On that last point about sharing knowledge with others, I just wanted to double-check with the community that sharing knowledge is still the status quo since what we share could conceivably help others achieve their goals and become our competitors - thoughts? 5/5
The #NeurIPS review process this year is very eye-opening. Reviewers seem disengaged (across all papers - maybe they are tired from being forced to battle LLMs?) 1/k
Also, since the LLMs have all read my papers (the prompt injections make it through to the reviews for those who couldn't afford more expensive LLMs) - I plan to release all of my submissions over the coming week so that humans can read them too 4/k
@gabriberton Let’s keep in mind, Schmidhuber anticipated the non-mystical version: intelligence = compression/prediction. Text, images and audio are different projections of the same world, so scale should push models toward the shortest shared latent structure. PRH is Plato meets AIT. </s>?
Is it conceivable that AGI eventually appears not in large data centres owned by massive corporations, but rather, within a global network of connected collaborating agents? 🤔
@fyang007@CVPR In doing things this way, we avoid punishing the “innocent” authors, and also denying the community / humanity of knowledge that advances the field. The fundamental objective is knowledge sharing, not handing out (or denying) individual tokens of prestige. 4/4
@fyang007@CVPR Perhaps a more suitable remedy is to publish that paper. But bar the irresponsible reviewer from participating in that particular channel for knowledge dissemination (and others maybe) for a defined period. 3/4
@CSProfKGD Thanks Kosta for posting these images for those who couldn’t be there. Even though I’m sure it doesn’t match the experience of hearing the talk, it is nevertheless thought provoking.
@MagicTony@J_D_Crawford I believe the more “dotty” version is attributed to Richard Gregory from The Intelligent Eye… or this is what I’ve always assumed.