You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Alison Gopnik on #AI:
- It's a different process from how kids learn.
- 4yos take a very small amount of data and make generalisations; current AI takes loads of data and is terrible at generalisation or invention.
- We can say "Maybe this isn't the right law." [AI can't]pic.twitter.com/BEjGXVLj5b – at SFJazz Center
I agree with 1, in a trivial sense. For 2, I think the following is a good point (&, of course, AI is doing more & more invention & generalization):https://twitter.com/karpathy/status/1084164152044572673 …
Programmers do invention and generalisation; AIs do not really (and AGIs will, once created):https://twitter.com/reasonisfun/status/939468092132003841?s=21 …
In this analogy the role of evolution is to create the architecture - once you have that (2) still holds right? The child is the endpoint of evolution - your neural network is the endpoint of your design. We need to find architectures and methods that allow for few-shot learning.
2 is *very* wrong w.r.t. kids learning from 'small amounts of data'. Children do not get told once "that's a dog" and that's then one labeled example for them to learn from. No - if they spend 1 min with this dog (or any dog sufficiently similar to it later on) (1/2)
(2/2) then their visual systems are seeing the dog tens (or hundreds) of thousands of times and this label of 'dog' can be re-used. The label is very likely re-used internally for different multi-modal views of this same dog, as long as the inputs keep resembling that same dog
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
