Twice a week, Data Points brings you the latest AI news, tools, models, and research in brief. In today’s edition, you’ll find:
- A company that mines data to mine critical minerals
- Doctors use chatbots to deal with insurers
- A travel agent powered by AI
- An LLM health coach backed by OpenAI and Huffington
But first:
AI systems reconstruct images from brain activity with remarkable accuracy
Researchers at Radboud University used fMRI scans of human subjects and electrode array recordings from a macaque monkey, applying an improved AI system using a new technique they call predictive attention mechanisms. The key innovation is the AI’s ability to learn which specific brain regions are most informative for image reconstruction, allowing it to focus its attention on the most relevant neural signals. This research could lead to advanced brain implants for restoring vision by stimulating higher-level visual processing areas in the brain. (New Scientist)
New attention algorithm speeds up AI model processing
Together.AI released FlashAttention-3, a new algorithm that significantly accelerates the attention mechanism in large language models. The update achieves up to 75% utilization of an H100 GPU’s maximum capabilities, a substantial increase from the previous 35%. This advancement enables AI models to process longer text more efficiently and could lead to faster training times and improved performance for large language models. (Together.AI)
Mining firm uses AI to unearth massive copper deposit in Zambia
KoBold Metals uses an AI-powered database called TerraShed to identify promising mineral deposits, including valuable metals. The system integrates diverse data sources, including century-old paper maps, modern satellite imagery, and novel technologies like muon detectors. By analyzing huge amounts of this information, TerraShed can reveal previously undetected underground mineral formations. The company’s approach aims to make mineral exploration more effective and efficient as demand for battery metals increases worldwide. (The New York Times)
AI becomes doctors’ ally in insurance battles
Physicians are using AI chatbots to draft prior-authorization requests and appeal insurance claim denials more efficiently. Doctors like Dr. Azlan Tariq report significantly higher approval rates when using AI-generated letters, cutting down on time spent fighting insurers and improving patient care. This development raises concerns about a potential “AI arms race” between doctors and insurance companies, as both sides adopt the technology to streamline their processes. (The New York Times)
AI-powered travel planner designs specialized trip itineraries
Byway’s JourneyAI draws on multiple data sources to create customized flight-free itineraries, including transport timetables, fare information, and customer preferences. The tool analyzes data from previously successful trips to match new customers with similar traveler profiles and preferences. JourneyAI aims to design resilient multi-stop journeys by incorporating fallback options to manage potential disruptions along the route. (TechCrunch)
OpenAI and Huffington team up on health coach project
Sam Altman and Arianna Huffington are backing a new AI health coach that promises to offer personalized wellness advice based on scientific research and user data. The project, called Thrive AI Health, aims to nudge users toward healthier habits in areas like sleep and nutrition, with support from several medical institutions. While AI shows potential in healthcare, experts caution about privacy risks and the importance of maintaining human medical oversight. (The Verge)
Still want to know more about what matters in AI right now?
Read last week’s issue of The Batch for in-depth analysis of news and research.
Last week, Andrew Ng wrote about how current attempts to regulate AI models in California could put developers at risk:
“If this law passes, the fear of a trial by a jury — leading to a verdict that can be very unpredictable with significant penalties in the event of a conviction — will be very real. What if someone releases a model today after taking what they genuinely felt were reasonable safeguards, but a few years later, when views on AI technology might have shifted, some aggressive prosecutor manages to convince a jury that whatever they did was not, in hindsight, ‘reasonable’? Reasonableness is ambiguous and its legal interpretation can depend on case law, jury instructions, and common facts, among other things.”
Read Andrew’s full letter here.
Other top AI news and research stories we covered in depth included: Claude’s introduction of Artifacts, Amazon hires agentic talent from Adept, cloud computing companies rethink their climate goals, and GaLore, a new optimizer that saves memory during pretraining.