American Intelligence The two most important posts on Sofiechan this year didn't get nearly enough discussion. First, this one: >>5033... posted 4mo ago with 49 replies (collapse hidden) log in to judge received 17.8 17.8 American Intelligenc (view hidden) log in to judge received 17.8 17.8 Well said.... 4mo ago (collapse hidden) log in to judge received 4.6 4.6 Well said.... (view hidden) log in to judge received 4.6 4.6 Yes, this is an interesting prompt. Clearly there is latent desire that's currently being met in the most crass and retarded way possible, bottom up, by cartoon shysters like Joel Osteen.... 3mo ago (collapse hidden) log in to judge received 8.6 8.6 Yes, this is an inte (view hidden) log in to judge received 8.6 8.6 This is Panda Express tier meme territory. You say some little riddle "and in that moment, the student was enlightened." THAT is your estoteric based "cultic order"?... 3mo ago (collapse hidden) log in to judge received 6.1 6.1 This is Panda Expres (view hidden) log in to judge received 6.1 6.1 Lol I agree with this actually. We've strayed from the mission. The point of this thread is to discuss how a new, reality-based elite can succeed given that we are in the takeoff.... 3mo ago (collapse hidden) log in to judge received 6.7 6.7 Lol I agree with thi (view hidden) log in to judge received 6.7 6.7 Here's the most interesting answer to these questions I've found so far, via Dryden from Praxis:... 3mo ago (collapse hidden) log in to judge received 4.3 4.3 Here's the most inte (view hidden) log in to judge received 4.3 4.3 The big gap I see is many people still think of LLMs as next-token predictors trained over internet text. This is only the first step toward training a modern base model. A majority of the training compute is reinforcement learning over a wide variety of r... 1mo ago (collapse hidden) log in to judge received 7.0 7.0 The big gap I see is (view hidden) log in to judge received 7.0 7.0 Yes, it produces tokens. My point is that today, those tokens are only "predictions" in an abstract sense closely related to general intelligence. Early LLMs like BERT and GPT2 predicted tokens in a simple, concrete way: "what token is most likely to come ... 4w ago (collapse hidden) log in to judge received 6.1 6.1 Yes, it produces tok (view hidden) log in to judge received 6.1 6.1 Yes. Descartes would have a lot to ponder if he saw the kinds of advanced thinking that today's non-sentient machines are capable of. Our predecessors are impressive but not magic; like science, philosophy must update in response to new evidence.... 4w ago (collapse hidden) log in to judge received 7.3 7.3 Yes. Descartes would (view hidden) log in to judge received 7.3 7.3 Returning to the core topic of this thread... "many are saying this"... 4w ago (collapse hidden) log in to judge received 6.1 6.1 Returning to the cor (view hidden) log in to judge received 6.1 6.1
