3 ms·

All the academic AI researchers are focusing on boring narrow AI problems, and nobody has the balls to really tackle the big problem[s] of Artificial General In

• machine 19y ago
  I disagree with you on a lot of points. First, I don't think AI has failed to deliver. There are a ton of things we can do now that we couldn't 30 years ago. Just to name a couple, how about simultaneous localization and mapping http://web.mit.edu/16.412j/www/html/readings/Eliazar+Parr-ijcai-03.pdf, http://web.mit.edu/16.412j/www/html/readings/Eliazar+Parr-ijcai-03.pdf, and face detection http://citeseer.ist.psu.edu/viola01robust.html. http://citeseer.ist.psu.edu/viola01robust.html. I also disagree that AI researchers are working on boring narrow problems. There are plenty of researchers working on huge open problems like object recognition, speech recognition, reinforcement learning, etc. And they're making great progress too. I think the reason most AI researchers don't talk about making "human-level AI" is that this isn't really a quantifiable goal. There are plenty of things computers can do better than people already (would you want a computer with "human-level" addition and subtraction?). And for just about every specific useful task humans do better than computers, there are a few dozen AI researchers working on it. Why is that not a reasonable approach to the problem?
  • jey 19y ago
    Yes, we can do a lot that we couldn't do 30 years ago. However, this only falls under the banner of "AI" because AI has been redefined from the promises of 50 years ago [1], when the goal of AI was to build "human-level" intelligent agents. The "Dartmouth Summer Research Project on Artificial Intelligence" proposal [1] in 1955 when revered AI researchers Marvin Minsky, (inventor of Lisp) John McCarthy, and (inventor of information theory) Claude Shannon claimed "a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer". They failed at this lofty goal of solving the AI problem over a summer, and over time the field of AI was redefined to be narrow fancy mathematical tricks to solve domain-specific problems. Most "AI" is not intelligent at all, instead they are approaches for searching some search space with a non-general-purpose tool, such as genetic algorithms, neural networks, or first-order logic theorem provers. A "true AI" would discover regularities [2], or patterns, in any search space, and exploit these patterns to incrementally improve its ability to navigate the search space. Examples of regularities in the world are that navigation is often best performed by sticking to the regularly appearing feature that we call a "road" instead of attempting to climb over buildings, and that navigation can be effectively guided by attempting to minimize the Euclidean distance between the current location and the goal location. Instead, most of the currently available tools are essentially performing a stochastic search through the state space, possibly guided by some heuristics. While a lot of progress is being made in the redefined and inaccurately named field of AI, object recognition and speech recognition are still narrow domains. They are not using general purpose approaches, but instead tweaking and tuning specific algorithms to these areas. In reality, I am not championing for "human-level AI", as I think it would be very difficult to duplicate a human's strengths and deficiencies [3] accurately. Instead, I am saying that we should be encouraging research that produces a general intelligence; software that can recognize and exploit regularities (patterns) in the state spaces of many problem domains, and apply the abstracted knowledge gained from earlier experiences to new domains and problems. Such software would have a tremendous impact on society, assisting us in solving and analyzing many problems and likely leading to an "intelligence explosion" after crossing some threshold at which the software is able to recursively improve itself. [4] You asked "just about every specific useful task humans do better than computers, there are a few dozen AI researchers working on it. Why is that not a reasonable approach to the problem?" How do we take all these domain specific solutions and get a cohesive general intelligence out of them? I see these domain specific algorithms as being possibly useful to a general intelligence as an optimization to shift some of the load off of the more general purpose algorithms, but I don't see how throwing a bunch of face recognition, path finding, neural network and genetic algorithms into a box would automatically cause a general intelligence to pop out. These research areas seem mostly tangential or supportive to genuine Artificial General Intelligence research. 1. http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html 2. http://en.wikipedia.org/wiki/Kolmogorov_complexity http://en.wikipedia.org/wiki/Kolmogorov_complexity 3. http://www.singinst.org/Biases.pdf http://www.singinst.org/Biases.pdf 4. http://en.wikipedia.org/wiki/Technological_singularity http://en.wikipedia.org/wiki/Technological_singularity
• ced 19y ago
  Donating to the institute you mention triggers the same reaction in me as the article's plea for funding: if anyone has a good idea for making AI, he should have no problem convincing other hackers to join in and form a startup. Computer ressources are probably not the limiting factor. I'm sure that a lot of researchers in academia have pet theories about how to make it happen, but they need grants. The NSF has funded enough failed AI projects, it's hard for them to justify funding more given the current lack of money in science. There are probably a lot of people working on AGI on their own, or in startups. I am. But you won't hear much about people like me unless we succeed to some level. IMO, the field is waiting for its Einstein to come up.
  • jey 19y ago
    I'm suggesting that we fund research into the problem of AGI. I'm not saying that we should be funding teams to code up someone's personal AGI hypothesis. In the past, too much attention has been focused on trying to code up an AI without any real justification for why that AI system would work, and this has led to the failures you cite. What we need is fundamental research into intelligence, theories that describe intelligence and how to attain it. This is how research progresses in other fields; rocket scientists do not come up with their personal favorite hypothesis of rocket propulsion then seek funding to try to build a rocket using that propulsion system. Instead, they form models and consistent theories to validate their hypotheses before actually attempting to build a rocket. Why should AI research be any different? If the implicit assumption is that general intelligence is fundamentally just too difficult to understand, I disagree, and have not seen a convincing argument making that case.
    • ced 19y ago
      Theories are ultimately validated through experimentation and observation. For "intelligence", that means coding it up, and see if it learns anything. That doesn't sound very different to me from "funding teams to code up someone's personal AGI hypothesis".