The first cellphone weighted about 2 kg and had nearly no functionality. Should we have given up and stuck to landlines? You don't give up on something when you just started to make progress, only because the ultimate goal is still far away.
-
-
-
changing tack ≠ giving up
-
Tweet unavailable
-
Gary's not saying this track is wrong, just that it's only part of the solution. Full general intelligence seems to have 3 components - (1) Reflection (forming symbolic causal models), (2) Pattern Recognition (machine learning), (3) Optimization - Achieving goals in the world
-
The statistical/pattern recognition approach of machine learning (2) is only 1/3rd of the solution! Still missing the other 2/3rds!
-
There's actually 3 different kinds of inference. (1) The logic-based approach (2) The statistical approach (3) The information-theoretic approach Machine learning only really develops (2), the statistical approach so again, 2/3rds of the answer is left out!
-
Interesting. I am new to this. How exactly do you distinguish (2) and (3)? The information theoretic connection that I'm aware of is this: https://arxiv.org/abs/1608.08225
-
They refer to different levels of abstraction - as an analogy, think of physics vs. chemistry. Links to my wiki-books - will bring you right up to speed! (2) Probability&Statistics https://en.wikipedia.org/wiki/User:Zarzuelazen/Books/Reality_Theory:_Probability%26Statistics … (3) Information&Complexity https://en.wikipedia.org/wiki/User:Zarzuelazen/Books/Reality_Theory:_Computation%26Complexity …
- 2 more replies
New conversation -
-
-
The problem is very hard but this is not the first time humans had one. The amount of incentive to solve it has to be taken into account - doing ML is one of the best jobs now in both academia and industry.pic.twitter.com/rYJFQxi5pD
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Category error. As Turing pointed out, when computers start to write poetry, it will be computer poetry, not human poetry. Computers will never think like humans, but they will think like computers and eventually as conscious computers.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Good article until the final recommendation to resort to knowledge engineering. Given the enormous complexity of not just language, but the world that can't work. I tried to hard code the rules of baseball recently, and that's a pretty simple system compared to language.
-
Not suggesting you hand code the whole thing, or to learn it all, either. We need strong primitives to learn good stuff. See my January arXiv article on AlphaGo and innateness.
-
So if I'm understanding you, we need to define some tiny small set of foundational knowledge that will allow better learning from there?
-
Yes, I think that's right. The key is understanding how we form symbolic models of reality that enable us to *communicate* (including self-reflection or talking to ourselves). So what are the basic 'building blocks' (or ontological primitives) of thought?
-
To find out, one needs to look at the structure of knowledge. Wikipedia is huge corpus of knowledge. I spent around a year and half trying to categorize wikipedia articles. I wanted to try to determine what the most natural categorization scheme for explanatory knowledge was.
-
I was also trying to answer the question What is the minimum number of concepts (wikipedia articles) I'd need to summarize all explanatory knowledge? What's the minimum number that would enable me to pick up any scientific paper on any subject and at least grasp the main ideas?
-
Well, I initially guessed I'd need about 1 000 - 2 000 articles, but in the end the actual number turned out to be quite a bit higher. I ended up topping out at around 9 000 wikipedia articles. This turns out to be enough to summarize all knowledge.
End of conversation
New conversation -
-
-
"Heavier than air flying machines are impossible" Lord Kelvin 1895.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Is Ray Kurzweil, that avatar of wishful thinking, still working at Google?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.