Loading video...
Video Failed to Load
Please enjoy benahorowitz.eth and me answering questions on the current state of AI! 01:14 How AI startups can compete with the bigger players 05:57 Will the "God models" get 100x better? 09:19 Counter argument: Are the tests too simple? 10:14 Internet data represents average human activity 11:50 AI's ability... show more
1,957,301 views • 2 years ago •via X (Twitter)
9 Comments

@bhorowitz Deleting the original post with 403k views (and an Elon share) in order to repost the video with working time-stamps — flex 💪

@elonmusk @bhorowitz

@elonmusk @bhorowitz AI should be unbiased, completely neutral without any emotions but with the morals

@elonmusk @bhorowitz I wish I was 20 years younger. Exciting, if a little scary, times ahead. Mustn't fall into the trap of becoming wholly reliant upon AI. The Machine Stops is a novel well worth reading in this respect. E M Forster. Another 'visionary'?

@elonmusk @bhorowitz In the near future Artificial Intelligence will be used for advanced totalitarianism as well as "undesirable" population reduction. These are must see videos on the subject

@bhorowitz At this point, anyone who isn't advocating for a fully open-source democratized AI, almost certainly has ties to ultra-wealthy elitists that seek to consolidate their power for generations by controlling the technology that will be woven into everything in the future.

@bhorowitz You two are sharp

@bhorowitz if u need faster than 2x video speed its on a16z you tube as well. 🎤

100x better is a big claim. What data are they going to use? The models are currently trained to the average human. The current architecture is bound by the average IQ of the data contributors. I think to get 100x better they will need self improving models that reference their own predictions and improve.



