AI Top-of-Mind for 7.11.24 — The City by the Bay
Today — AI funding, xAI-Oracle nixed, Gradio, DreamFlare AI, physicians and chatbots, and do we really understand LLMs?
Top-of-mind is AI funding, and it should come as no surprise that the Bay Area is out in front. As reported by ‘Crunchbase,’ last year, 50% of all AI funding went to Bay Area companies.
Note — many who moved away from the Bay Area and SF in particular during COVID are now moving back, to be part of the AI ‘scene.’
Last week I covered the xAI-Oracle server deal, where the former was poised to spend up to $10B on Nvidia chip rentals. Well, per ‘The Information;’ the deal is off, and Musk is building his own data center which he claims can be built faster without Oracle’s involvement. Whether this is or will evolve into the previously announced ‘Gigafactory of Compute’ is unknown. And xAI continues to rent chips under an earlier deal.
In other vendor news, AMD is purchasing Silo AI out of Finland. The company is still trying to close the gap with Nvidia across both hardware and software.
On the model front, Simeon Emanuilov looks at MiniCPM-Llama3-V 2.5 (a real mouthful), an open-source MLLM developed by Tshighua University. Strong points include overall performance vs both Gemini and GPT-4V, as well as capabilities to avoid hallucinations via RLAIF-V.
But on a more sobering front, Alberto Romero concludes that no one knows how AI works. From his post:
If you ask the experts at the forefront of interpretability research, they readily admit it. Dario Amodei, CEO of Anthropic, says “Maybe we . . . understand 3% of how [neural nets] work.” Leo Gao, researcher at OpenAI, says it plainly: “We don’t understand how neural networks work,” a statement “strongly seconded” by Neel Nanda, lead of interpretability at Google DeepMind.
And, Thomas Reid writing in ‘AI Advances’ details ‘Gradio,’ a UI rapid prototyping tool for AI models. An example from his post:
If you are looking to increase your AI knowledge, some good tutorials from the VC firm a16z. Check out their ‘Emerging Architectures for LLM Applications.’ From the site:
Moving to healthcare, the ‘NY Times’ reports on how physicians are leveraging chatbots to cut through insurance paperwork. From the article:
· For a growing number of doctors, A.I. chatbots — which can draft letters to insurers in seconds — are opening up a new front in the battle to approve costly claims, accomplishing in minutes what years of advocacy and attempts at health care reform have not.
· Dr. Tariq said Doximity GPT, a HIPAA-compliant version of the chatbot, had halved the time he spent on prior authorizations. Maybe more important, he said, the tool — which draws from his patient’s medical records and the insurer’s coverage requirements — has made his letters more successful.
· Since using A.I. to draft prior-authorization requests, he said about 90 percent of his requests for coverage had been approved by insurers, compared with about 10 percent before.
Lastly, turning to the creative and entertainment, ‘Forbes’ reports on the launch of a new platform, DreamFlare AI, where subscribers can create AI-generated interactive experiences. Link to the launch video.
