AI Reading for Thursday March 21, 2024

Microsoft's hire of Suleyman may reflect a desire to hedge dependency on OpenAI and have their own heavy-hitting internal AI product team - Business Insider

---

GitHub rolling out beta of AI autofixes for advanced security customers. When security scanner finds a possible bug like not sanitizing inputs, it can generate a PR with a proposed fix - InfoWorld

---

Dev says your private GitHub repos may be The Stack of LLM training data and you might want to check. - post.lurk.org

Hacker News discussion. If your GitHub is public momentarily, a scraper is likely to grab it, which doesn’t seem something most people would want.

---

Nvidia's new Blackwell to use same node as Hopper, opening a door for Intel if its bleeding-edge node can ramp up, giving a significant edge to Intel's Gaudi GPU or even an opportunity to manufacture Nvidia chips despite longstanding rivalry - The Motley Fool

---

LLM model targets multimodel understanding of scientific papers including complex dataviz, reasoning and processes

---

WSJ on open-source LLMs

---

Cohere seeks $5b valuation on $22m annual revenue run rate.

---

QuietSTaR prompting improves LLM reasoning by having the AI generate multiple theses and then pick the best one - livescience.com

---

Friends don’t let friends upload their private parts to an AI STI test - Engadget

---

Using generative AI to match ads with content improves performance - Business Insider

---

Neuralink livestreams brain-computer interface patient. Impressive, but hey Grok, list all the companies considered more advanced than Neuralink in BCI?

---

ChatGPT store fills up with sketchy apps, from de-plagiarizing your papers to impersonating your favorite celebrity to imitating copyrighted content.

---

Key Stable Diffusion researchers leave controversial AI company - Forbes

---

A(nother) template for structured prompting - FlexOS

---

Best gen-AI education tools - Forbes

---

Follow latest AI headlines via SkynetAndChill.com on Bluesky