Ok, so… today is crazy. There's a mysterious Mistral-Next model that just dropped, and I am testing it for coding—it's GOOD. Possibly GPT-4 level. I even ran a few tests that I asked GPT to evaluate the outputs, and it gave Mistral-Next a perfect score.
RESEARCH
-
AI Needs Spirit Beyond Movable Type Technology
By
–
The printing press was replacing woodcarvings with movable type. AI is going to replace the construction of reality in the mind, but movable type is not enough. This technology is fundamentally movable type, it lacks intention. The AI needs to be infused with spirit.
-
AI Lecture Series: Mines Paris Alumni and Addendum
By
–
[#Event] "Artificial Intelligence, Intelligence of the Future" the lecture series by Mines Paris Alumni and Addendum https://actuia.com/actualite/%80intelligence-artificielle-intelligence-du-futur-le-cycle-de-conferences-de-mines-paris-alumni-et-addendum/
… #AI #artificialintelligence
@MinesParisAlumn -
AI and Machine Learning Synergy Infographic Guide
By
–
Unveiling the synergy between AI, Machine Learning, and more with this crisp infographic by @appypie
. Follow @ingliguori for a deeper dive into each field! #AI #MachineLearning #DeepLearning #NLP #ComputerVision -
Model Capabilities and Limits: Ground Truth Dissolves
By
–
This particular loop captures perfectly what the model is good at and where it fails. It's also a metaphor for postmodernism, but meanwhile just stare at it until the illusion of ground truth dissolves https://t.co/OotYElgJkf
— Joscha Bach (@Plinz) 16 février 2024This particular loop captures perfectly what the model is good at and where it fails. It's also a metaphor for postmodernism, but meanwhile just stare at it until the illusion of ground truth dissolves
-
AI Still Lacks Actual Groundwork and Manual Labor Capabilities
By
–
But please look at the actual leg work, in no way done by the AI yet
-
OpenAI Sora milestone prediction assessment February 2024
By
–
(I don't consider OpenAI Sora to have hit this mark yet. Though there's still 6 months until August 2024 to prove me wrong on the early side of this prediction.) https://
x.com/Im_actuallyaca
/Im_actuallyacat/status/1758268812640538946
… -
ICL vs SFT: Scale and long context favor In-Context Learning for larger models.
By
–
SFT of smaller models will be around for a while I’m sure, but those are smaller models. With each increase in model scale the tradeoff between ICL and SFT tips in favor of the former, and even more so with long context I think.
-
LLMs exhibit distinct intelligence pathological in unique ways
By
–
LLMs and diffusers exhibit intelligence distinct from animal intelligence, better in many ways worse in many ways, pathological in distinct ways. Unfortunately the use of a single term "intelligence" confuses people; "meme" personalities arbitrage this to fame lol
-
Gemini 1.5: 10M tokens, 100K-shot, prompting eats everything
By
–
Gemini 1.5 — 10M tokens and the demos to prove it can use them. So many implications here. Why SFT when you can 100K-shot? And if it can translate Kalamang given a grammar and dictionary, what *can’t* the right words teach it? Prompting is poised to eat everything.