It would be interesting to add a "Debunk" or "Deflated" section, because the more money we inject into AI, the more "fake it until you make it" there is, and this becomes a bit dangerous for people, as some may believe things work, when they don't work at all.
There's a fierce and interesting review of Github Copilot Workspace there:
https://www.youtube.com/watch?v=75Hv0RUFIrQ
It would be interesting to add a "Debunk" or "Deflated" section, because the more money we inject into AI, the more "fake it until you make it" there is, and this becomes a bit dangerous for people, as some may believe things work, when they don't work at all.
The link to "Better & Faster Large Language Models via Multi-token Prediction" should be:
https://arxiv.org/abs/2404.19737