of the body is not important: a car, a human, a network of IOT devices. What matters is that the brain is in constant contact with external events, and adapting to those external inputs.

I think embodiment certainly gives you more avenues for learning; you can pick things up or move around them to look at them from different angles or you can interact with them to see what happens. And I think our language is ultimately grounded in reality in a pretty fundamental way (which is why I did my MSc thesis on language grounding). But I do think there may be other ways to general intelligence (although an embodied AI may be more relatable and recognizable to us). Ultimately learning in AIs is just doing really complicated statistics. I think an AI that learned from text or databases could maybe learn enough theory about the world that we might say it "understands" it and that it would feel "intelligent", albeit in a way that would probably be far more foreign than an embodied AI.
What I suspect it definitely needs is stimulation, i.e. some sort of input that it can learn from and reason about. You can be born paralyzed, but you can still learn about the world, and I imagine you can have a rich inner life even if you barely ever actively interact with the rest of the world. But for a hypothetical person who does not and has never possessed any senses, I have a hard time imagining that their mind would be anything but completely alien to the rest of us.

Embodiment may be broader than you're thinking. Senses are just intermediation between the brain and the 'real' world. There are senses as we think of them and then there are *sensors* that can interact and report on the world in ways that we can not even imagine.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: werecow on March 01, 2019, 08:30:34 AM
Well usually when we talk about embodied cognition in AI, we're talking about an AI (typically a mobile one, or at least one with multiple sensory modalities like vision or sound) that is sensing the world directly in some way that goes beyond someone typing a question into a keyboard and interacting with databases or the web.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: brilligtove on March 01, 2019, 09:35:37 AM
Well stated. The critical factor is the capacity to directly interact with the real world - that is, the sense must be able to alter the behaviour of the system. Put another way, the system uses the inputs from its senses to alter its behaviour.

In that context, a text I/O interface could be a sense - but not if it is just dumping out data.
Title: Re: Will A.I replace most if not all of our jobs?
Post by: Billzbub on March 01, 2019, 11:45:02 AM
The reason I brought up the whole embodiment thing is just to point out that attempting to simulate and entire mammalian brain will either require embodiment in said mammal, or require the modification of the brain to not require it.  The first will be super creepy and I can't wait to see it happen, and the second will be...I don't know.  Is it easier or harder to just create part of a brain that can be used by someone or something else?  I look forward to seeing how this pans out.