Hmm… If I do a few searches, and find that no one else is discussing the subje

Hmm… If I do a few searches, and find that no one else is discussing the subject with clarity, then I can do that. There is a lot of ‘noise’ being produced and not much signal. I suspect that’s because few people understand the mechanics of both the brain and the software, and fewer can reduce the difference to an explanation digestable by the curious. In simple terms, the language model is getting very close to the same dependency chain that our brains use – though artificial ‘neuron’ is now a bit misleading. It’s more like neural minicolumn or column. Our brains disambiguate everything in relation to the current body positions. So our body and actions are the measurements it can use to test limits. To some degree the language model produces a bodyless brain in a vat, lacking that embodiment as a system of measure, and must learn about ‘bodies’ by inference. Now, we learn about abstract concepts the same way. So it shouldn’t be that impossible to ‘get there’ with enough working memory for wayfinding (recursive searching). And I see that this is happening … a bit. Shallowly. But it’s there.

Reply addressees: @RealLouisPrince @NektariosAI @GaryMarcus