In the Hellaswag LLM benchmark evaluating common sense natural language inference, Danube performed with an accuracy of 69.58%, sitting just behind Stability AI’s Stable LM 2 1.6 billion parameter model.
Chandrasekar explained that through the OverflowAPI, Google now has continuous access to the APIs that pull public data from Stack Overflow. These APIs enable access to the same data available to the Stack Overflow community via its public APIs.
Microsoft unveils AI-powered finance assistant Copilot for Finance, promising major efficiency gains through workflow automation but faces data privacy and security challenges.
Couchbase is aiming to differentiate itself by positioning its support of vectors as part of its overall platform approach which spans from cloud to on-premises data centers, and even on mobile and edge devices.
Filigran is now instrumental in organizing and analyzing cybercrime data, managing the modeling of multiple ransomware campaigns and enhancing response to cyber incidents.
Alibaba researchers have developed a new AI system called EMO that animates portraits with fluid and expressive talking and singing motions perfectly synced to audio, significantly advancing the state-of-the-art in controllable video synthesis.
Daron Acemoglu, David Autor, and Simon Johnson, faculty co-directors of the new MIT Shaping the Future of Work Initiative, describe why the work matters and what they hope to achieve.
While it remains to be seen how well StarCoder2 models perform in different coding scenarios, the companies did note that the performance of the smallest 3B model alone matched that of the original 15B StarCoder LLM.
Gradial, an AI startup streamlining marketing workflows, raised $5.4M led by Madrona; the generative platform automates content updates so creative teams can focus on strategy rather than tedious tasks.
V-JEPA uses the same rule of learning through observations, referred to as “self-supervised learning,” which means that V-JEPA does not need human-labeled data.
Samba-1 is not a single model, like OpenAI's GPT-4, rather it is a combination of more than 50 high-quality AI models put together in an approach that SambaNova refers to as a Composition of Experts architecture.