3 ms·
The really nice thing about this is that the AI can now acquire these newly-decoded texts as part of its training set, and begin learning at a geometric rate.
by zozbot234 2y ago
The really nice thing about this is that the AI can now acquire these newly-decoded texts as part of its training set, and begin learning at a geometric rate.
- mistrial9 2y agoerrors in => errors out
- WhereIsTheTruth 2y agoDon't forget to spice it up with some bias! https://x.com/i/grok/share/uMwJwGkl2XVUep0N4ZPV1QUx6 https://x.com/i/grok/share/uMwJwGkl2XVUep0N4ZPV1QUx6
- rzzzt 2y agoBut do I want to see ancient programming advice written in Linear B?
- zeofig 2y agoWhy not just feed it random data? It's so smart that it will figure out which parts are random, so eventually you will generate some good data randomly, and it will feed on it, and become exponentially smarter exponentially fast.
- Validark 2y agoThis is actually hilarious and I'm sad you are getting downvoted for it.
- nitwit005 2y agoWith our current methods, feeding back even fairly small amounts of outputs back in as training data leads to declining performance. Just think of it abstractly. The AI will be trained on the errors the previous generation made. As long as it keeps making new errors each generation, they will tend to multiply.
- red75prime 2y agoDegradation of autoregressive models being fed their own unfiltered output is pretty obvious: it's, basically, noise being injected into the ground truth probability distribution. But. "Our current methods" include reinforcement learning. So long as there's a signal indicating better solutions, performance tends to improve.