9 ms·

I just don't buy it. I understand I'm in a minority, so let me indulge myself and explain, for a moment, why I think AI / Robots are nothing to worry about. My

I just don't buy it. I understand I'm in a minority, so let me indulge myself and explain, for a moment, why I think AI / Robots are nothing to worry about. My claim is: Robots will not replace humans in all jobs. Further, it's more than because of Human preference, I think that Robots will lack cognitive abilities that we have. My expectation is that in each area where they are deficient then they may well become a tool for a Human overseer, and that might replace some workers. I don't believe AI will effectively deal with: 1) Cultural concerns. e.g. Appropriate imagery in advertising. 2) Major strategic considerations when faced with an uncertain future. (Physical war, trade war, geopolitical events, natural disasters). 3) Legal frameworks, as they are not well defined. 4) Human relationships. e.g. Don't make Alice and Bob work together. 5) Intuitive leaps. e.g. Rubics cube. 6) Empathy. The reason I believe this is not because I believe in a soul, or anything mystical, but because I believe that the nature of our brains, our senses, our learning experiences, and the evolutionary processes that got us here can not be matched by our technology. I think this means Robots / AI will not independently create new businesses, new products, build factories, kit out factories, buy other robots, produce marketing, deal with customer services, returns, courts, investors, regulators, suppliers and so on. I can imagine products that will exist forever, like, rope, or cutlery, or plates, and someone automating a factory to produce these commodities. But they will still use humans to manage product development, supplier and customer relations, advertising and security, and, probably, looking after the robots. I think that Robots / AI will at best become a tool for people to use, but the power will remain with Human employees.

taberiand 9y ago
Humans aren't so complicated that an AI won't be able to model those things you've listed.
- notahacker 9y ago
  Alternate view: humans are so fucking simple they'll pay extra for a machine-made coffee if it comes with a human attendant. Humans might be so bad at figuring out whether Alice can work with Bob or whether we should bomb Iran that a magic 8 ball could do a better job, never mind the anthropomorphised superintelligent computers which exist only in the imaginations of futurists past and present. But humanity is going to continue to insist on getting people to sort that sort of thing out.
  - beams_of_light 9y ago
    Get the robot! We got a sentient one here, folks!
- lostboys67 9y ago
  ah this is we will have true AI in 5 years which has been said every year since the 1960's - likewise the fusion "to cheap to meter" which has been going since the 1950's
eksemplar 9y ago
I work in business development in the public sector. First we did digitzation, turning paper archives into something digital. Then we started standardizing and connecting the archives while improving worker interaction with them. These days we're doing automation. None of this has led to people getting fired. It has cost a massive amount of jobs though. All unnoticed because we simply didn't rehire for positions when someone retired because the digitization hadn't eaten whole workflows but only 5 minutes here and there. The next thing we're going to do is AI. We've already begun the proofs of concepts. We've gone through 6 million case files with 5-70 page documents, and sorted them in a particular way. A process which too 25 full time jobs half a year to do. Our proof of concept AI and 1000 servers in azure, did the same thing in 5 hours. Computers didn't really change the office landscape like robots did to the assembly lines. AI will. At one point in our history, one of the most important jobs in a large city, was removing the endless tones of horse shit from the streets every day. Cars made those guys as obsolete as AI will do almost everything. Sure there will be new jobs as well, but technology stopped creating more jobs than it ate almost 10 years ago, and it's an exponential curve.
- mdorazio 9y ago
  > we simply didn't rehire for positions when someone retired because the digitization hadn't eaten whole workflows but only 5 minutes here and there I think this is a very important part of automation in the knowledge worker sector that people often miss. Automation doesn't have to get a single person fired to lead to high unemployment, it just has to make managers stop and ask, "do we really need to replace that person who left?". I've been seeing the same thing as you over the last decade with companies I've consulted with where headcount gets pared down here and there every year, and the remaining employees are expected to use more technology tools and wear more hats to fill in the gaps.
- goatlover 9y ago
  > At one point in our history, one of the most important jobs in a large city, was removing the endless tones of horse shit from the streets every day. Cars made those guys as obsolete as AI will do almost everything. Which was replaced by lots of mechanics, factory workers, marketers, office workers and salesmen. A lot of jobs were created by automating away the horse. I wonder how many people predicted that back then? I don't see much evidence yet that AI is going to automate most existing jobs, unless Google, FB, Hitachi, etc is secretly on the cusp of AGI. The current machine learning isn't anywhere near that. This is being blown out of proportion, but it's always been since the beginnings of AI. Maybe someday. Or 20 years from now, along with cold fusion and flying cars.
twoquestions 9y ago
Trouble is, there's a bunch of people out there who have no aptitude for those things, and have neither the desire nor the means to acquire that aptitude, and/or don't have the means to sell this aptitude for enough money for bare necessities. What do they do? This isn't a rhetorical question, I don't know the answer, and I'm kept up at night for the coming storm.
- achamayou 9y ago
  Invest in education rather than robots ;)
  - philipkglass 9y ago
    Even if every barista could learn to write software or practice law, it would just mean that software developers and lawyers would come to be paid like baristas. An individual can rise by educating-up to a kind of job that offers high pay. A whole generation of graduates can't. (Literacy used to be a specialized skill that provided its wielders with better-than-average jobs. Not so much today.)
    - alien_at_work 9y ago
      Further, what if by the time the barista learns a new trade that trade has been eliminated?
    - CM30 9y ago
      > it would just mean that software developers and lawyers would come to be paid like baristas. In quite a few places, the former are basically paid like baristas and office clerks. That certainly seems to be the case at the web development end (and I guess more mainstream side in general) in a lot of places in Europe.
hasbroslasher 9y ago
I'm unsure how you would expect humans to do better with cultural concerns or strategy in the face of uncertainty. We're known for routinely making mistakes in both of these domains - just look at POTUS for a case study. Computers can already infer sentiment, not to mention recognize patterns extremely well, and are likely trainable to provide example feedback to any given cultural current. Is culture really any more diffuse than something like conversation, which computers already do reasonably well? Furthermore, it seems odd to suggest that computers would have trouble in numbers games like war/trade war/resource allocation. These are essentially reducible to games, and with Go out the way, it doesn't seem that figuring out the best tariffs to pass or best places to bomb would require much more logic. Maybe I'm being naive, but I think your conceptions of the limitations of machines are founded on essentially nothing. There's little, philosophically or scientifically, in the way of physicalist/functionalist conceptions of the mind that supports your viewpoints! There's little reason to doubt, in my eyes, that our brains are extremely complex machines, running on electrical impulses and neurotransmitters, to produce responses to stimuli. As such, I don't see why you would privilege our machines over another machine. Why is our learning experience different from that of a machine? Do we not learn from trial-and-error, from explicit (social) rule-programming, from inference? Is evolution really at odds with computers? Some ML algorithms already boast "evolutionary" traits. And are our 5 senses really so sharp? What about beings who "see" all EM radiation or feel rumblings deep in the earth, like a seismograph does?
- arichard123 9y ago
  Our differences may well just come down to a difference of beliefs. In reference to the first 3 paragraphs of your response, I believe there are a few orders of magnitude (at least) difference between "computers can already infer sentiment" to, for example, "a computer that manages a construction firm that uses Mexican commodities realises that Trump has come to power and THAT wall, that is promised, isn't going to happen. Mexico won't be happy, and there might be a trade war. Perhaps start supporting Anti-Trump causes. Certainly don't waste time bidding to build the thing." In reference to the flesh vs machinery question. I think I don't believe it because I don't think we have a good way of designing a good AI. An AI needs to be like a slave that is motivated to serve some goal other than itself. A human brain would not be a good model for such an AI, as our brains are trained to serve ourselves. (Or our selfish genes?) So we cannot copy human brain design and development processes to make an AI that we can trust. We need to design a new brain, based on perhaps similar, but different processes. Evolution with different selectors? Deterministic design? I don't think we could do it.
michaelbuckbee 9y ago
It was some article here on HN a while back that framed this much better for me. It's NOT that Robots are going to take all the jobs, it's that automation and productivity advances are going to reduce the people necessary by a drastic degree. There's lots of stats like Ikea factories making 17x the bookcases they did in 1985 but with the same number of staff. But even in our technical and development niche, look at something like Instagram. They went from idea to billions of users with what a decade ago would have been a laughably small staff. The future is the top 10% of staff executing with an army of automated services at their command to do the work that 90% of their colleagues used to do.
- mooreds 9y ago
  Some might say it's a "Brave New World" :) The real question is what happens to the 90%.
  - CobrastanJorji 9y ago
    The best outcome is the "Culture" endpoint: a post-need society. 10% of people are called upon to work, and the rest thrive on basic income and become artists and inventors, philosophers and musicians and actors. The worst outcome, on the other hand, is rather more dystopian. A society with a very large percentage of young, unemployed people with no particular path to success is not a stable society. I suppose it matters how comfortable the 90% is when they become unemployed. Is basic income enough to fulfill individual passions and interests? What happens to a large body of humans whose needs are taken care of but who aren't strictly needed for anything?
    - germinalphrase 9y ago
      Also, do the 10% believe that they are a static group deserving of persistent participation in the "working class" - or is this a competitive and fluid group in which the children of the 90% may move into/out of as their skills/ability apply? Which is to ask - is it worth maintaining a 90% at all or should we 'solve that problem'?
pdeuchler 9y ago
Reducing employment by 10% would be catastrophic for the US economy and will cause massive social unrest, not to mention the effects in developing countries where people who have just been pulled out of abject poverty in the last generation will suddenly see those opportunities disappear again. Now remember truck drivers are the most populous single occupation in the US, and commercial trucking is expected to be largely automated within 20-30 years by most major shipping companies. We don't need an science fiction style singularity to destroy our world, just a couple of small changes in the fringes will be enough.
- the8472 9y ago
  > Now remember truck drivers are the most populous single occupation in the US Based on non-uniform binning criteria. Bin differently and other jobs might be on the top. Although they might be just as susceptible to automation. It's a nice example, but it is about as solid as pluto's planetary status.
  - mdorazio 9y ago
    If you're interested in non-binned numbers, Business Insider summarized the BLS data [1] a couple years ago. Nationally, the most common occupations are retail and food service workers. You could argue that Amazon is going to put some significant portion of those out of work in the next 20 years. Note that truck drivers are not in the top 10 chart even though the American Trucking Association estimates 3.5 million truck drivers in the US since BLS puts the number at about 1.7 million. I don't know where the discrepancy comes from. [1] http://www.businessinsider.com/most-popular-jobs-in-america-2014-4 http://www.businessinsider.com/most-popular-jobs-in-america-...
georgeecollins 9y ago
It's like they say, Watson can beat any human at Jeopardy. But Watson plus a human can beat any human and Watson by itself.
enraged_camel 9y ago
>>1) Cultural concerns. e.g. Appropriate imagery in advertising. This seems to assume advertising will continue to be aimed at humans. Why would it be? Why would purchasing stuff not be done by AIs in the future as well? In that scenario, things like culture would become irrelevant, as marketing AIs would advertise products/services to purchasing AIs.
letlambda 9y ago
>5) Intuitive leaps. e.g. Rubics cube. I'm not sure what a Rubics cube has to do with intuition. But solving a Rubics cube is so firmly within the realm of doable that no one is going to call it AI anymore.
- arichard123 9y ago
  Sorry, that was an incredibly bad example. I meant the invention of a Rubics cube as a product that might sell, as opposed to it's solution. I thought it was an example of a product that had no similar competitor when it was released but was wildly successful.
visarga 9y ago
> 5) Intuitive leaps. e.g. Rubics cube. You're trolling, right? Intuitive leaps are the bread and butter of neural nets.
- justinjlynn 9y ago
  In fact, one might argue that's all they do -- depending on how one defines "intuition".
  - goatlover 9y ago
    In a vary narrow sense, sure. But humans are great at generalizing across different domains, and finding new connections and patterns.
- arichard123 9y ago
  No, I wasn't trolling. I think we have a different concept of what an intuitive leap is. A neural net, as far as my understanding goes, needs training in the field that it makes decisions in. I think you believe that process itself contains intuitive leaps. Neural nets trained to walk develop weird walking styles, and I think that you think that is evidence of an intuitive leap. If that's what you mean, I would disagree. You train it to walk, and it won't understand how to climb. You train it to climb on a climbing wall and it won't understand that it can't climb on glass, or string, or a waterfall without additional training. Whereas you and I wouldn't need to try and fail to climb a waterfall. We would just know it isn't going to work. We would also understand, without having tested it, that those climbing skills would enable us to climb a tree, or a rope net, or ladder. We might even be able to make a judgement about how well we could climb each before disaster. If I were wrong on this, and I could see an AI trained to do X and then apply it to Y then it would cause me to review my beliefs on AI.
adrianN 9y ago
Machines didn't take all the jobs in agriculture either, but they changed the society from >90% of the population working the fields to <10% what we have today. Imagine what would happen if a similar development would affect all kinds of work. Even if structural unemployment is "just" 40% our current social system would be wholly inadequate.
- goatlover 9y ago
  Yes, but that's a really bad comparison because the net result ended up being a huge increase in new jobs that were created, thanks to the machines. We have tons more people and jobs today.
  - TuringTest 9y ago
    That's because machines needed lots of manual operators to take care of them while working. Now that we have computers, it's not at all clear that this will be true anymore; machines need to be designed, but that requires much less people working on them.
    - goatlover 9y ago
      We've had computers for 70 years now. In order for the machines to seriously cut into available jobs this time around, the software and robots have to be able to do tasks across a wide swath of activities humans are currently doing.
      - TuringTest 9y ago
        You haven't been watching the news recently?
        goatlover 9y ago
        Yeah, but I've also read what people like Rodney Brooks have to say on the matter. And I take tech hype from the media with a healthy dose of salt.
VLM 9y ago
You may want to elaborate on the "can not be matched by our technology" aspect. I'll try. An "80 IQ" manager will fail at the list of six tasks, and a "120 IQ" manager will be a success, by most peoples analysis. Surely, an intelligent but inadequate AI would be mentored like an intelligent but inadequate human manager. We have millennia of experience proving humans can't do that successfully. Therefore if we invent an inadequate AI, regardless of being close to human performance, we know we can't improve performance to make them useful. We play games with re-rolling the dice, but the clock time and human labor time make that too expensive (although it works for mere numerological metric number optimization problems). We play games with pretending that the history of flipping honest coins implies something about the honest coin although it implies more about self delusion and statistics. We play political games where the leader is useless but at least the leader is our guy, our kind of people. Sure, for the sake of argument, I'll take it on faith we can build a mentally disabled AI manager. But we know based on centuries of experience we have no idea how to "fix" a mentally disabled manager into something successful. This is not necessarily a complete black pill. A developmentally disabled AI might, if its merely disabled and not evil, be handy to provide minimal supervision to small children or animals in a safe controlled environment. Its just not going to be very useful. Maybe rephrased, that's nice that some believe we can run a marathon, and maybe we can get part way, but if we absolutely know based on substantial experience we can't upgrade an athlete to run the critical last 100 feet to break the tape and finish, then why bother? Or if we have to build a bridge 1010 feet long if in separate experiments we know we can't extend a bridge past 1000 feet long then it doesn't matter if we have faith or belief that we can build it 10 feet or 950 feet or 999.9 feet because we know once we hit 1000 feet we have no idea how to get it to 1010 feet based on centuries of trying really hard using off the shelf cheap self reproducing biological bridges of 1000 feet length. My greatest fear of AI is it will turn political and we'll no longer be able to talk about the science and technology of AI; it'll just be another dog whistle signal for one side or another, like many other topics "discussed" here on HN. Kool kids will have the right blind faith, the enemy will have the wrong blind faith.

taberiand 9y ago

Humans aren't so complicated that an AI won't be able to model those things you've listed.

notahacker 9y ago
Alternate view: humans are so fucking simple they'll pay extra for a machine-made coffee if it comes with a human attendant. Humans might be so bad at figuring out whether Alice can work with Bob or whether we should bomb Iran that a magic 8 ball could do a better job, never mind the anthropomorphised superintelligent computers which exist only in the imaginations of futurists past and present. But humanity is going to continue to insist on getting people to sort that sort of thing out.
- beams_of_light 9y ago
  Get the robot! We got a sentient one here, folks!
lostboys67 9y ago
ah this is we will have true AI in 5 years which has been said every year since the 1960's - likewise the fusion "to cheap to meter" which has been going since the 1950's

eksemplar 9y ago

I work in business development in the public sector. First we did digitzation, turning paper archives into something digital. Then we started standardizing and connecting the archives while improving worker interaction with them. These days we're doing automation. None of this has led to people getting fired. It has cost a massive amount of jobs though. All unnoticed because we simply didn't rehire for positions when someone retired because the digitization hadn't eaten whole workflows but only 5 minutes here and there. The next thing we're going to do is AI. We've already begun the proofs of concepts. We've gone through 6 million case files with 5-70 page documents, and sorted them in a particular way. A process which too 25 full time jobs half a year to do. Our proof of concept AI and 1000 servers in azure, did the same thing in 5 hours. Computers didn't really change the office landscape like robots did to the assembly lines. AI will. At one point in our history, one of the most important jobs in a large city, was removing the endless tones of horse shit from the streets every day. Cars made those guys as obsolete as AI will do almost everything. Sure there will be new jobs as well, but technology stopped creating more jobs than it ate almost 10 years ago, and it's an exponential curve.

mdorazio 9y ago
> we simply didn't rehire for positions when someone retired because the digitization hadn't eaten whole workflows but only 5 minutes here and there I think this is a very important part of automation in the knowledge worker sector that people often miss. Automation doesn't have to get a single person fired to lead to high unemployment, it just has to make managers stop and ask, "do we really need to replace that person who left?". I've been seeing the same thing as you over the last decade with companies I've consulted with where headcount gets pared down here and there every year, and the remaining employees are expected to use more technology tools and wear more hats to fill in the gaps.
goatlover 9y ago
> At one point in our history, one of the most important jobs in a large city, was removing the endless tones of horse shit from the streets every day. Cars made those guys as obsolete as AI will do almost everything. Which was replaced by lots of mechanics, factory workers, marketers, office workers and salesmen. A lot of jobs were created by automating away the horse. I wonder how many people predicted that back then? I don't see much evidence yet that AI is going to automate most existing jobs, unless Google, FB, Hitachi, etc is secretly on the cusp of AGI. The current machine learning isn't anywhere near that. This is being blown out of proportion, but it's always been since the beginnings of AI. Maybe someday. Or 20 years from now, along with cold fusion and flying cars.

twoquestions 9y ago

Trouble is, there's a bunch of people out there who have no aptitude for those things, and have neither the desire nor the means to acquire that aptitude, and/or don't have the means to sell this aptitude for enough money for bare necessities. What do they do? This isn't a rhetorical question, I don't know the answer, and I'm kept up at night for the coming storm.

achamayou 9y ago
Invest in education rather than robots ;)
- philipkglass 9y ago
  Even if every barista could learn to write software or practice law, it would just mean that software developers and lawyers would come to be paid like baristas. An individual can rise by educating-up to a kind of job that offers high pay. A whole generation of graduates can't. (Literacy used to be a specialized skill that provided its wielders with better-than-average jobs. Not so much today.)
  - alien_at_work 9y ago
    Further, what if by the time the barista learns a new trade that trade has been eliminated?
  - CM30 9y ago
    > it would just mean that software developers and lawyers would come to be paid like baristas. In quite a few places, the former are basically paid like baristas and office clerks. That certainly seems to be the case at the web development end (and I guess more mainstream side in general) in a lot of places in Europe.

hasbroslasher 9y ago

I'm unsure how you would expect humans to do better with cultural concerns or strategy in the face of uncertainty. We're known for routinely making mistakes in both of these domains - just look at POTUS for a case study. Computers can already infer sentiment, not to mention recognize patterns extremely well, and are likely trainable to provide example feedback to any given cultural current. Is culture really any more diffuse than something like conversation, which computers already do reasonably well? Furthermore, it seems odd to suggest that computers would have trouble in numbers games like war/trade war/resource allocation. These are essentially reducible to games, and with Go out the way, it doesn't seem that figuring out the best tariffs to pass or best places to bomb would require much more logic. Maybe I'm being naive, but I think your conceptions of the limitations of machines are founded on essentially nothing. There's little, philosophically or scientifically, in the way of physicalist/functionalist conceptions of the mind that supports your viewpoints! There's little reason to doubt, in my eyes, that our brains are extremely complex machines, running on electrical impulses and neurotransmitters, to produce responses to stimuli. As such, I don't see why you would privilege our machines over another machine. Why is our learning experience different from that of a machine? Do we not learn from trial-and-error, from explicit (social) rule-programming, from inference? Is evolution really at odds with computers? Some ML algorithms already boast "evolutionary" traits. And are our 5 senses really so sharp? What about beings who "see" all EM radiation or feel rumblings deep in the earth, like a seismograph does?

arichard123 9y ago
Our differences may well just come down to a difference of beliefs. In reference to the first 3 paragraphs of your response, I believe there are a few orders of magnitude (at least) difference between "computers can already infer sentiment" to, for example, "a computer that manages a construction firm that uses Mexican commodities realises that Trump has come to power and THAT wall, that is promised, isn't going to happen. Mexico won't be happy, and there might be a trade war. Perhaps start supporting Anti-Trump causes. Certainly don't waste time bidding to build the thing." In reference to the flesh vs machinery question. I think I don't believe it because I don't think we have a good way of designing a good AI. An AI needs to be like a slave that is motivated to serve some goal other than itself. A human brain would not be a good model for such an AI, as our brains are trained to serve ourselves. (Or our selfish genes?) So we cannot copy human brain design and development processes to make an AI that we can trust. We need to design a new brain, based on perhaps similar, but different processes. Evolution with different selectors? Deterministic design? I don't think we could do it.

michaelbuckbee 9y ago

It was some article here on HN a while back that framed this much better for me. It's NOT that Robots are going to take all the jobs, it's that automation and productivity advances are going to reduce the people necessary by a drastic degree. There's lots of stats like Ikea factories making 17x the bookcases they did in 1985 but with the same number of staff. But even in our technical and development niche, look at something like Instagram. They went from idea to billions of users with what a decade ago would have been a laughably small staff. The future is the top 10% of staff executing with an army of automated services at their command to do the work that 90% of their colleagues used to do.

mooreds 9y ago
Some might say it's a "Brave New World" :) The real question is what happens to the 90%.
- CobrastanJorji 9y ago
  The best outcome is the "Culture" endpoint: a post-need society. 10% of people are called upon to work, and the rest thrive on basic income and become artists and inventors, philosophers and musicians and actors. The worst outcome, on the other hand, is rather more dystopian. A society with a very large percentage of young, unemployed people with no particular path to success is not a stable society. I suppose it matters how comfortable the 90% is when they become unemployed. Is basic income enough to fulfill individual passions and interests? What happens to a large body of humans whose needs are taken care of but who aren't strictly needed for anything?
  - germinalphrase 9y ago
    Also, do the 10% believe that they are a static group deserving of persistent participation in the "working class" - or is this a competitive and fluid group in which the children of the 90% may move into/out of as their skills/ability apply? Which is to ask - is it worth maintaining a 90% at all or should we 'solve that problem'?

pdeuchler 9y ago

Reducing employment by 10% would be catastrophic for the US economy and will cause massive social unrest, not to mention the effects in developing countries where people who have just been pulled out of abject poverty in the last generation will suddenly see those opportunities disappear again. Now remember truck drivers are the most populous single occupation in the US, and commercial trucking is expected to be largely automated within 20-30 years by most major shipping companies. We don't need an science fiction style singularity to destroy our world, just a couple of small changes in the fringes will be enough.

the8472 9y ago
> Now remember truck drivers are the most populous single occupation in the US Based on non-uniform binning criteria. Bin differently and other jobs might be on the top. Although they might be just as susceptible to automation. It's a nice example, but it is about as solid as pluto's planetary status.
- mdorazio 9y ago
  If you're interested in non-binned numbers, Business Insider summarized the BLS data [1] a couple years ago. Nationally, the most common occupations are retail and food service workers. You could argue that Amazon is going to put some significant portion of those out of work in the next 20 years. Note that truck drivers are not in the top 10 chart even though the American Trucking Association estimates 3.5 million truck drivers in the US since BLS puts the number at about 1.7 million. I don't know where the discrepancy comes from. [1] http://www.businessinsider.com/most-popular-jobs-in-america-2014-4 http://www.businessinsider.com/most-popular-jobs-in-america-...

georgeecollins 9y ago

It's like they say, Watson can beat any human at Jeopardy. But Watson plus a human can beat any human and Watson by itself.

enraged_camel 9y ago

>>1) Cultural concerns. e.g. Appropriate imagery in advertising. This seems to assume advertising will continue to be aimed at humans. Why would it be? Why would purchasing stuff not be done by AIs in the future as well? In that scenario, things like culture would become irrelevant, as marketing AIs would advertise products/services to purchasing AIs.

letlambda 9y ago

>5) Intuitive leaps. e.g. Rubics cube. I'm not sure what a Rubics cube has to do with intuition. But solving a Rubics cube is so firmly within the realm of doable that no one is going to call it AI anymore.

arichard123 9y ago
Sorry, that was an incredibly bad example. I meant the invention of a Rubics cube as a product that might sell, as opposed to it's solution. I thought it was an example of a product that had no similar competitor when it was released but was wildly successful.

visarga 9y ago

> 5) Intuitive leaps. e.g. Rubics cube. You're trolling, right? Intuitive leaps are the bread and butter of neural nets.

justinjlynn 9y ago
In fact, one might argue that's all they do -- depending on how one defines "intuition".
- goatlover 9y ago
  In a vary narrow sense, sure. But humans are great at generalizing across different domains, and finding new connections and patterns.
arichard123 9y ago
No, I wasn't trolling. I think we have a different concept of what an intuitive leap is. A neural net, as far as my understanding goes, needs training in the field that it makes decisions in. I think you believe that process itself contains intuitive leaps. Neural nets trained to walk develop weird walking styles, and I think that you think that is evidence of an intuitive leap. If that's what you mean, I would disagree. You train it to walk, and it won't understand how to climb. You train it to climb on a climbing wall and it won't understand that it can't climb on glass, or string, or a waterfall without additional training. Whereas you and I wouldn't need to try and fail to climb a waterfall. We would just know it isn't going to work. We would also understand, without having tested it, that those climbing skills would enable us to climb a tree, or a rope net, or ladder. We might even be able to make a judgement about how well we could climb each before disaster. If I were wrong on this, and I could see an AI trained to do X and then apply it to Y then it would cause me to review my beliefs on AI.

adrianN 9y ago

Machines didn't take all the jobs in agriculture either, but they changed the society from >90% of the population working the fields to <10% what we have today. Imagine what would happen if a similar development would affect all kinds of work. Even if structural unemployment is "just" 40% our current social system would be wholly inadequate.

goatlover 9y ago
Yes, but that's a really bad comparison because the net result ended up being a huge increase in new jobs that were created, thanks to the machines. We have tons more people and jobs today.
- TuringTest 9y ago
  That's because machines needed lots of manual operators to take care of them while working. Now that we have computers, it's not at all clear that this will be true anymore; machines need to be designed, but that requires much less people working on them.
  - goatlover 9y ago
    We've had computers for 70 years now. In order for the machines to seriously cut into available jobs this time around, the software and robots have to be able to do tasks across a wide swath of activities humans are currently doing.
    - TuringTest 9y ago
      You haven't been watching the news recently?
      - goatlover 9y ago
        Yeah, but I've also read what people like Rodney Brooks have to say on the matter. And I take tech hype from the media with a healthy dose of salt.

VLM 9y ago

You may want to elaborate on the "can not be matched by our technology" aspect. I'll try. An "80 IQ" manager will fail at the list of six tasks, and a "120 IQ" manager will be a success, by most peoples analysis. Surely, an intelligent but inadequate AI would be mentored like an intelligent but inadequate human manager. We have millennia of experience proving humans can't do that successfully. Therefore if we invent an inadequate AI, regardless of being close to human performance, we know we can't improve performance to make them useful. We play games with re-rolling the dice, but the clock time and human labor time make that too expensive (although it works for mere numerological metric number optimization problems). We play games with pretending that the history of flipping honest coins implies something about the honest coin although it implies more about self delusion and statistics. We play political games where the leader is useless but at least the leader is our guy, our kind of people. Sure, for the sake of argument, I'll take it on faith we can build a mentally disabled AI manager. But we know based on centuries of experience we have no idea how to "fix" a mentally disabled manager into something successful. This is not necessarily a complete black pill. A developmentally disabled AI might, if its merely disabled and not evil, be handy to provide minimal supervision to small children or animals in a safe controlled environment. Its just not going to be very useful. Maybe rephrased, that's nice that some believe we can run a marathon, and maybe we can get part way, but if we absolutely know based on substantial experience we can't upgrade an athlete to run the critical last 100 feet to break the tape and finish, then why bother? Or if we have to build a bridge 1010 feet long if in separate experiments we know we can't extend a bridge past 1000 feet long then it doesn't matter if we have faith or belief that we can build it 10 feet or 950 feet or 999.9 feet because we know once we hit 1000 feet we have no idea how to get it to 1010 feet based on centuries of trying really hard using off the shelf cheap self reproducing biological bridges of 1000 feet length. My greatest fear of AI is it will turn political and we'll no longer be able to talk about the science and technology of AI; it'll just be another dog whistle signal for one side or another, like many other topics "discussed" here on HN. Kool kids will have the right blind faith, the enemy will have the wrong blind faith.