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
My takeaway from Rich Sutton’s “The Bitter Lesson” is that AI methods that survive the test of time tend to scale. Systems that scale have simple and elegant designs that are easy to understand. Designing such systems require even greater human creativity. http://www.incompleteideas.net/IncIdeas/BitterLesson.html …
I like your takeaway @hardmaru but think Rich’s piece is way too dismissive of innateness, and doesn’t recognize that specific prior knowledge framed in the right way — eg convolution or machinery Monte Carlo Tree Search — is often critical.
The question is whether these innate structures *are* what they *seem* to be. If we can find a clever meta-method of guiding the system to rapidly re-discover them from scratch in the way which functions well, that would work beautifully. For example, this is what Bengio wants:pic.twitter.com/3pTd0x0Ur8
Ideally, you would be able to justify the architecture of a neural net in terms of axioms that hold for the problem to solve. Eg. geometric shapes are invariant if shifted around and this is exploited in convolutional networks. So knowledge may justify the structure of a network.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
