News
1. Next generation of AlphaFold
The latest version of AlphaFold demonstrates significantly improved accuracy and expands its capabilities beyond proteins to include other biological molecules like ligands, nucleic acids, and those with post-translational modifications (DeepMind).
2. A small AI model is improving
Microsoft researchers added multimodal capabilities to their small, open-source AI model Phi 1.5, showing that less expensive models can have features similar to OpenAI’s massive GPT-4 (Semafor).
3. Generating images from MEG signals
Meta developed an AI system that can generate images from brain activity recorded with magnetoencephalography signals. The images lack fine details but show it’s possible to decode visual information from the human brain (Meta).
4. Biden’s AI Executive Order
The order tasks federal agencies with developing more detailed AI guidance and requires companies developing powerful AI systems to disclose safety testing results, but does not specify consequences for unsafe models (Time).
5. Fact-checking images
Google has launched a new set of tools to provide users with more contextual information about images to combat the spread of misinformation. These tools allow users to view an image’s history, metadata, and how it has been described on different sites, which can help debunk false claims (TechCrunch).
6. Waymo driverless vehicles become available
Uber has partnered with Waymo to offer driverless vehicle rides in the Phoenix area, a service first reported in May 2023. This collaboration allows users to hail a Waymo autonomous car through the Uber app (TechCrunch).
Articles
1. Big problems that demand bigger energy (MIT Tech Review)
Some hardest problems worth solving with technology, selected by top experts in different fields.
2. Innovative new cell therapies could finally get at tough-to-target cancers (MIT Tech Review)
After years of limited progress, new CAR T cell therapies are showing encouraging response rates in early trials against solid tumors, but researchers are still working to optimize factors like targeting, precision, and safety. Approaches include targeting unique proteins, engineering ‘logic gate’ T cells, adding tumor tags, and controlling T cell activity.
3. The Skills Your Employees Need to Work Effectively with AI (HBR)
Interviews with company leaders suggest the following: invest more in two important human skills – interpersonal skills to communicate and engage with others, and domain knowledge to maximize the value of AI tools.
