Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

wired.com

Post reply on HN
Print view View original

101–110 of 239 posts

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#101

Post by ewjordan »

Houshalter wrote: post #98
Animats wrote: post #91

Earlier quoted context omitted.

We don't even have good generalized robot manipulation. Think of that as monkey-level AI. After half a century of hard work, robot manipulation in unstructured situations still sucks. Go watch DARPA Humanoid Challenge videos or old Willow Robotics towel-folding to see how badly it sucks. Factory robotics works because the work situation is highly structured. On the other hand, once someone cracks that, a huge number…

That's exactly what is about to happen. Deep learning has the potential to do robot control. Currently researchers are beating tons of video games using reinforcement learning with deep networks. Applying the same methods to robots shouldn't be too hard. And we've also come a long way with machine vision as well over the past 5 years.

I'm as bullish on AGI as anyone in the medium term, but deep learning is not even playing the same game as AGI, let alone in the same ballpark or having the potential to achieve it.

Deep learning is still mere perception. It doesn't handle memory or processing, it just transforms input into output, typically trained by Big Data, way bigger than necessary statistically speaking, given the world we live in.

AGI requires super aggressive unsupervised learning in recurrent networks, likely with specialized subsystems for episodic and procedural memory, as well as systems that condense knowledge down to layers of the network that are closer to the inputs. At a minimum. And nobody is really working on any of that yet (or at least succeeding) because it's really damn hard.

That's why everyone in "AI" is rebranding as a deep learning expert, even though deep learning is really just 1980s algos on 2016 hardware - you gotta sex up feed forward backprop or you don't get paid.

Edit: to be fair, robot control is much simpler than AGI, and might be mostly solved with deep learning somewhat soon, I forgot the context of your post.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#102

Post by ensiferum »

The future of humanity? The few and select Big Corp will own the infrastructure and technology and employ and sell it to reduce human labor as much as possible thereby creating massive value increases for the stock holders and displacing millions of blue collar workers into poverty. (In later stages also white collar workers). The execs will all be trillionaires and live off their own walled societies while the rest of us fight for our lives for tooth and nail. Eventually we can't even have a revolution since the goverments are owned by the Big Corp, which has manipulated and coerced the system into activities such as spying on their own citizens on the pretext of terrorism etc. Any act of revolution will be quickly and swiftly dealt with immediate force.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#103

Post by Houshalter »

zaroth wrote: post #35
pmyjavec wrote: post #11

Earlier quoted context omitted.

Why aren't you "concerned" about people losing their jobs? The alternatives are tech companies start to pay their taxes and humans start to use tech to care for one another, genuinely, not just wealthy Silicon Valley types trying to make buckets because their "entitled" to do so and pretend it's all in the name of progress. The idea of the Internet and Automation were exciting to me as it was about liberation and dec…

Consider a broader timeline. It used to take humans over 500 hours of labor spinning, weaving, and sewing to produce a single shirt. Automating textile production has "put more people out of work" than computers ever will. Do you think for a second we would ever want to go back to how it was? Do you ask the inventor of the sewing machine how they will start giving back? These inventions free humanity for more worthwh…

Well yeah a lot of unemployed textile workers did go into poverty or even starve to death at the time. I've read some speculation that the first industrial revolution was a net negative on society initially, and took many decades to actually improve the average standard of living. There was a reason people gathered in mobs to break the new machines. They were hungry.

But still the industrial revolution didn't really replace laborers. The machines were still quite limited, and humans were still needed to do the jobs machines couldn't do. What's different this time, is soon the machines will be able to do everything humans can do. Or at least everything an unskilled worker can do. Operating a machine in a factory, driving a truck, entering data into a computer, making phone calls, these are all things machine learning is capable of.

Lastly look at horses. The invention of trains would seem to have competed with them and taken many of their jobs. But instead horses vastly increased, because trains couldn't do everything horses could do. Then cars were invented, and the horse population crashed over just a decade.

Why did this happen? Didn't the invention of the train prove that automation doesn't take horse jobs? Shouldn't there always be new jobs for horses? Can't horses specialize in the 1% of tasks that cars can't do, like transportation in places without roads?

But that didn't happen. The cost of just feeding the horses was much higher than the cost of buying an automobile. There were some obscure jobs for horses left, but nowhere near enough.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#104

Post by Houshalter »

ewjordan wrote: post #101
Houshalter wrote: post #98

Earlier quoted context omitted.

That's exactly what is about to happen. Deep learning has the potential to do robot control. Currently researchers are beating tons of video games using reinforcement learning with deep networks. Applying the same methods to robots shouldn't be too hard. And we've also come a long way with machine vision as well over the past 5 years.

I'm as bullish on AGI as anyone in the medium term, but deep learning is not even playing the same game as AGI, let alone in the same ballpark or having the potential to achieve it. Deep learning is still mere perception. It doesn't handle memory or processing, it just transforms input into output, typically trained by Big Data, way bigger than necessary statistically speaking, given the world we live in. AGI require…

I'm not talking about AGI at all! Just robot control. It's a difficult problem sure, but not that difficult. There has been massive progress on it, and related problems. I have no doubt we will have 'solved' it in at a decade.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#105

Post by downandout »

pashariger wrote: post #62

I don't know about you, but I'm very thankful that we have a president that can understand and intelligently talk about these issues (rather than thinking it's all just black magic).

I'm curious...which President are you saying thought it was "all just black magic"? Let's also remember that Obama is the man that unapologetically expanded covert/illegal surveillance programs to an unprecedented scale, directly costing American technology companies billions of dollars in sales. I'd say his actions reflect neither intelligence nor understanding of the tech industry.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#106

Post by Houshalter »

Animats wrote: post #99
Houshalter wrote: post #98

Earlier quoted context omitted.

That's exactly what is about to happen. Deep learning has the potential to do robot control. Currently researchers are beating tons of video games using reinforcement learning with deep networks. Applying the same methods to robots shouldn't be too hard. And we've also come a long way with machine vision as well over the past 5 years.

The systems which learn video games work only for games where the state of the game is entirely visible, and the desired player action can be decided based only on the current state. PacMan, yes. Doom, not that way.

That's only because they didn't use recurrent neural networks which save information over time. RNNs make it possible to play games with hidden state. Deepmind is currently working on that with starcraft, which is vastly more complicated than pacman. They also have some work on 3d games like doom.

A few weeks ago there was a paper posted on "Synthetic Gradients" which should make it much more practical to train RNNs for games. Before it required saving every single computation the computer makes to memory, which uses a huge amount of memory and computation. Using synthetic gradients they need only store a few steps in the past. And it can learn online.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#107

Post by blahi »

moeabbas999 wrote: post #44
bitL wrote: post #41

Earlier quoted context omitted.

Specialized AI - think about retina neurons, capable of detecting direction, edges, depth etc. General AI - think about thinking machines We can do the former, have no clue how to do the latter.

It's not such a huge jump to the latter when you nail the former.

Can you give a little bit of details about your background in AI or at least statistics/machine learning?

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#108

Post by dwaltrip »

golergka wrote: post #15
pmyjavec wrote: post #11

Earlier quoted context omitted.

Why aren't you "concerned" about people losing their jobs? The alternatives are tech companies start to pay their taxes and humans start to use tech to care for one another, genuinely, not just wealthy Silicon Valley types trying to make buckets because their "entitled" to do so and pretend it's all in the name of progress. The idea of the Internet and Automation were exciting to me as it was about liberation and dec…

A lot of comments like yours are based on the assumption of "giving back". I don't quite understand it. Let's suppose I've founded booking.com and put a LOT of travel agents out of business. At what point did I assume the responsibility of giving back to those travel agents? I agree that they are miserable because of my actions. However, in real life, nothing happens because of a single reason. They are also miserabl…

Your successful web business would be a very small event at the end of an innumerably long causal chain involving billions of humans and tens of thousands of years [1].

Can you name every event that would have enabled you to be ready to create the business, and all aspects of the world that would allow for its success? It's impossible. This is why we have to give back.

[1] We could obviously go back further, but it becomes less meaningful in the context of this discussion.

Re: Barack Obama on A.I., Autonomous Cars, and the Future of Humanity

#109

Post by imtringued »

greeneggs wrote: post #70
gefh wrote: post #38

I wish people would stop bringing up the trolley problem as somehow relevant to self driving cars. In any situation,a car is going to brake as hard as possible while maintaining control, and only serve if there's no way to stop and a clear space. Other than that, a good enough interview that I can't see taking place again next presidential term.

The trolley problem is just an extreme example, that people can easily understand and think about. Less extreme versions of it occur all the time. For example, say you are driving along a road, passing a bicyclist. You'd like to give the cyclist more room, but there's an oncoming car in the lane next to you. How much room do you give? At exactly what threshold do you decide to wait to pass the cyclist? What if the cy…

If there is an oncoming car you don't overtake.
Post reply on HN