@fchollet It seems increasingly likely that "consciousness" is simply an emergent capability of self-modeling. The consciousness we think of as "subjective experience" is likely many of these discrete building blocks being chained together by local threads.

@jeethu @bryan_johnson What if microplastics are just what the body craves? What a fortuitous discovery

@teortaxesTex Wait I actually just wrote about this as "discrete self-modeling," but I see no reason not to call it consciousness even without subjective experience. This is eerie

I used to think that LLMs would be "capable of abstract understanding up until the point where there needs to be a relation between the knower and the concept to understand." Now I just think they're conscious. But not in the way you might think.

@RaoulGMI A sense of "self" is likely an emergent feature in LLMs and their successors. In the forward pass of an LLM, if it reasons about itself in order to answer a question unrelated to itself, why isn't it conscious?

@pragmaticml @IndicoDataLabs Ok, just trained on it and getting 100% now. Can you guys release the next level?

@RaoulGMI So it's roughly the year 2000 for crypto? Did anything happen with the internet then?

@SethDeGroot Careful, that algebra may lead to linear algebra, which may lead to AI, which apparently threatens Zack's skillset

@zen_llama @ercwl So you're saying they're scarce?

@fishnets88 @probabl_ai Definitely true. It does seems like ColumnTransformer is more of a first class citizen lately. Excited to see the next episode in the series. @carlolepelaars and I are suckers for complex sklearn pipelines

RT @fishnets88: Just a friendly reminder to anyone interested: I'll be attempting a live stream this Friday. It'll be about temporal data l…

@laurashin Don't interview him

@GregKamradt That early emergent behavior is just consciousness. But don't be alarmed, we'll get used to it pretty quickly

@jonathanrlarkin

@xwang_lk @calabi_and_yau This is more like testing the "Chihuahua or Muffin with extremely advanced positional reasoning" problem

@Prutyke @ykdojo And it might not even be a big deal. We just have to kind of rethink how we view consciousness and intelligence. And it'll happen very quickly, probably in the same way we panicked about releasing the weights to GPT-3. And then we'll all go back to our lives

@jonathanrlarkin This is about 5 years sooner than I initially thought we'd be able to add charts/images to our models, but I guess I'll take it