2 ms·
It isn't exactly niche yet since it doesn't exist at this point but my speculation is that we will have AGI within that time frame. One reason I think that is
by ilaksh 8y ago
It isn't exactly niche yet since it doesn't exist at this point but my speculation is that we will have AGI within that time frame.
One reason I think that is because most of the core capabilities missing from AI systems are things that animals can do. Most of the difference with humans is a matter of degree.
It's still a very challenging list of capabilities of course. The core requirements for AGI are fairly well understood at this point. No one knows exactly how to achieve all of them together but many geniuses are working on it. And the tools such as DL, RL and less mainstream approaches such as SNNs and Sparse Predictive Hierarchies are very powerful. AGI is a real field now and intersects with things like developmental robotics.
It's a long shot but there are so many great minds involved and the rewards are so large that I think making a bet for within ten years is not a bad bet. Sure we have said that before. Things are not the same as before.
- AnimalMuppet 8y ago> The core requirements for AGI are fairly well understood at this point. That seems wildly improbable to me. It seems far more likely that anyone who thinks they understand them are grossly underestimating what's involved. Yeah, I know that I can't prove my position...
- ilaksh 8y agoProbably you misunderstood what I meant. What I mean is we have lists of capabilities that AGI needs to have and know how it differs from narrow AI. For example: online continuous learning, perception, generalization, transferring skills between domains, etc. These aspects of the _what_ (rather than how) of AGI have been well researched.