- Children learn by playing because not much is expected of the outcome in play. Improvement happens when you can play. When AI has a play environment to learn with reinforcement. When entrepreneurs are allowed to try and fail and do better. Doctors learn by practicing under supervision, or on corpses, until they can do it for real. No straight line goes up without a jiggle in the beginning.
- I like the networking perspective, but the ML perspective is such a loose analogy that it's hard to even judge. I mean, we've known forever softening constraints allows you to reach solutions otherwise unreachable, for one? There's a gulf of difference between succeeding at something deterministic by allowing failure vs. good pattern matching by optimizing over a rough landscape of examples.
by dataviz1000
0 subcomment
- The LLM reasoning models behave strikingly similar to superscalar out-of-order execution processors with decomposition, verification, and error correction steps.
Moreover, the LLM reasoning models are reliably consistent solving the same task with the same prompt using different variables. This can be demonstrated.
Not everything has to be deterministic to be useful. Nonetheless, understand how LLM models can be applied and be useful will help a lot of people to be less frustrated and spend less tokens.
by booleandilemma
0 subcomment
- Interesting. I could apply this to some people I've worked with. They work so well because they don't have to.
by simianwords
0 subcomment
- The interesting and useful parts of internet and AI come from their emergent properties - they were not handcrafted but just allowed to grow.
People hate that for some reason
- Another way to phrase this is that guarantees can impose constraints that are overly limiting.
Having said that, the parallel with ML seems to be bit of a stretch. What exact high level guarantee is given up by being probabilistic?
- I suppose another interesting thing about this observation is that this is true about the universe too! Einstein thought God doesn't play dice with the universe and Niels Bohr proved him wrong.
So either it's an interesting statement because it's infinitely generalizable or not interesting for the same reason.
Hacker News