Pretty much exactly what happened.

"The earth is degenerating today. Bribery and corruption abound. Children no longer obey their parents, every man wants to write a book, and it is evident that the end of the world is fast approaching." - Assyrian tablet, c. 2800 BC

Absolute awesomness. Please read this. All ML research has to be like this. Accessible to broad community, open source. Huge bonus: training on decentralized commodity hardware, low bandwidth, no expensive supercomputers required. I am a very big fan. Together AI

Thanks to everyone who participated in the NeurIPS Workshop on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems! At our poster session, I counted approximately 140 people - a spectacular success! Lots of work, but well worth it. gp-seminar-series.github.io/neurips-2022/

So discouraging to see this, independently of its original intent. “Abandon” (all) your research in favor of mine, seems like a bad “recommendation” to put on a slide. One can always be nice & acknowledge parallel efforts. Research shouldn’t be synonymous with polarization.

So machine learning researchers are happier than mathematicians?

Prompts are not Lipschitz. There are no “small” changes to prompts. Seemingly minor tweaks can yield shocking jolts in model behavior. Any change in a prompt-based method requires a complete rerun of evaluation, both automatic and human. For now, this is the way.

Max Tegmark Rishi Sunak Ursula von der Leyen Altman, Hassabis, and Amodei are the ones doing massive corporate lobbying at the moment. They are the ones who are attempting to perform a regulatory capture of the AI industry. You, Geoff, and Yoshua are giving ammunition to those who are lobbying for a ban on open AI R&D. If

⚡️ Speed up LLM inference by 5x. ⚡️ We introduce a new framework, coalescence, that makes structured generation several times faster than standard generation. Coalescence is very flexible, and raises unexpected questions 🧐 blog.dottxt.co/coalescence.ht…

“The fact that [transformer neural nets] model language is probably one of the biggest discoveries in history. That you can learn language by just predicting the next word with a Markov chain—that’s just shocking to me,” Mikhail Belkin says. By Will Douglas Heaven. technologyreview.com/2024/03/04/108…

People are overestimating the short-term potential of LLMs (productionizing is tough) while underestimating their medium-term potential. With a proper system (UX/UI/agents) around GPT-4-level models, GenAI is set to redefine human-computer interactions.

A good way to judge defensibility of a startup in GenAI is to ask the question “if a better model comes out tomorrow, does it make my product better or obsolete?”

Demos showing a LLM "generating" a working game of Space Invaders or Tetris with a single prompt would be *very* impressive if you could generate any game of similar complexity you might come up with. But you can't. You can only generate specific games that the model has

This is a remarkable advantage by Google AI. 1LOC grounding on the world's best search engine. Now available on the AI SDK.

SB 1047 has been amended, reflecting some of the changes suggested by Anthropic. This is likely the final version of the bill we will see before it goes to a vote in the Assembly. 🧵below with details.

"The researchers found that 24 of the brain samples, which were collected in early 2024, measured on average about 0.5% plastic by weight." theguardian.com/environment/ar…

It's not just that driverless Waymo vehicles get into far fewer serious crashes than human-driven ones. The crashes that do happen are overwhelmingly the fault of the other driver. Lots of human drivers rear-ending Waymos and running red lights. understandingai.org/p/human-driver…

2,850 days. Almost 8 years. That's how long it took to save a life. In August of 2016 my First Round partner, Bill Trenchard, led Ossium Health's Seed round. And in May of 2024, a 68 year-old woman in Michigan became the first patient to receive a bone marrow transplant from a

The largest impact of GenAI currently is probably automating the processing of PDFs. The number of man-hours spent worldwide extracting information from documents is wild.

hardmaru Sakana AI How about applying to the automated production of reproducibility reports of existing papers? This should be easier than producing entirely new papers, and be quite useful in that this is a type of work that we probably don't do enough as a community.