Andrew McAfee: Are droids taking our jobs?

[Tylor]

Robots and algorithms are getting good at jobs like building cars, writing articles, translating -- jobs that once required a human. So what will we humans do for work? Andrew McAfee walks through recent labor data to say: We ain't seen nothing yet. But then he steps back to look at big history, and comes up with a surprising and even thrilling view of what comes next. (Filmed at TEDxBoston.)

Andrew McAfee studies how information technology affects businesses and society.

http://www.ted.com/talks/andrew_mcaf..._our_jobs.html



Thought that this was a very interesting video. He seems very optimistic about overcoming the limitations on human intelligence. For me, it has raised a many questions about how we are going to transition to a new era of wealth as certainly our view of capitalism will most likely not function in this kind of society driven by an artificial intelligence work force.

Being only 18 and just starting college, it makes me worried about what kind of world I will be living in when I graduate and what it will be like by the time I am in my late twenties.

[Loki]

This is the flip side of Luddism. The world isn't going to end in a decade, and neither will you be living on Mars. Technological change is pretty gradual, and advances in AI aren't so great that computers have any shot of replacing humans in the foreseeable future. The way I see it, it might be possible to automate tasks that are menial in nature, and require no thought or intelligence. If that's the kind of job you're going for, you might be in trouble. Otherwise, I really wouldn't worry about it.

[Tylor]

This is the flip side of Luddism. The world isn't going to end in a decade, and neither will you be living on Mars. Technological change is pretty gradual, and advances in AI aren't so great that computers have any shot of replacing humans in the foreseeable future. The way I see it, it might be possible to automate tasks that are menial in nature, and require no thought or intelligence. If that's the kind of job you're going for, you might be in trouble. Otherwise, I really wouldn't worry about it.

Kind of unrelated...
Not just looking this specifically, though, there is no question that things are going to be very different which factors into my fear for what the future may hold. The next twenty years will certainly not be anything like the last. I've been doing a lot of research on the state of the US and Global economy lately and that has me quite worried. Not to mention government has gotten seemingly more corrupt, some people will watch Fox News or MSNBC as their only source of news, there is a growing dependency on government entitlements, and a large number of other observations I have been making about modern society, government, and our current economy.

[Loki]

If society is going to hell, you probably don't have to worry about technological innovation leading to the loss of your job. As for the current state of society, it's no worse than it was during previous crises, most notably the staglation of the '70s and early '80s. There's nothing happening that we haven't seen before. It is the arrogance of every era to assume that it's fundamentally different to all that came before. Your criticisms could be and have been made by members of every generation going back to the dawn of civilization.

[Steely Glint]

For me, it has raised a many questions about how we are going to transition to a new era of wealth as certainly our view of capitalism will most likely not function in this kind of society driven by an artificial intelligence work force.

Artificial intelligence has been promising great breakthroughs for decades, but has achieved little. Turns out, the problem is actually really hard. As Loki says, we will increasing be able to automate menial tasks but I don't see us being able to build machines that can replicate skilled work in any of our lifetimes. And when we can, it will be a change to civilization and our species so profound that worrying about having a job is a bit like a hunter gatherer worrying about where he'll find berries and elk the middle of a modern city.

[Wraith]

As Loki says, we will increasing be able to automate menial tasks but I don't see us being able to build machines that can replicate skilled work in any of our lifetimes.

Software is already capable of more skilled tasks than most people know; one of the problems faced is that people have a far lower tolerance for computer mistakes than human ones (I.E. if a skilled person averages 92% accuracy, and a computer is 98% accurate, the person will still be prefered for the job). It's not a universal bias, but this shows up in a lot of weird places.

That's a bit of a digression though. In the field of general AI, it's hard to predict when the advances will come, but once they do they'll come fast. As Moore's law progresses, the amount of cleverness we need to make advances in the field shrinks, and we only need one good breakthrough. I think "not in our lifetimes" is extremely pessimistic. I'd put it at 10 years optimistically, 25 years more realistically, 40ish at the outside.

[Steely Glint]

I think the barriers to making a strong AI are conceptual rather than hardware based, so the breakthrough could come tomorrow or in 200 years.

[Wraith]

I think the barriers to making a strong AI are conceptual rather than hardware based, so the breakthrough could come tomorrow or in 200 years.

There's always the brute force method available, it's just not feasible yet. At the risk of sounding like Alber, simulating an entire human brain will eventually be a possibility. We won't really need anything but the most basic understanding of how it works (which we mostly already have), we can just run it and go from there. IMO, that's not a very good approach, but it's one that'll be available to us eventually if we can't come up with anything more elegant before then.

Although, that is assuming that Moore's law doesn't collapse completely - which is a real possibility.

[Steely Glint]

The idea is to come up with something better than a human, though. If we're just going to simulate a human brain, why not just use a real person? They're pretty easy to make. A computer brain would obviously be much faster than a human one, but there's more to generating a being which can do useful work than just having a functioning facsimile of a brain.

[Tylor]

I may be misinterpreting what you're saying but, to use an example that was discussed in the video, it is a real possibility that truck drivers may lose their jobs to AI in the somewhat near future. Google already has a car out on public roads that drives itself perfectly fine. The advantage to using AI rather than a human is that you won't have to pay the AI to drive the truck(along with any other benefits a trucking company may provide, I'm not in the trucking industry so I wouldn't know), the AI doesn't need sleep or breaks, it has the potential to be a much safer driver than a human and can drive more efficiently. Of course, this can apply to many other jobs as well and not just the trucking industry.

There would be a monetary benefit to using AI instead of a human.

[Loki]

Surely truck driving qualifies under the menial labor exception I mentioned above?

Did the unemployment rate go up when computers replaced most secretaries? Or when cars replaced horse buggies?

[Steely Glint]

I think lorry driver falls under the category of "menial worker" that Loki talked about.

Consider also that a new technology may create jobs in ways we cannot foresee. For example, the computer was originally devised as a method for automating the process of, well, doing computations. For most of history 'computer' was a job title. Then, clever people figured out how to build machines to do that, faster and better. I guess those human computers were out of the job. Half a century has it was invented, the digital computer had spawned entire industries. How many people work in IT in some fashion around the globe day? How would they explain what they do for a living to someone who lived before the PC? Before the web? They couldn't.

TL: DR: we come up with new technology to make work easier and then think of all kinds of new things we could do for it and in the process create a bunch more work.

[GGT]

This is the flip side of Luddism. The world isn't going to end in a decade, and neither will you be living on Mars. Technological change is pretty gradual, and advances in AI aren't so great that computers have any shot of replacing humans in the foreseeable future. The way I see it, it might be possible to automate tasks that are menial in nature, and require no thought or intelligence. If that's the kind of job you're going for, you might be in trouble. Otherwise, I really wouldn't worry about it.

Tech changes are faster than you recognize or admit. Because your career path hasn't been replaced by AI or IT yet. But you are, make no bones about it, easily replaceable.

If society is going to hell, you probably don't have to worry about technological innovation leading to the loss of your job. As for the current state of society, it's no worse than it was during previous crises, most notably the staglation of the '70s and early '80s. There's nothing happening that we haven't seen before. It is the arrogance of every era to assume that it's fundamentally different to all that came before. Your criticisms could be and have been made by members of every generation going back to the dawn of civilization.

Tylor posed a great question, on a great topic. Stop replying like a know-it-all-professor, and admit you don't know it all. BTW, there IS something happening now that we've never seen before. It won't work to compare today's challenges to the 70s or 80s. There IS a "new normal". If there's any arrogance going on here, it's from you...and the myopic elitist pov from academics in your peer group, that dare to say criticisms aren't worthy, because criticisms have been made since the dawn of time. And regular folk just don't understand. FFS.

Surely truck driving qualifies under the menial labor exception I mentioned above?

Did the unemployment rate go up when computers replaced most secretaries? Or when cars replaced horse buggies?

Yes, that was a matter of fact. It was temporary, of course. We now live in an era when every new computer application can do the work of ten.twenty.thousands of actual people. An age when every new Apple iPhone iteration raises our GDP by half of one percent, without translating into more domestic jobs.

More importantly, we live in the new age of droids and drones doing the dirty work of War. So "real people" don't have to set foot on foreign soil, or risk losing their limbs or life to hidden bombs, right? Would that include the truck driving soldier and military personnel transporting fuel or IT hardware? Is that still "menial work" in your mind?

[Wraith]

The idea is to come up with something better than a human, though. If we're just going to simulate a human brain, why not just use a real person? They're pretty easy to make. A computer brain would obviously be much faster than a human one, but there's more to generating a being which can do useful work than just having a functioning facsimile of a brain.

You start there, it's easier to make improvements on it. At the start, such a simulated person wouldn't need sleep, and could potentially think faster than an actual person. Once you have that, it's easier to gain a greater understanding of how the whole system works, and to figure out improvements you can make.

Surely truck driving qualifies under the menial labor exception I mentioned above?

Interestingly, the driving problem is actually a harder problem than a lot of the 'non-menial' labor you're probably thinking of.

[Loki]

It's still a computing problem. If x, then y. If y, then z. Etc. It doesn't require it to create anything new.

[GGT]

Would you like to play a game?

Game theory is a crappy way to improve the Human Condition.

How about global thermonuclear war?

[Lewkowski]

I think people use advancing technology as a scapegoat for their own unemployment. Fact is we have more people employed today then we did in the past. Fact is that advancements in technology provide a lift in the quality of life. Quit worrying and enjoy your iPhone.

[wiggin]

There's always the brute force method available, it's just not feasible yet. At the risk of sounding like Alber, simulating an entire human brain will eventually be a possibility. We won't really need anything but the most basic understanding of how it works (which we mostly already have), we can just run it and go from there. IMO, that's not a very good approach, but it's one that'll be available to us eventually if we can't come up with anything more elegant before then.

Although, that is assuming that Moore's law doesn't collapse completely - which is a real possibility.

From a neuroscience point of view, I think you're wrong, Wraith. There's a lot we don't know about how the brain works (let alone complex functions like cognition and problem solving), and even cracking that problem will likely take decades if not longer. We don't even have all of the tools to do it yet.

I've been friends for a long time with a chap who was way more Alber than Alber (Alber actually had several links in his sig to organizations this friend helps run). He's a true believer and is working diligently to get there for the last 15 years or so. Yet in reality they haven't really gotten anywhere.

This isn't to say that it's not possible in the time frame you suggest - after all, I do buy into the basic idea of a singularity, so it suggests that development of world-changing AI will likely happen in a very compressed period. But the technical challenges to getting there are very significant. Raw computing power isn't going to do it (I don't buy any of that nonsense about emergent consciousness in sufficiently sophisticated networks); you need some very clever design with a deep understanding of cognitive science. I'm not even sure we have the computer language to express the necessary functions.

[Tylor]

Surely truck driving qualifies under the menial labor exception I mentioned above?

Did the unemployment rate go up when computers replaced most secretaries? Or when cars replaced horse buggies?

I meant to respond to Steely Glint's Q of why we would rather just use an AI instead of a human. Either way when the 3.5 million truck drivers in the US lose their jobs, then what? As we enter an era where AI can surpass our own intelligence there will unquestionably be a large change in employment and I don't understand how there wouldn't be. Labor and work will become more rare as truckers/factory workers are replaced and eventually knowledge workers as AI becomes stronger. At the rate that technology is growing, this is entirely possible within my lifetime and I am sure it will start to show in 15-20 years from now.

[Loki]

Think of it like this, what would have to happen for 3.5 million truck drivers to lose their jobs? At the very least, millions of specialized trucks would have to be built. Who's going to build those trucks? Who's going to mine and process the metals that will allow others to build those trucks? Who will market those trucks? Repair them? More importantly, who will design them, who will come up with new models? Last I checked, these all produce new jobs, many of which the truck drivers would be qualified to obtain.

[Tylor]

We've already got computers and robots taking the place of people in car factories. In 5-10 years these robotics will be far more improved than they are now, looking at how quickly technology has grown in just the past few years. Very little human workers, if any at all, will be needed to build them.

Who will market these? I haven't seen any commercials or advertisements for semis in my short 18 years of life. And with the technology of today's internet and constant connection to news this won't need much, if any, marketing. Any trucking company will want automated trucks over a human driven one so any whiff of this technology that can decrease their expenses immensely will catch their attention.

Repair them? If we have robots building these trucks, they can repair them too. Technology for identifying what part of a car needs maintenance is already out there. My grandmother's BMW has this technology and she's had it for 4 years.

Who will design them? The technology for driving these trucks is already being worked on and researched. The models? People had to design current models so I'm sure the same people will be working on the future design.

Mining these materials for the trucks won't bring in 3.5 million more jobs. And this is only the trucking industry.

There are stock market programs that trade investments on a news and a chart based algorithm. In the video, a short news article on a company's earnings was written by a computer and you wouldn't even be able to tell the difference.

[Wraith]

From a neuroscience point of view, I think you're wrong, Wraith. There's a lot we don't know about how the brain works (let alone complex functions like cognition and problem solving), and even cracking that problem will likely take decades if not longer. We don't even have all of the tools to do it yet.

That's the thing though; you don't need to know everything about how the brain works. You don't even need a high-level understanding at all. Just the very basic stuff; how neurons and synapses work. You can gain the bigger stuff later and more easily once you start just start simulating all the pieces together. At that point, it becomes easy to quickly test hypothesis, work out what's happening, and make adjustments to improve fidelity, or just improve beyond fidelity.

I think the better route is to skip trying to mimic evolved neurology entirely, but if we don't work that out, my point was just that eventually raw computing power will make it viable and very likely easier than coming up with clever algorithms.

[Nessus]

It's still a computing problem. If x, then y. If y, then z. Etc. It doesn't require it to create anything new.

How much experience do you have working with computer programming? I'm not trying to assault your person here, but my working experience (which isn't programming-intensive, I just have to do some stuff occasionally and have a gut feeling about what computers do) makes me agree with Wraith's statement that automation is already available for a lot of things Joe Six-Pack wouldn't think of. It's not exactly intuitive which things are doable for computers and which aren't; I'm willing to make a bet dollars to donuts that your government is researching things that'd facilitate the manufacture of automated drones that can differentiate between US troops and sand niggers. (If these don't already exist; I'm under the impression the drones are remote-controlled right now?)

I don't remember if I ever linked it here, but Charlie Stross has a good blog post about a potential future where the main job market is for security related tasks; human beings monitoring other human beings, both inside and outside prisons. Mostly inside, however, as we don't really have the ability to manufacture prisoner-beating robots, but further automating the monitoring of the public is probably feasible within a relatively short time-frame.

Even creative jobs can be automated based on your argument there; we can monitor fairly closely what the consuming populace desires (this is becoming easier and easier as we're moving away from mass transmissions to direct product purchasing), and it probably isn't that hard to make an MCMC algorithm based content generator. Like Steely said, 'computer' used to be a job for a person; picture Orwell's MiniTruth except mostly staffed by computer algorithms.

Disclaimer: I'm placing all of this roughly around Wraith's stipulated time-frame, maybe a bit more pessimistic than his but in that ball-park. Innovation is driven by lazy people, and the corollary is that innovation always obsoletes people. Eventually we'll have obsoleted a huge chunk of humanity in the Western World, do we put them back into poverty (and in some cases slavery) along with the third world, or what?

Speaking of menial tasks, sexual pleasure via another actor probably won't be automated any time soon compared to these other things (real dolls are fairly Uncanny Valley material, and despite the efforts of many Japanese engineers, making them motile and capable of basic dialogue is way beyond us), so whoring could be a fairly solid career path down the line. Of course two prostitutes won't really profit purchasing goods from one another, but you're better at this invisible hand than I so you can figure out what that economy would look like And if we return to a society with slaves, hunting handling and training slaves for sale would be an industry that could employ a lot of people; presumably there'd be demand for human slaves for 'irrational' reasons such as sadism, ego, and sexual desires. And by introducing slavery, we also reduce the pressure to automate everything, see the Greeks and Heron's device.

[Steely Glint]

Why would we return to a society of slaves if we're going to automate everything under the sun?

[Nessus]

As I said, see Heron's device. We can only consign so many as Lebensunwertes leben before society collapses, much easier to have a helot system

[Wraith]

It's still a computing problem. If x, then y. If y, then z. Etc. It doesn't require it to create anything new.

I somehow missed this post. It is not anywhere near as simple as a series of if/then/else statements. It is a much more complicated problem than you think.

[Loki]

I don't think it's simple - if it was, it would be implemented by now. But I also don't think it requires the AI to produce new knowledge.

Nessus, you seem to completely ignoring the cost dimension. Just because a robot can conceivably do something as well as a person doesn't mean that it makes any financial sense to replace the person with a robot.

[Nessus]

Nessus, you seem to completely ignoring the cost dimension. Just because a robot can conceivably do something as well as a person doesn't mean that it makes any financial sense to replace the person with a robot.

Hence why a re-surgence of slavery seems pertinent. All the same, the goal of any capitalist ought be workerless factories, non?

[Steely Glint]

If cost were the only thing preventing a re-resurgence of slavery it wouldn't have disappeared in the first place.

[Loki]

Hence why a re-surgence of slavery seems pertinent. All the same, the goal of any capitalist ought be workerless factories, non?

Why? The goal is a ratio of capital to labor that's based on the marginal cost of one relative to the other.