These are the top 5 AI developments in the last 24 hours

Hi friends, here are the 5 top stories in AI today. I did not include the announcement of new ChatGPT features, but you might find that valuable as well. The new multi-file upload capability still can't read PDFs and thus doesn't seem to do as good a job as Claude.

I can already see some good stories percolating up through our news discovery system for tomorrow, make sure to join us then!

Cheers, Marshall

Today’s stories:

Paul Graham on the Many AI Opportunities Still Available. Paul Graham, YCombinator's co-founder, said last night that while a majority of startups in the current YC batch are not focused on AI, a huge number of them are. "Not one I talked to seemed bogus. There are that many legit opportunities in AI." Asked how he determines a startup is bogus, he said his heuristic is that "the founders seem like opportunists and there is no connection between their backgrounds and what the startup is doing." [via @paulg]

IBM and NASA deploy open-source geospatial AI foundation model on Hugging Face
IBM and NASA partnered to launch the watsonx.ai geospatial model on Hugging Face. Trained on NASA's satellite data, it's optimized for tasks like burn scar and flood mapping. The model improves flood prediction by 15% due to quicker training and higher accuracy. [Venturebeat]

How MIT’s Liquid Neural Networks can solve AI problems from robotics to self-driving cars
MIT's Computer Science and Artificial Intelligence Laboratory has developed liquid neural networks (LNNs). The developers claim these compact, efficient systems surpass conventional deep learning models in some use cases, like with continuous data streams from robotics and self-driving cars. LNNs use adjustable differential equations, allowing them to adapt to new circumstances post-training while maintaining extreme efficiency. [VentureBeat]

Vicuna Models Now Based on LLaMA-2, Offers Open Language Model Alternative to GPT-4.
The student-run Large Model Systems Organization announced an updated version of the Vicuna model family, an open LLM alternative to ChatGPT and GPT-4, now on HuggingFace for commercial use. Meta's Yann LeCun notes the new version is now based on LLaMA-2. [via _philschmid]

Hugging Face Joins the PyTorch Foundation as a Premier Member
"The PyTorch Foundation, a neutral home for the deep learning community to collaborate on the open source PyTorch framework and ecosystem, is announcing today that Hugging Face has joined as a premier member." [PyTorch.org]

That's it!

Marshall