Welcome to the AI Ecosystem Report, featuring practitioner analyst and entrepreneur Toni Witt. This series is intended to deliver the timely intelligence about artificial intelligence (AI) you need to get up to speed for an upcoming client engagement or board meeting.
Highlights
Innovation (00:25)
ServiceNow, Nvidia, and Hugging Face announced their joint release of StarCoder 2, a family of open-access large language models (LLMs) designed for generating code. Hugging Face and ServiceNow have collaborated with the BigCode community for machine learning (ML) developers to share code, datasets, applications, and more. Nvidia joined this partnership to train the largest model within the StarCoder family that can generate application source code, review code, and assist human developers.
ServiceNow customers have a direct need for AI technology. The company can deliver that in a product format. Hugging Face not only contributes to the development of the model but also makes it usable and accessible through its hosting platform. Nvidia is powering the infrastructure, the technology, and the development of the StarCoder project. Together, the companies have produced a set of powerful cogeneration models that are open source, which the BigCode community and companies can learn from and apply.
Funding (02:26)
Figure is a startup building humanoid robots powered by AI. It recently raised a $675 million round at a $2.6 billion valuation from investors including Jeff Bezos, Nvidia, Microsoft, and OpenAI. Building these robots is a very capital-intensive undertaking; there’s a lot of machinery, engineering talent, and AI needed, which creates a huge upfront investment cost.
Founded in 2022, this startup has made great progress with developing its bot. The size of the bot is approximately 5 feet 6 inches and 130 to 140 pounds. It can lift a 45-pound load and run for five hours. Figure partnered with OpenAI to develop some of the models it’s running on the bot, as well as Microsoft Azure for its internal development team.
Solution of the Week (05:24)
Google released a new family of open-source LLMs called Gemma in partnership with Hugging Face. Gemma comes in two sizes: Gemma 7B has 7 billion parameters, which a consumer-grade GPU or TPU could handle, and Gemma 2B has 2 billion parameters which a standard CPU or smaller device could handle.
With this partnership, Google has continued to demonstrate its dedication to open-sourcing its technology. As AI models become more powerful, it’s useful to have open-source options to test the weights, training data, and other elements.
Ask Cloud Wars AI Agent about this analysis
