7 ms·

Ray Kurzweil Wants to be a Robot

30 points by haidut 17y ago

jamesk2 17y ago
Kurzweil is a genius but maybe he should try not to put such specific dates on his predictions. It makes it easy to say he was wrong by detractors.
- sp332 17y ago
  Kurzweil usually makes predictions about things that have strong trends over time. His favorites are the trends that increase exponentially, because most people extrapolate the early, slow growth linearly and come to a completely wrong conclusion. If you watch his TED talk, he explains his technique. http://www.ted.com/talks/lang/eng/ray_kurzweil_on_how_technology_will_transform_us.html http://www.ted.com/talks/lang/eng/ray_kurzweil_on_how_techno...
  - bitwize 17y ago
    Uh, yeah, except the funny thing about exponential curves is that at small scales they look a hell of a lot like sigmoid curves.
- burke 17y ago
  Kurzweil's oddly-specific dates have usually been correct.
- pj 17y ago
  It also makes it easy to verify that he was right.
Periodic 17y ago
When I first heard about the whole Singularity thing it made sense. I thought it was cool. I wanted to be there. But as I've learned more about the limitations of our understanding, our biology, and computers I've realized that this isn't something that will be happening any time soon without some amazing breakthroughs. I don't expect the Singularity to happen my lifetime, and I'm still in my 20s. We really understand very little about the brain. We don't know exactly how neurons work and the most information we can get from the brain is on the order of 1-7 BITS per second when sampling from implanted electrodes. It's also not a lack of data. We don't even know how to fully decode the data we already have. Our understanding of AI really is pretty poor. Knowledge representation and machine learning has become a hard enough task that many researchers are back to developing algorithms for specific tasks, which embeds the knowledge of the researcher in the algorithm, precluding the system from learning anything truly novel. They say we can't see past the Singularity because things will change too much and too fast. Right now we can't even see the Singularity because too many things would have to change too much and quite fast.
- speek 17y ago
  We're not gonna get there using purely modeled AI or purely modeled machine learning. It's too much work to do it that way. We need a biologically inspired computer architecture... one that's non von Neumann; where the stored data and computation aren't separate.
- Estragon 17y ago
  There are thousands of people working furiously on all of those problems, though. I was working at Janelia Farm until last year. You should see what those guys are planning for neuroscience. It's going to be huge. It's a mistake to assume that our capacity to query neural circuitry is going to remain stuck at such a low rate.
  - TrevorJ 17y ago
    And they have been working furiously for a good while now. The romantic in me hopes that they can make it happen, but I'm not holding my breath. Mainly because a lot of people confuse speed of processing with intelligence, but that's not really the spark of life we need. There's nothing out there so far that can mimic human thought even at hundreds of times slower than human responses. We still don't have something quite right with the basic principles yet.
    - Devilboy 17y ago
      What Kurzweil is saying is that you can't just linearly extrapolate the progress of these projects. Even if we only know a fraction of a percentage of what's going on in the brain after 20 years doesn't mean we'll know only twice as much in another 20 years. Rather we're likely to see the pace accelerate. Going from 50% to 100% might take less time than going from 1% to 2% knowledge. Keep in mind that he's been making predictions like these for decades and he's got a pretty good track record.
      - TrevorJ 17y ago
        I guess that I am saying is that I'm not convinced that the direction we have been heading is the direction that will ultimately produce the end result we want. The model T and the present day automobile share the same basic concepts. I'm wondering if human-level AI will bear any real resemblance to today's solutions on the concept level. I'm not convinced it will.
joechung 17y ago
The last kind of person I would help become a robot is someone that wants to become one.
Tichy 17y ago
"advance in a linear fashion but rather at an exponential rate. It's the difference between 1-2-3-4-5 and 1-2-4-8-16." Don't know, should I laugh or should I cry?
danbmil99 17y ago
hey 20-somethings, stop complaining it might not happen in your lifetime and just pitch in. You could live another 70 years. Think about 1906 vs 1976. I'm almost 50 and I'm doing what I can. I almost definitely won't make it, but you guys have a 30 year extra chance. And if you don't, well fuck, your kids or grandkids will. If you've had someone you love get old and sick and die, you know what I'm talking about. Get to work. It's not just about you.

30 points by haidut 17y ago

Kurzweil is a genius but maybe he should try not to put such specific dates on his predictions. It makes it easy to say he was wrong by detractors.

sp332 17y ago
Kurzweil usually makes predictions about things that have strong trends over time. His favorites are the trends that increase exponentially, because most people extrapolate the early, slow growth linearly and come to a completely wrong conclusion. If you watch his TED talk, he explains his technique. http://www.ted.com/talks/lang/eng/ray_kurzweil_on_how_technology_will_transform_us.html http://www.ted.com/talks/lang/eng/ray_kurzweil_on_how_techno...
- bitwize 17y ago
  Uh, yeah, except the funny thing about exponential curves is that at small scales they look a hell of a lot like sigmoid curves.
burke 17y ago
Kurzweil's oddly-specific dates have usually been correct.
pj 17y ago
It also makes it easy to verify that he was right.

Periodic 17y ago

When I first heard about the whole Singularity thing it made sense. I thought it was cool. I wanted to be there. But as I've learned more about the limitations of our understanding, our biology, and computers I've realized that this isn't something that will be happening any time soon without some amazing breakthroughs. I don't expect the Singularity to happen my lifetime, and I'm still in my 20s. We really understand very little about the brain. We don't know exactly how neurons work and the most information we can get from the brain is on the order of 1-7 BITS per second when sampling from implanted electrodes. It's also not a lack of data. We don't even know how to fully decode the data we already have. Our understanding of AI really is pretty poor. Knowledge representation and machine learning has become a hard enough task that many researchers are back to developing algorithms for specific tasks, which embeds the knowledge of the researcher in the algorithm, precluding the system from learning anything truly novel. They say we can't see past the Singularity because things will change too much and too fast. Right now we can't even see the Singularity because too many things would have to change too much and quite fast.

speek 17y ago
We're not gonna get there using purely modeled AI or purely modeled machine learning. It's too much work to do it that way. We need a biologically inspired computer architecture... one that's non von Neumann; where the stored data and computation aren't separate.
Estragon 17y ago
There are thousands of people working furiously on all of those problems, though. I was working at Janelia Farm until last year. You should see what those guys are planning for neuroscience. It's going to be huge. It's a mistake to assume that our capacity to query neural circuitry is going to remain stuck at such a low rate.
- TrevorJ 17y ago
  And they have been working furiously for a good while now. The romantic in me hopes that they can make it happen, but I'm not holding my breath. Mainly because a lot of people confuse speed of processing with intelligence, but that's not really the spark of life we need. There's nothing out there so far that can mimic human thought even at hundreds of times slower than human responses. We still don't have something quite right with the basic principles yet.
  - Devilboy 17y ago
    What Kurzweil is saying is that you can't just linearly extrapolate the progress of these projects. Even if we only know a fraction of a percentage of what's going on in the brain after 20 years doesn't mean we'll know only twice as much in another 20 years. Rather we're likely to see the pace accelerate. Going from 50% to 100% might take less time than going from 1% to 2% knowledge. Keep in mind that he's been making predictions like these for decades and he's got a pretty good track record.
    - TrevorJ 17y ago
      I guess that I am saying is that I'm not convinced that the direction we have been heading is the direction that will ultimately produce the end result we want. The model T and the present day automobile share the same basic concepts. I'm wondering if human-level AI will bear any real resemblance to today's solutions on the concept level. I'm not convinced it will.

joechung 17y ago

The last kind of person I would help become a robot is someone that wants to become one.

Tichy 17y ago

"advance in a linear fashion but rather at an exponential rate. It's the difference between 1-2-3-4-5 and 1-2-4-8-16." Don't know, should I laugh or should I cry?

danbmil99 17y ago

hey 20-somethings, stop complaining it might not happen in your lifetime and just pitch in. You could live another 70 years. Think about 1906 vs 1976. I'm almost 50 and I'm doing what I can. I almost definitely won't make it, but you guys have a 30 year extra chance. And if you don't, well fuck, your kids or grandkids will. If you've had someone you love get old and sick and die, you know what I'm talking about. Get to work. It's not just about you.