@osma I predicted one of its defenses before I heard of Gandalf, which happened to be the level I gave up on and never got past. So far.

@orbiterlab Our world could be a leaf node of a world able to simulate more than one world. If m is the number of worlds simulated per world and m is greater than one, then the likelihood is not small.

@davidmortensen I agree and wish I had the same grasp of the field as you. My point is there are people out there having quasi-religious experiences and someone soon will try to sell them a bridge.

@aran Does the "M" in "LLM" refer to a trained latent space representation or its architecture? Or both?

@davidmortensen Does the M refer to a trained latent space representation or its architecture? Or both? Isn't a function approximator the sum of its parts? Not calling them algorithms is manipulative to those who do not understand a distinction between output generated by symbolic AI versus inference from data.
https://www.bostonglobe.com/2023/05/08/opinion/google-bard-chatgpt-dishonest-anthropomorphism-evan-selinger/

Word on the street is @garymarcus keepin' it real.

@Niessner I thought Altman said they weren't training GPT-5, which implied to me there is no LLM equivalent to Moore's Law, and they've hit the ceiling of what can be acheived with next token prediction.

@yonomitt Have you explored Javascript's BigInt or maybe its typed arrays?

@minimaxir I don't believe A.I. will have the ability to advance user interface animation at all. It will use already established techniques which are "good enough". I'm excluded by traditional means, gatekeeping, which will not change.