Will the Robots Take Our Jobs? Or Worse?

Fresh off his defeat (according to me!) in the signaling battle, Tyler is on fire in the rematch. Some heavy blows are thrown in our battle over artificial intelligence and the future of jobs. Is there a knockout? You be the judge! Either way, I guarantee a better debate than any you have seen recently.

Comments

I'm with Tyler on this one, Alex's points seemed more like handwaving theoretical concepts, while Tyler had more concrete points. But I write software for a living, and I am watching firsthand as we continue to march software/robot solutions up the IQ curve.

To clarify - let's imagine you have a guy, Joe, who can do 9 things well, and it takes him about three years to learn how to do a new thing well. And you have BoX1, a general purpose robot, who can only do one thing well. But with a legion of software developers behind it, BoX1 can learn to do another thing well every year. By year 15, BoX1 has surpassed Joe, even if Joe has continued to learn new things. What is Joe, and the millions of others like him supposed to do when there's literally no way he can contribute to society that a robot can't do better?

The obvious response is for Joe to learn new skills that BoX1 can't master. But training isn't magic - not everyone can be an architect, or a physicist, or any other "machine-resistant" career. So that particular assumption falls flat to me, especially given what I know about the average person.

"But I write software for a living, and I am watching firsthand as we continue to march software/robot solutions up the IQ curve."

Me too -- and I'm just not seeing that march (at least not at a rapid or accelerating pace). The 'Roomba' has been on the market for going on 15 years. It remains a marginally useful gimmick and isn't remotely close to being able to handle simple problems that low-skilled humans handle easily (opening and closing doors, climbing stairs, picking up stray items, moving furniture). How long before we have household robots that can go around the house to collect stray dishes, load, unload and run the dishwasher, and put clean dishes away in the cupboards? Or collect dirty laundry, wash and dry it, fold it, and put it away? My guess -- a very, *very* long time. Not in my lifetime. Probably not in my kids' lifetimes. And these are considered really low-skilled tasks.

This youtube clip of a robot folding laundry is from 5 years ago. At the end, one of the designers thought they would be ready for wide use, including setting tables and washing the dishes, in 5 to 10 years - so around 2016 to 2021.

https://www.youtube.com/watch?v=5FGVgMsiv1s

This article by Tim B Lee is pretty good, IMO.

http://www.vox.com/new-money/2016/10/4/13161708/burrito-drones-overhyped

Shorter version: Dont believe the hype, we have grabbed a lot of the low hanging fruit and further incremental improvement will take more money and time than will be returned in value.

That robot laundry folder is neat, but how cheap would it have to be for you to buy one?

Yes, and here's what the intervening 5 years have us in state-of-the-art laundry folding:

https://www.youtube.com/watch?v=QztsrEfenf8

A huge, stationary, expensive, single-purpose appliance that folds and stacks a few shirts (but not, apparently, pants, socks, underwear, sheets, etc) provided that you carefully clip them, one at a time, onto the input belt.

The robot in the clip I linked to was a prototype of something that might be available around 2022 while the machine you linked to is expected to be on the market for $800 soon.

2011 nothing
2017 the $800 folder
2022 ?

This seems much sooner than "probably not in my kids' lifetime"stated above.

There's a lot to do with laundry and halfway measures are more trouble than nothing so it's hard to incrementally work your way up.

Make something that will open my dryer, figure if the things there are really dry or need some more spinning, and then fold them all into piles. I'd pay $500 for that.

"The robot in the clip I linked to was a prototype of something that might be available around 2022"

Uh huh. It's kind of standard procedure for people in that researcher's position to envision a commercial product in 5-10 years. It's a time close enough to be intriguing now but also safely far enough out that nobody's really going to remember and come back a decade later to ask what happened to it. Where has the 2011 laundry folder robot gone? Without doing a literature search, my guess is nowhere much. The grad students working on it probably have graduated and the researcher and his current students are doing something else more eye-catching than incremental improvements in laundry folding.

Their website says it will do pants and towels. But yeah, no socks, underwear, or sheets.

Why go to folding clothes? Most people fold their own. That is not a large job segment.

On the other hand manufacturing, shipping, material handling, warehousing .. huge and hugely impacted.

(not sure if Tyler is arguing his, or a more extreme position, but yes as many of us have been saying, smart robots are different in kind from dumb machines.)

Already happened a long time ago. What new advancement in automated manufacturing, shipping, material handling, warehousing is possible today that wasnt possible in 1980? How much of that is practical?

For sure since 1980. We didn't have end to end barcode and automatic routing then. Zip code was still cool.

Im not sure i understand what you're getting at.

The number of times a human hand touches a package has dropped dramatically since 1980, and continues to fall.

That sounds more like marginal improvement in existing systems than some sea change. Also, i question the word "dramatically". Do you have any proof of this? I mean, im sure the efficiency of shipping has improved in the last 36 years, but again, not a paradigm shift so much as a continuation of a trend that has been going on for some time.

Yes, we are talking about long term trends. Robots have been getting slowly smarter, haven't just woke.

Automation will get marginally better in the coming decades. Yawn.

That's the odd thing, you "yawn" at falling employment in traditionally important segments. I could see some sleepiness if these sectors were growing, adding workers, and someone just warned of a future decline, but that is not at all where we are.

The whole segment of "Goods producing, excluding agriculture" has had –2.0% annual job "growth" from 2002 to 2012.

I yawn because that is entirely expected, even in mature industries. Its not the end of the world, it humans getting naturally better at what needs to get done.

Have you seen the Amazon warehouse robots? They are amazing but also a clear indication of how limited robots remain:

https://www.youtube.com/watch?v=quWFjS3Ci7A

They have to move entire shelves all over the warehouse just to get them to a human who picks one item out of a bin and packs it in a box--since robots can't handle that last step.

As above, compare to Sears fulfilment circa 1980.

http://www.wearethepractitioners.com/library/the-practitioner/2016/05/02/wms-the-idea-of-the-distribution-center

"In the fulfillment center of today, computers have replaced the army of people who performed these tasks a century ago. Under close examination, however, each of these steps exists in any e-commerce operation today. In fact, some e-commerce operations still receive orders via the postal service. Over the years telephone and the Internet have added new ways for customers to place an order, but they have not completely replaced the original paper methods. The difference between the Montgomery Ward operations of a century ago and the e-commerce fulfillment center of today is the speed of the process, and the number of people needed to move the process."

"On the other hand manufacturing, shipping, material handling, warehousing .. huge and hugely impacted."

But therein lies the problem. Robotics are feasible in manufacturing only because it's worthwhile to install expensive, inflexible machines to do a particular task (e.g. paint an auto body panel) exactly the same way over and over many thousands of times in a row. And most of the tasks of this nature have already been automated. But tasks requiring flexibility, adaptability, hand-eye coordination -- that's a whole different level.

Actual current events

http://www.pwc.com/us/en/industrial-products/next-manufacturing/robotic-trends-changing-manufacturing.html

https://www.youtube.com/watch?v=dm-QiFNW8sg

Because to learn something new takes a legion of software developers behind it.

What I see is very well funded software/hardware being well done, the legion's cost is extremely high. Outside of that broad slice of mass market stuff is awful; where the legions aren't there, the tasks are very specific and right now the legions are creating solutions that reduce productivity.

When the legions are replaced by robots, then maybe something like you envision will occur.

And you assume a world where learning and knowledge exists somewhere important, when in fact most knowledge of how things work and how they are done exists not where the legions have access to it, but are in the minds and hands of those currently doing it.

What we will likely see is a further reduction in employment in repetitive manufacturing or distribution tasks, partly by robotic, but mostly by the shrinking of offerings because of the extremely high fixed costs to be able to exist in those worlds.

The sticking point is the cost of the legions. There are very few people who can actually generate a productivity advancing piece of software, and they will always be very very expensive. It works because those very few can spread their brilliance over a wide market.

Recently, talking to an elevator mechanic, the Province changed inspection requirements for elevators, doubling the frequency. These are safety tests; a series of tests of common malfunctions and the safeties that prevent injury.

Why? These industries don't move at a torrid rate; an installation will last for decades. The old mechanical relay ladder logic has been replaced by digital systems.

I'd suggest, based on what I see in my field, that the digital stuff is fragile and prone to unknowable failures. Obscure or unknown bugs in almost one of a kind software implementations means that a sensor failure or combination is unpredictable, and regular inspections to test the validity of the inputs is necessary.

So there will be lots of work keeping the robots working. They will require well trained and qualified technicians to service and maintain them. This stuff isn't like a fleet of 787's where they are all the same, the real world is full of short run productions, multiple versioning, etc.

The costs involved because of the limits of availability of trained people will mean less simply gets done.

I don't see a future of abundance, but one of shrinking, of uniformity, of limits. Simply because the cost differential between a drop in mass manufactured gizmo of any size vs. a specific need being satisfied by an on site designed and erected solution will be so high that the second simply won't occur.

" The old mechanical relay ladder logic has been replaced by digital systems.

I’d suggest, based on what I see in my field, that the digital stuff is fragile and prone to unknowable failures."

LOL, no that's not correct. Mechanical relay ladder logic was extremely simple. It was literally one line of current PLC code. I've been in a room that had 20 "rungs" of relay logic on a large wall. It was replaced with 20 lines of code in a PLC. If you kept the functionality the same, then the PLC code isn't anymore prone to failure than the relay ladder logic itself.

Also, relays can and do fail over time. PLC's are built pretty well and can also run for decades. Granted, most PLC's are replaced in a decade or so. But that's almost always driven by a desire for higher levels of production, lower consumption of expendables or safety reasons.

A legion was thousands of men. Many impactful things were created by a handful.

"The obvious response is for Joe to learn new skills that BoX1 can’t master. But training isn’t magic"

Neither is software. People assume that computers/robots/software will simply continue to advance at the rate they have been forever. This has not been the case for anything else and wont be the case for machines either.

With regards to the Roomba, a "solution" may be robot friendly homes with no doors. The world will be designed around robots to get around difficult problems. The "Robot Overlord" scenario is not Skynet sending out Terminator robots, but humans voluntarily making the world friendlier to robots and more hostile to low IQ humans.

Funny. The best way the Roomba kept my living room clean was that it forced me to tidy up enough that it could run every night.

My biweekly maid service has the same effect.

Seems like the answers hinges on the question: "Which jobs?"

Will machines take some? Undoubtedly. But how many? And will that number be what we'd consider "significant"?

There are still lots of jobs that need done, but society has labelled them low-status. Lots of things need to be cleaned or maintained. Environmentalists ought to be really freaked out by the concept of skipping all this work, because it means a lot more waste and trash.

I really enjoyed this, thank you.

Technologies are developed to provide some advantage over what nature provides e.g. a violin is a technology, that reliably provides sounds that are seldom heard in nature. These inventions necessarily create an trade imbalance - we have the violins and you don't - which is transformed into profit for the owner of the technology. Smart machines will just be more of same, except that the owners of the machines will be directly profiting off the failure of human labor to keep up. I do think this is a fundamental difference as compared to other technologies.

What happens when the owners patents on the machines expire?

That was pretty good. I think Tyler under rates human resilience and resourcefulness. Stay tuned.

Tyler has a little bias, machines are as good as the humans behind them. COBOL is not only alive, it still used a lot. https://www.microfocus.com/products/cobol-development Today computers and robots look cheap because it's the very first time they're applied to a problem. But, as technology advances and better solutions are launched, the real challenge is in the update of automation done 20-30 years before without shutting down a service.....such as getting rid of COBOL and using better development environment. Not so long ago, Tyler or Alex linked to this rant on why it takes longer to repair a bridge compared to building a new one https://www.washingtonpost.com/news/wonk/wp/2016/05/26/why-americans-dont-trust-government/?postshare=3951464268844830&tid=ss_tw

Writing new code and developing a new services is easier to adapting old ones to new circumstances, like the Anderson bridge. That could explain why new technologies look so shiny and invincible, but they're not. They require lots of humans behind them. Perhaps 10 humans to develop, but 100 humans to keep running. I'm not software industry insider but it would be good if someone knows this developer to maintenance worker ratio.

Its complicated (some code is flaky, some industries are more stable) but I would guess that it is like 1 hour of maintenance per 10 hours of development. Most code trends toward stability, but most industries replace it for more functionality before that occurs. So something like smartphone software becomes continuous development and continuous bug fixes (in a ratio determined by software quality and/or strategic needs).

When you have minimum wages plus payroll taxes plus ACA mandates that mean a no-skill fry cook will cost $60k/yr in California soon, you'll either see a lot fewer restaurants or a lot more robot fry cooks.

Not only do left-leaning governments love the poor; they love technology!

Nope. You will see fewer restaurants. At least ones that meet mandated standards.

I don't think they worry about such things in the favelas.

Winner, winner, rice and bean dinner!

This.

The very rich will still have dining options, but these will be fine dining. The very poor will never dine out (as per normal). And the rest of us will have to find small owner-operated nutrition delivery systems like food trucks.

There's always been a fair amount of paid-under-the-table work in restaurants and bars, an not just in favelas.

I think Tyler came out slightly ahead in this one, mostly because he was able to draw on more specific information. Alex had strong anecdotes but weak arguments. His inability to wield specific information about the points he was trying to argue undermined the philosophical brushstrokes he was trying to paint with.

With regards to automation, I wish they would've brought up the point that technology tends to automate industries, but not entirely. And the 'not entirely' part is important because the people that are needed to fill that void are usually better paid than people who were doing the rote work.

I'm in the law biz. The unrelenting pressure of "better, faster, cheaper", especially in a stagnant economy, has meant that our business must scrutinize its costs; and our biggest cost is people.

At first, reducing headcount was painful and we feared that it would be disruptive. But now, like James Bond after his first kill, we've learned that each successive round of cuts gets easier. Meanwhile we discover, a bit perversely perhaps, that we can continue to drive profits per partner (PPP) upward. And when PPP is seven figures simple math dictates that next year's new and improved income expectation by the partnership will require even greater efficiency as we try to win new business with the promise of same - by doing more with less; less "full time equivalents" fka "people". We've been doing it for 8 years now; so long that it's become a habit. It's hard to see how we break it, or why we (the survivors anyway) would want to.

"and our biggest cost is people."
I thought your biggest asset was people. You are confusing me.

No, his biggest asset is clients.

So people is his second biggest asset.

One of the places where this debate will play out in the real world is Singapore, where there are over 1 million foreign laborers doing mostly unskilled work like construction and cleaning homes.

The pressure to reduce overcrowding in the country is immense at the same time Singapore is increasingly concerned about terror connections among these workers.

They just introduced a National Robotics Program with $450 million to fund automation and the media frequently have stories about automatic medicine sorters and robot construction inspectors.

And of course the impact on employment will mostly be felt by foreigners so the social impact is lower there than elsewhere.

Most places have different problems than an expensive city filled with high-skill people does.

They are not all 'high skilled'.

Never visited, have you?

" mostly unskilled work like construction"

There's nothing unskilled about construction, except perhaps in the minds of those that find it disgusting to ever break into a sweat at work. Construction workers are object manipulators, who work in the real world. Their nemeses are orthographic conceptualists, who deal in the printed or enpixelated word in an abstract sense. An orthographic conceptualist can come up with the most interesting discovery but it's meaningless until an object manipulator actually physically moves something. All of the brain power that went into the Manhattan Project was just so much low voltage synapse activity until some mechanic actually built the bomb and some other clods delivered it to the target.

"When you work in a modern factory, you are paid, not only for your labor, but for all the productive genius which has made that factory possible: for the work of the industrialist who built it, for the work of the investor who saved the money to risk on the untried and the new, for the work of the engineer who designed the machines of which you are pushing the levers, for the work of the inventor who created the product which you spend your time on making, for the work of the scientist who discovered the laws that went into the making of that product, for the work of the philosopher who taught men how to think and whom your spend your time denouncing.
“The machine, the frozen form of a living intelligence, is the power that expands the potential of your life by raising the productivity of your time. If you worked as a blacksmith in the mystics’ Middle Ages, the whole of your earning capacity would consist of an iron bar produced by your hands in days and days of effort. How many tons of rail do you produce per day if you work for Hank Rearden? Would you dare to claim that the size of your pay cheek was created solely by your physical labor and that those rails were the product of your muscles? The standard of living of that blacksmith is all that your muscles are worth; the rest is a gift from Hank Rearden.
“Every man is free to rise as far as he’s able or willing, but it’s only the degree to which he thinks that determines the degree to which he’ll rise. Physical labor as such can extend no further than the range of the moment. The man who does no more than physical labor, consumes the material value-equivalent of his own contribution to the process of production, and leaves no further value, neither for himself nor others. But the man who produces an idea in any field of rational endeavor-the man who discovers new knowledge-is the permanent benefactor of humanity. Material products can’t be shared, they belong to some ultimate consumer; it Is only the value of an idea that can be shared with unlimited numbers of men, making all sharers richer at no one’s sacrifice or loss, raising the productive capacity of whatever labor they perform. It is the value of his own time that the strong of the intellect transfers to the weak, letting them work on the jobs he discovered, while devoting his time to further discoveries. This is mutual trade to mutual advantage; the interests of the mind are one, no matter what the degree of intelligence, among men who desire to work and don’t seek or expect the unearned." -- Ayn Rand

Wow, every time I read Ayn Rand, I'm amazed at what a terrible writer she was! :-) I don't see how anyone ever finished her books.

But this speech only took 90 pages. And it probably would have sounded better in Russian.

In theory, as robots replace humans the price of robot-produced goods should drop in a way that on net compensates for the loss of jobs. Otherwise, robots would not, on net, be more profitable to operate than humans. That's the theory in a perfect market where there are no artificial monopolies. Of course if the ownership of the robots is limited by patent to a small number of people, those people will reap the majority of the net gains, and inequality will rise.

Imagine this thought experiment: Suppose there was an ultimate machine that produced all the goods and services needed by the whole of humanity in one hour of labor a day by one person. The person that owned the machine would have a kind of God-King status, and would perhaps have one employee who was extremely well compensated (the machine operator). Everyone else would end up becoming servants and entertainers for the God-King and the machine operator. Now, on the pittance made by juggling for their entertainment, they would make more than enough to purchase all of the necessities of life at the vastly reduced price enabled by the machine, so everyone would be better off in absolute terms. But it would be much more *unequal*. It would be a society with extreme differences in status between the top and the bottom of the social hierarchy.

Now imagine that one day the patent on the machine expires. Suddenly (presuming that there are enough people left capable of doing so) a thousand other companies start producing ultimate machines, and in a couple of years everyone has their own personal ultimate machine in their kitchen or basement. No more social hierarchy, or much less of one.

The point of this exercise is that there's a dramatic difference between a society where a small group of people own all the robot technology, which is enabled by patent law, and one where robot technology is radically decentralized.

So... buy land?

You've just describe the feudal world where the land, the productive asset of the economy, was owned by the king, then as book land expanded, feudal lords. That system was broken when individuals could be productive on infertile land and small lots via industries.

This analogy is fundamentally flawed, Hazel.

If there were "an ultimate machine that produced all the goods and services needed by the whole of humanity in one hour of labor a day by one person," then the effective marginal cost of production is zero (assuming an infinite supply or raw materials). In this world, the masses of humanity have nothing to trade with the God-King, because by the very definition you use, the machines produce anything--goods or services--that anyone wants in unlimited quantity.

The God-King therefore has two choices. First, he can conclude he has nothing to trade with ordinary humans, and doesn't, in which case he will be very satisfied alone and his technology will have no effect on commerce otherwise. Think a reclusive Howard Hughes.

On the other hand, since the marginal cost of production approaches zero, the God-King could provide for everyone on a purely altruistic basis. Think King Abdullah of Saudi Arabia during a period of high oil prices. I would guess altruism would be the likely response, with a massive welfare state the result--much as in Saudi Arabia.

That's why everyone would end up being exotic dancers and artists. The God-King doesn't need anything, but a robot can't provide art. Only humans can caper for your amusement. And the God King could easily afford to have a veritable army of servants just because it's more fun to have human servants than robot ones.

Hazel, you respond "but a robot can’t provide art," which contradicts your premise that the ultimate machine produces "all the goods and services needed by the whole of humanity." So you may want to revise that and put some exclusions around services.

This is a degenerate thought experiment. Any actual ultimate machine still needs input of land, raw-materials, energy. The output remains scarce, and there is no reason it wouldn't be sought after by the god-king and his inner-circle (actually the class of people who own the means of production).

"It's huge uply for women."

Game set match Alex.

lol

Yup, in the heat of the moment I misspoke bigly.

... and won the debate.

"uply" was my favorite part.

...that an "we're practically cyborgs already." ;-)

I wish these various debates would stop headlining "robots." Robotics, in the sense of a physical machine moving around, is hard. Algorithms are the real threat to jobs over the next ~10 years. Not because of artificial intelligence, but because of artificial work experience.

I just accept it as a general term, covering 3d printers and LegalZoom.

In kids' vernacular a 'bot is more likely virtual. The language is defined by usage.

As an aside ..

http://shop.evilmadscientist.com/productsmenu/846

Kind of a retro robot writer

Depends on whether people are worthy of the 'jobs'.
Thinking that you are entitled to a 2-bedrm house with a new car every 5 years and 3 weeks of vacation on a job where you scraped through high-school and need only show up 9 - 5, M-F and were able to acquire the skills in a month (manual, manufacturing, clerical, etc.) is the issue. I think that robots/AI should crop off the lowest 25% of the jobs just on principle. That being said, robots/AI should only be mobilized to promote work diversity (remove max 90% of drudgery and facilitate data intensive tasks), opportunity (allow for humans to control/ manage/ create robot/AI tasks (free promotions)), health/welfare support (remaining coal/ gas jobs), and overall economic growth (sprinkle robots/AI in deficient supply areas). That being said, there should be an overall plan for augmenting robot/AI as a strategic commodity from an economy or wide-industry level (subsidies) - not necessarily business by business. Robots/AI should use should promote toward the world we want not be another 'healthy' competition competitor for individuals, businesses, and industries to contend with.

Well, nobody is entitled to anything, but it would be nice if the economy were able to produce enough 2-bedroom houses and cars that everyone could afford such things, even on the income of a high-school graduate with a full-time 9-5 job. What keeping the price of houses and cars high?

In regards to houses ,a word answer: land. Robots will not be making more of it whatever else they do.

There's plenty of land. The problem is it's inaccessible, either because of geography or legal barriers.

"There’s plenty of land." Especially when people no longer need to live where the jobs are.

What people are or aren't entitled too is immaterial. What is material is what those people will do, how they will vote, and how they will act, when they no longer have a healthy balance of work, family, and leisure. Whether you or I like it or not, a substantial set of our population does not have the IQ or the intrinsic motivation to study hard topics and apply themselves. Things like a robust culture and incentive system can move it on the margins, but at the end of the day human ability has a huge variance.

In time, we can provide people with what most consider most of the basic necessities as considered at the time, but those basic necessities always grow. And some needs are potentially unlimited: it's easy to give a hungry person food, but it's much harder to make a sick person healthy.

Alex, Alex, Alex! (And Tyler, too ;-))

You need to get you mind around the implications of something that is changing not just exponentially (compound interest does that), but changing exponentially with a very short doubling time. For example, if a human is 1000 times smarter than a computer, but a computer is getting twice as smart every 1.5 years, when does a computer get as smart as a human, and when does it get 1000 times *smarter* than the human? The answers are, of course, 15 years to be as smart as a human, and 30 years to be 1000 times *smarter* than the human.

So even Tyler is wrong, when he says at 1:52: "This is not an apocalyptic scenario, but it's a slow trickling away of a lot of the jobs that make our lives meaningful."

It's not going to be a "slow trickling away." I'm on record (on my blog) as predicting that more than 50 percent of the jobs in the top 15 job categories in the U.S. will be gone by 2044:

Even Tyler's brushing away of the Terminator scenario is not something that is wise. Think of all the things that humans are vulnerable to (bacteria and viruses, toxic gases, etc.) that machines are not. Once machines have both human intelligence and bodies to move around, there is a real possibility of apocalyptically bad things happening.

The serious predictors (Tetlock) don't like if-then-else predictions, but I don't know how we can avoid them in this case. If "AI" merely advances, it will peel off jobs at some rate. On the other hand, if "AI" actually achieves the acronym, everything will be very different, yes. So if no real "AI" then Tyler, else Bahner.

We're not anywhere near "real AI".

We aren't anywhere near 2044.

"We aren’t anywhere near 2044."

That's less than 30 years. I'd be willing to bet that more than half of the debaters and the people who have commented on this debate will be alive in 2044.

Any economist who thinks that it's likely that the top 15 job categories in the U.S. in 2012 will shed 50 percent of their jobs in less than 30 years should agree that such a change is unprecedented in human history, and therefore should be a matter of extreme interest to the economics profession.

Not the date specifically. I just have no confidence in my own ability to judge claims beyond 5 or 10 years. Questions become nonlinear.

"We’re not anywhere near 'real AI'."

You don't think IBM Watson is "real AI"? It can beat the best human Jeopardy players, it can have a conversation. In fact, it is already better than humans at diagnosing lung cancer:

http://www.businessinsider.com/ibm-watson-cancer-2013-10

Moreover, it's learning at a rate that dwarfs the rates at which humans learn. There are predictions that it will be the "best doctor in the world" in less than a decade. If the "best doctor in the world" isn't "real AI," what is?

No. Watson is not "real AI".
Most people actually mean something like "artificial consciousness" when they talk about "real AI". They aren't talking about a machine that can rattle off facts from an encyclopedia or compute complicated math. They're talking about a machine that can contemplate mortality.

"No. Watson is not 'real AI.' Most people actually mean something like 'artificial consciousness' when they talk about 'real AI.'"

OK. I see we clearly attach separate meanings to "real AI."

If I could go to Watson and tell him:

"I have essentially normal blood pressure (125/75), but I have these episodes of incredibly high blood pressure (like 180/100 or even 220/120). But my heart doesn't race at all during these episodes. Sometimes they're triggered by emotions such as anger of fear, but just as often they have nothing to do with emotions. They just happen. During the episodes, I feel like hell. I really feel like I'm going to die."

...and Watson said, "My good man, it sounds like you have pseudopheochromocytoma," I'd think Watson was a friggin' genius!

And I wouldn't think Watson to be any less of a genius if I then asked, "Have you contemplated your mortality?" and Watson said, "No."

P.S. But I also think that if I can tell my cell phone the name of a doctor and it can give me turn-by-turn directions to that doctor, the cell phone is pretty smart. If a dog or a parakeet could do that, I would think it was a friggin' genius!

If i created a mouth swab that turned blue when exposed to the saliva of people with pseudopheochromocytoma, that would also be friggin amazing, and yet i doubt you would call it AI.

Do the same thing on a computer and suddenly everyone swoons.

If you were trying to find your way to the highway and a series of signs guided you right there, would you think that they were artificial intelligence?

"If i created a mouth swab that turned blue when exposed to the saliva of people with pseudopheochromocytoma, that would also be friggin amazing, and yet i doubt you would call it AI. Do the same thing on a computer and suddenly everyone swoons."

How in the world do you view a computer listening to a patient, and then going through all the millions of potential human health problems, and coming up with an extremely rare disease that probably less than 1 in 50 doctors would come up with, as being the same as a mouth swab that performs a chemical reaction?

"If you were trying to find your way to the highway and a series of signs guided you right there, would you think that they were artificial intelligence?"

Of course not...unless the signs were recognizing my car approaching, and changing just for me.

I think you ought to read more about artificial intelligence. Comprehending human speech and using that comprehension to provide turn-by-turn directions to a destination (and even revising the instructions if the person misses a turn!) is pretty obviously different than unchanging signs posted by a highway.

"How in the world do you view a computer listening to a patient, and then going through all the millions of potential human health problems, and coming up with an extremely rare disease that probably less than 1 in 50 doctors would come up with, as being the same as a mouth swab that performs a chemical reaction?"

Because it told you correctly that you have pseudopheochromocytoma, a disease, i should point out that was not discovered by a computer.

This gets to the real problem with discussing AI, everyone has a different definition of AI. You clearly think that anything that would require a human to think must be AI if performed by a computer. I think that Intelligence is much more than that. Now maybe this would simply be a difference of semantics, but people like you assume that because a computer does something that looks like intelligence and that you define as AI, that therefore, anything that falls under the category of "requires intelligence" will also be possible by AI.

People like me think you are simply assuming your conclusion. You are making no distinction between things that are impressive looking but only really require having a good dataset and the ability to draw new conclusions and insights based on information that isnt already codified somewhere.

"Tress do not grow up to the sky". At some point computers/robots stop getting smarter (and probably the curve flattens out asymptotically before that). The upper bound on the intelligence of anything created by human beings is human intelligence.

Re: Think of all the things that humans are vulnerable to (bacteria and viruses, toxic gases, etc.) that machines are not.

And machines are vulnerable to things human are not. EMPs for example. Or simply immersion in water for any device not specially made for it.

This is incorrect. Computers have been advancing at an exponential rate for a long time, therefore they will continue to do so at the same rate forever.

Actually there are lots of examples of things that were true until they weren't. Evolution took a long time to produce "real AI." Man took a long time to "really fly."

So I think "it hasn't happened yet" is not a good guarantee that it will or won't. Only time will tell. Mark Bahner makes a long range prediction which I can't really judge.

2044?

"Only time will tell. Mark Bahner makes a long range prediction which I can’t really judge. 2044?"

I assume you're asking why 2044, of all possible years. I made the prediction in 2014, which was why 2044. But I made interim predictions within that time period...for 2024 and 2034. For example, I predicted that the number of tractor trailer truck drivers in 2024 would be 1.06 times as many as in 2012. (All values are relative to the year 2012, which was the last year for which I had data in 2014.) For example, my predictions for the number of tractor-trailer truck drivers are:

2024 = 1.06 (i.e., 6 percent *more* tractor trailer truck drivers per unit of U.S. population).
2034 = 0.56 (i.e. a decline of 44 percent in the number of tractor trailer truck drivers per unit of U.S. population, as computer-driven trucks rapidly replace human truck drivers from 2024 to 2034.
2044 = 0.06 (i.e., a decline of 94% in the number of tractor trailer truck drivers per unit of U.S. population, as virtually every tractor trailer truck on the road is computer-driven).

http://markbahner.typepad.com/random_thoughts/2014/11/jobs-vulnerable-to-artificial-intelligence-part-2.html

"Trees do not grow up to the sky.” -->Computers/robots are not trees.

"The upper bound on the intelligence of anything created by human beings is human intelligence."

No, there's already strong evidence that's not so. There's ample evidence that the intelligence of creatures is related to the number of operations per second their brains can perform, and the memory storage capacity of their brains. There is no obvious limit on the number of calculations per second and memory capacity of a computer.

We've also seen in events like the recent "Go" victory of a computer over the best human that we can create devices that can be "smarter" than any human without us even understanding exactly why they do what they do.

"And machines are vulnerable to things human are not. EMPs for example." -->The problem with relying on this for safety is that machines can evolve and reproduce much faster than humans can. And a human requires a good 15+ years before he/she can perform as an adult. A robot comes up to speed much faster.

The social consequences are that robots seem to be first replacing the male factor of production. Historically in America if you were a hard-working man, even if you were not intelligent, you could earn a respectable income to put yourself and your family in a middle-class (relative) lifestyle. This meant these men would work hard, earn money, and be attractive to the opposite sex. This is crucial as young men without jobs or family prospects are incredibly incredibly dangerous for a countries stability.

So even if the path to full automation is slow, we are already seeing the effects of lower returns to non-high IQ male factors of production.

Well, people at the lower ends of the bell curve were always a lumpenproletariat who barely survived and in bad times often did not. Two things have changed: a larger fraction of the lower-intelligence population is being relegated to that status, however we do not allow them to starve or die without medical care.

Tyler won. Alex's point about how in the 1930s people blamed unemployment caused by depression on technology is interesting. During the depression, there was a clear decline in production caused by the idleness of all those workers. If it is automation rather than some other factor causing the decline in working, you'd expect no change or positive change in production. What has been the change in production associated with lower labor force participation rates? Tyler wrote this back in 2010 about "zero marginal product workers:"

Matt Yglesias suggests the notion is implausible, but I am surprised to read those words. Keep in mind, we have had a recovery in output, but not in employment. That means a smaller number of laborers are working, but we are producing as much as before. As a simple first cut, how should we measure the marginal product of those now laid-off workers? I would start with the number zero. If a restored level of output wouldn't count as evidence for the zero marginal product hypothesis, what would? If I ran a business, fired ten people, and output didn't go down, might I start by asking whether those people produced anything useful?

It is true that the ceteris are not paribus, But the observed changes if anything favor the hypothesis of zero marginal product. There has been no major technological breakthrough in the meantime. If anything, there has been bad monetary policy and a dose of regulatory uncertainty. And yet again we can produce just as much without those workers. Think of "labor hoarding" yet without…the hoarding.

http://marginalrevolution.com/marginalrevolution/2010/07/zero-marginal-product-workers.html

Jesus Christ, Tabarrok, buy a new suit. That one makes you look like a poorly built robot yourself.

I liked it.

Given that you're Brazilian, I'd say that this underscores my point.

Why?

Your compatriots are not exactly known for the quality of their dress - at least not the men, that is.

I would be interested in a discussion on the point reconciling the dominance of capital over labor while at the same time seeing increasingly low returns on capital.

Tyler absolutely killed it.

Yes, nothing about free market economics implies that people can adapt quickly.

It is of course widely known that bumble bees can not fly, and men can not go faster than 27 mph, or was it the speed of sound? The same quality of thinking, imho, concludes that human competence has plateaued or that intelligence is limitless. Neither are justified based on the data. We know that our educational system has a submaximal number of graduates, and that those graduates have a submaximal comprehension of the subjects they studied. While those who pay attention know that IQ is a poor predictor of economic or social success, it continues to capture the thought processes of those dealing with humanity's limits. We don't know how much we can teach our children, all we can measure is how much we have taught them. Alex made a good point, that with cell-phones and perhaps AR devices of the near future, information will not be what distinguishes the successful from the unsuccessful. While we have data proving that people as smart as Da Vinci, Newton, Einstein, Goethe can exist (or at least, could exist), there is zero evidence that an IQ of 500 is theoretically possible (if you'll excuse my poetic license). Intelligence quality may be limited by the fact that the world is non-linear, and that many things are chaotic in the formal sense. As far as quantity; bandwidth, memory, and processing rate are obviously finite. (whether the limits are significant or not is another question). With a declining birth rate, the problem seems to me to be to fit the population to the jobs that give the most people the best quality of life, and give none an unacceptable QoL. I note that so far only one post has mentioned the gray and black economies. Aren't criminal enterprises likely to employ humans for a long time? Who's going to pay a robot under the table? I was going to mention the oldest profession, but then I realized that on-line porn has probably displaced some sex workers, and may continue to. As far as hands-on jobs (a pun is intended), I suspect that is another area where humans will have employment for the foreseeable future. How well will we accept face-to-face help from artificial psychotherapists, and can a machine replace most of human touch from professions such as physiotherapy, chiropractic, medicine and bar tending? I have concluded just by looking around that far fewer employees are required than what currently exists. That is, given no new inventions, people will lose jobs. It is hard to argue that enough new jobs will be invented/created to absorb these job losses - in a world where decisions are made by homo economicus. It seems to me that this thread also avoids the issue of underemployment. How many people won't be satisfied if the only way they can earn a living is to work 1 hour a week at 300 - 600 - 1200 $/hr? So even the premise that there will be enough "higher paid" jobs seems to rest on unproven assumptions about what is required for human happiness.fulfillment.satisfaction. Is it really true that the "Average Joe/Josephine" will be able to, or be willing to learn the equivalent of a BS degree every 5 - 10 years? At age 30? At 50? At 70?? I doubt it. The current political mantra, especially from Progressives is "More education", and seems to be an almost exact equivalent to the mantra of abstinence being the best method of birth control. I suspect, arguably, that many kids who have gone to college only to be underemployed after graduation would have been better off had they not gone. My best guess is that society will be recast with a lot more value/status being placed on educational and artistic achievements and volunteerism and less on simply earned income, once the dust has settled. The only alternative I see is a system of technology controls which limit the application of technology to certain jobs. Replace a restaurant's wait staff with wi-fi devices? Nope, against the law. Use a robot to change your oil? Nope - too risky. (Use an unlicensed hair saloon to braid your hair? Nope, not that either. Oh wait, we already have that.) What is certain, imho, is that sooner or later a human level (although very unlikely to be "human-like") will be built. The only thing that would prevent this (aside from our extinction or profound social upheaval) is the existence of some supernatural aspect of the human mind. That is almost certainly not present.

hair saloon? hair salon, lol. although, that gives me an idea...

human level intelligence will be built. sorry for typos.

These robots already have taken away the jobs of many graduate interns.

https://www.techkokobot.com/product/letter-folding-machine-lf283b

Next will be robots for taking orders and making coffees.

TC by a series of TKO's

I was impressed.

I agree with you, Alex. I'd also humbly suggest people read the article on this topic I've written recently. ttps://medium.com/@daniilgor/will-robots-really-create-a-permanent-underclass-81a1b05b9cca#.l9uoadquh

Comments for this post are closed