Tiny AI Startup Vicarious Says It’s Solved CAPTCHA
singularityhub.com
singularityhub.com
https://news.ycombinator.com/item?id=6625245 (forbes.com)
https://news.ycombinator.com/item?id=6625247 (kurzweilai.net)
https://news.ycombinator.com/item?id=6625351 (technologyreview.com) <- Main discussion
https://news.ycombinator.com/item?id=6626405 (vimeo.com) <- video of process in action
https://news.ycombinator.com/item?id=6627848 (vicariousinc.tumblr.com)
https://news.ycombinator.com/item?id=6628086 (cbc.ca)
https://news.ycombinator.com/item?id=6628092 (wired.com)
https://news.ycombinator.com/item?id=6629173 (wired.com)
https://news.ycombinator.com/item?id=6629559 (newscientist.com)
https://news.ycombinator.com/item?id=6629656 (dailydot.com)
https://news.ycombinator.com/item?id=6629708 (mashable.com)
========
There are two other comments worth reading.
https://news.ycombinator.com/item?id=6629173:
That's a lot of fancy words to say that they overfit
to their training data.
That makes this sound like a very typical result in
supervised machine learning (if it's a result at all).
They have used an algorithm to learn a brittle heuristic
that works in the cases it was trained to work on.
https://news.ycombinator.com/item?id=6628092: Please look at this comment from Yann LeCun:
https://plus.google.com/104362980539466846301/posts/Qwj9EEkU... [0]
"Beware: It's a textbook example of AI hype of the worst kind"
[0] https://plus.google.com/104362980539466846301/posts/Qwj9EEkU...