Excellent ...for those (like me) who have no clue how to drink from the arXiv fire-hose nor the context to judge importance. YES, all the #LLM summaries are 3 (three) sentences! AI Research Highlights In 3 Sentences Or Less (June -July 2023), by Sebastian Raschka open.substack.com/pub/sebastianr…

Noted by #TheSequence, this #BAIR paper gives math insights into #SSL training behavior based on firm #PCA concepts. Pay attention to those eigenvalues! It's the secret sauce for LLMs. On Stepwise Nature of Self-Supervised Learning bair.berkeley.edu/blog/2023/07/1… gitlab.com/generally-inte…

Hilary Mason Simple (but fascinating) question! Obvious answer is your (cosine-sim) app for your vector database. Harder answer is how you understand structural relationships (manifolds?) among object embeddings, especially within local neighborhoods. Smashing 1000s dim into 2? 👎

VERY INFORMATIVE vis of social/political/economic impact of the Internet over 3 decades. Top 10 website. Note scale! Starts Jan-1993 with 20M views for world pop of 5.5B. Ends 29 years later with 80B views - 4000x usage with 49% pop growth. Wikipedia! ...start in Dec 2005.

IMPRESSIVE!!! If you really want to know the currently accepted facts (and its many mysteries) for the beginning and end of our Cosmos... I just backed ENCYCLOPAEDIA COSMOLOGICA (RELAUNCHED) on Kickstarter kickstarter.com/projects/cosmo…

Fascinating #AI startups... Check out #TxtAI, an "all-in-one embeddings database" wrapping vector-indexing with useful services. Nice repo/docs with new Medium channel. Now, how to make hype-free money? github.com/neuml/txtai medium.com/neuml/introduc…

Best column for highly-current substantive technical coverage of LLMs. Excellent insights into LLaMA2. If you understand this column's implications, you are riding the #AI wave crest... Ahead of AI #11: New Foundation Models, by Sebastian Raschka magazine.sebastianraschka.com/p/ahead-of-ai-…

Had fun with ChatGPT4 analyzing MNIST images. It FAILED to recognize images as 5-5-8, guessing 5-3-2. Did great on explaining overall analysis. "Image 3 is "2" b/c has a loop at base, which is not common in typical "2"s, and has straight slanting line descending from top loop."

Check out my latest article: Decision Support to Organizational Semantics... summarizes a journal paper on research issues for AI-assisted decision support in organizations. linkedin.com/pulse/decision… via LinkedIn

Interesting analyst briefing with Salesforce/Tableau on their evolving analytic semantic models. Nontrivial work to connect classic data semantics with 'GenAI ready' semantics so that humans and AI agents can properly collaborate. OpenAI, Google AI, & the like should take note.

I got my Kaggle certificate for the Google 5-Day Agent Intensive course and capstone project. I was impressed by Google's thoughtfulness in their Agent Developer Kit (ADK) towards enterprise-level agentic implementations. kaggle.com/learn-guide/5-…

The capstone project for Google 5-Day Agent Intense was a kick because I teamed with my son Eric Hackathorn on a "AI-for-Good" Zyra project to streamline workflow utilization of NOAA's massive scientific data sets. github.com/NOAA-GSL/zyra