6 ms·

I have frequently wondered if the lag in productivity is partly due to the distracting power of the internet and social media. This isn't really paralleled by t

I have frequently wondered if the lag in productivity is partly due to the distracting power of the internet and social media. This isn't really paralleled by the other technical advances that the authors of the study this piece summarizes were examining. Also, AI? We don't have anything even approaching AI in play now nor on the immediate horizon. The energy costs it would have are still enormous, even if we really could implement it. Maybe that's a revolution that's yet to begin. We got really good at problems that can be mapped into single precision matrix algebra, where the inputs and outputs are relatively small. It's a big jump to hard AI from there.

rectang 9y ago
> We got really good at problems that can be mapped into single precision matrix algebra, where the inputs and outputs are relatively small. General AI is a "wave of the future -- always has been, always will be" technology. Only when a piece gets broken out from general AI do we get practical benefits. The benefits from applying all that matrix algebra aren't small, though!
- eggie 9y ago
  So we are breaking off small pieces of a bigger thing. If that thing is AI, does it already have a form and an older name in the world? Would agree that in this article AI is a stand-in for what used to be called Math, Statistics, and Computer Science? Our scientific output is an artifice, and it is "intelligent" in its ability to guide our action. To me it isn't then strange to think of the sciences as the "general AI".
  - garmaine 9y ago
    As soon as you solve a problem in AI, people start labeling that problem as “just” math, statistics, or computer science. It is no longer considered AI. Shifting goal posts mean “real AI” is always in the future.
    - eggie 9y ago
      Even if we claim that "real AI" is a system that can actively perform at human levels in creative tasks, I wouldn't believe that it will always be in the future. However, I don't think this is very near, and I think this is sort of what people are implying when they talk about AI. To me saying "general AI is what we're exploring and making mini-discoveries in" is a kind of apology. Unfortunately, this apology pushes us into a grey area where "AI" is really just the scientific artifice. That's not the same as AI that has auto-creative capacities.
      - garmaine 9y ago
        What sort of creative task? Music composition? Graphic design? Surrealistic modern art? AIs have been programmed to do all these things and be indistinguishable from human creativity in blind studies. I have no doubt through that you'll be able to come up with a reason why each of these examples isn't what you really meant by "human level in creative tasks." Which is rather the point.
        sli 9y ago
        How specific are they to those tasks, is what I'm curious about. Can the surrealist bot produce musical compositions as well as the musical composition bot? The generality angle would say that one bot can do all of those things mostly or entirely on its own (i.e. trains itself as well, and was not necessarily "told" to train in that area), not multiple discrete bots that are all heavily trained by a human. Just to be clear, I'm not trying to argue with you in the sense that you convey in that second part of your comment, I just enjoy supposing about this particular subject.
    - mamon 9y ago
      People are trained by Hollywood to envision AI as robots. When presented an actual AI algorithm/model they see it as boring math, totaly unlike C-3PO, or Terminator, or whatever your favourite movie robot is.
      - deleted 9y ago
Eridrus 9y ago
"Productivity" as economists defines it is purely GDP/hours worked, it is only loosely correlated to what a lay person would consider productivity. It doesn't consider the value of consumer surplus or whether prices for goods have fallen due to foreign competition or anything else, it is purely productivity in the sense of our ability to generate dollars for every hour that someone is working.
- arbie 9y ago
  Is this definition flawed in the age where software can sometimes replace an entire profession? In such cases, can "productivity" sometimes fall in the face of technological advance?
  - Eridrus 9y ago
    It's tricky to predict what would happen to a statistic that measures the entire economy. At a first approximation all the money going to that profession would be removed from GDP, but so would result in productivity losses if we automated industries with higher than average "productivity". However, someone is doing the automating - are those people even more productive over the course of a year? Maybe, or maybe not and there's just a lot of investment that will bear fruit over the long run. But some amount of money is still being spent on the service, and the set of workers has gone down, so this would probably pull the number back up, but by how much? But that money isn't gone, it remains in someone's pocket, rather than disappearing, and the price of some good has fallen. Does this cause people to buy more of the good or some other good? If they buy more of some other good that has low "productivity", such as food and drinks, then that would pull productivity down. It would be hard to argue that such an outcome is unilaterally good though; we would hope that in the long run people find other high "productivity" activities, but whether we actually will is uncertain.
YaxelPerez 9y ago
Your brain could be considered a general intelligence, and it only uses a couple of watts. Energy consumption isn't the issue. However, I agree. AI is a bit overhyped right now, and AGI is still pretty far away imo.
- Robotbeat 9y ago
  It only uses a couple of watts (well, ten times that, actually, about 20 Watts), but that's because neurons are incredibly efficient. The brain is an Exa-ops computer that only requires 20 Watts. If we used today's state of the art TPUs, we're looking at 10,000 to 100,000 times the energy consumption, so on the order of a Megawatt. The electricity costs alone would be higher than the typical human wage. A corollary: If you can find a way to grow neural tissue (I'm thinking of all the lab-grown meat startups here) and connect to it efficiently and reliably (I'm thinking of Neuralink, here), then you would have access to a computational medium that is orders of magnitude more efficient than even the best of today's processors, and probably way cheaper, too.
  - chillacy 9y ago
    > an Exa-ops computer I think it's misleading to compare the brain to a computer, the architecture differences mean that the brain is great at some stuff (vision for instance) and terrible at others (multiply 2 64-bit integers in your head). I don't know if computational meat will ever be faster than machines at number crunching. (when I say machines I'm including next-generation computers beyond silicon too). But that's okay, each seems to complement each other's weaknesses. http://www.righto.com/2014/09/mining-bitcoin-with-pencil-and-paper.html http://www.righto.com/2014/09/mining-bitcoin-with-pencil-and...
gdubs 9y ago
If you're willing to commit to a very long (and IMO enjoyable) read, Robert Gordon addresses these possibilities in The Rise and Fall of American Growth.

rectang 9y ago

> We got really good at problems that can be mapped into single precision matrix algebra, where the inputs and outputs are relatively small. General AI is a "wave of the future -- always has been, always will be" technology. Only when a piece gets broken out from general AI do we get practical benefits. The benefits from applying all that matrix algebra aren't small, though!

eggie 9y ago
So we are breaking off small pieces of a bigger thing. If that thing is AI, does it already have a form and an older name in the world? Would agree that in this article AI is a stand-in for what used to be called Math, Statistics, and Computer Science? Our scientific output is an artifice, and it is "intelligent" in its ability to guide our action. To me it isn't then strange to think of the sciences as the "general AI".
- garmaine 9y ago
  As soon as you solve a problem in AI, people start labeling that problem as “just” math, statistics, or computer science. It is no longer considered AI. Shifting goal posts mean “real AI” is always in the future.
  - eggie 9y ago
    Even if we claim that "real AI" is a system that can actively perform at human levels in creative tasks, I wouldn't believe that it will always be in the future. However, I don't think this is very near, and I think this is sort of what people are implying when they talk about AI. To me saying "general AI is what we're exploring and making mini-discoveries in" is a kind of apology. Unfortunately, this apology pushes us into a grey area where "AI" is really just the scientific artifice. That's not the same as AI that has auto-creative capacities.
    - garmaine 9y ago
      What sort of creative task? Music composition? Graphic design? Surrealistic modern art? AIs have been programmed to do all these things and be indistinguishable from human creativity in blind studies. I have no doubt through that you'll be able to come up with a reason why each of these examples isn't what you really meant by "human level in creative tasks." Which is rather the point.
      - sli 9y ago
        How specific are they to those tasks, is what I'm curious about. Can the surrealist bot produce musical compositions as well as the musical composition bot? The generality angle would say that one bot can do all of those things mostly or entirely on its own (i.e. trains itself as well, and was not necessarily "told" to train in that area), not multiple discrete bots that are all heavily trained by a human. Just to be clear, I'm not trying to argue with you in the sense that you convey in that second part of your comment, I just enjoy supposing about this particular subject.
  - mamon 9y ago
    People are trained by Hollywood to envision AI as robots. When presented an actual AI algorithm/model they see it as boring math, totaly unlike C-3PO, or Terminator, or whatever your favourite movie robot is.
    - deleted 9y ago

Eridrus 9y ago

"Productivity" as economists defines it is purely GDP/hours worked, it is only loosely correlated to what a lay person would consider productivity. It doesn't consider the value of consumer surplus or whether prices for goods have fallen due to foreign competition or anything else, it is purely productivity in the sense of our ability to generate dollars for every hour that someone is working.

arbie 9y ago
Is this definition flawed in the age where software can sometimes replace an entire profession? In such cases, can "productivity" sometimes fall in the face of technological advance?
- Eridrus 9y ago
  It's tricky to predict what would happen to a statistic that measures the entire economy. At a first approximation all the money going to that profession would be removed from GDP, but so would result in productivity losses if we automated industries with higher than average "productivity". However, someone is doing the automating - are those people even more productive over the course of a year? Maybe, or maybe not and there's just a lot of investment that will bear fruit over the long run. But some amount of money is still being spent on the service, and the set of workers has gone down, so this would probably pull the number back up, but by how much? But that money isn't gone, it remains in someone's pocket, rather than disappearing, and the price of some good has fallen. Does this cause people to buy more of the good or some other good? If they buy more of some other good that has low "productivity", such as food and drinks, then that would pull productivity down. It would be hard to argue that such an outcome is unilaterally good though; we would hope that in the long run people find other high "productivity" activities, but whether we actually will is uncertain.

YaxelPerez 9y ago

Your brain could be considered a general intelligence, and it only uses a couple of watts. Energy consumption isn't the issue. However, I agree. AI is a bit overhyped right now, and AGI is still pretty far away imo.

Robotbeat 9y ago
It only uses a couple of watts (well, ten times that, actually, about 20 Watts), but that's because neurons are incredibly efficient. The brain is an Exa-ops computer that only requires 20 Watts. If we used today's state of the art TPUs, we're looking at 10,000 to 100,000 times the energy consumption, so on the order of a Megawatt. The electricity costs alone would be higher than the typical human wage. A corollary: If you can find a way to grow neural tissue (I'm thinking of all the lab-grown meat startups here) and connect to it efficiently and reliably (I'm thinking of Neuralink, here), then you would have access to a computational medium that is orders of magnitude more efficient than even the best of today's processors, and probably way cheaper, too.
- chillacy 9y ago
  > an Exa-ops computer I think it's misleading to compare the brain to a computer, the architecture differences mean that the brain is great at some stuff (vision for instance) and terrible at others (multiply 2 64-bit integers in your head). I don't know if computational meat will ever be faster than machines at number crunching. (when I say machines I'm including next-generation computers beyond silicon too). But that's okay, each seems to complement each other's weaknesses. http://www.righto.com/2014/09/mining-bitcoin-with-pencil-and-paper.html http://www.righto.com/2014/09/mining-bitcoin-with-pencil-and...

gdubs 9y ago

If you're willing to commit to a very long (and IMO enjoyable) read, Robert Gordon addresses these possibilities in The Rise and Fall of American Growth.