Join top executives in San Francisco on July 11-12 to hear how leaders are integrating and optimizing AI investments for success.. Learn more
based in california H2OAIa company that helps companies with the development of AI systems, today announced the release of two open source products: a generative AI product called H2OGPT and a no-code development framework called LLM Studio.
The offerings, available starting today, provide businesses with an open and transparent ecosystem of tools to build their own instruction-following chatbot applications similar to ChatGPT.
It comes as more and more companies look to adopt generative AI models for business use cases, but are wary of the challenges associated with sending sensitive data to a centralized large language model (LLM) provider serving a proprietary model. behind an API.
Many companies also have specific needs for model quality, cost, and desired behavior that locked-in offers fail to meet.
How do H2OGPT and LLM Studio help?
As H2O explains, the no-code LLM Study provides enterprises with a fine-tuning framework that users can simply jump into, choose between fully permissive and commercially usable code, data, and models (ranging from 7 to 20 billion parameters, 512 tokens) and start building a GPT for your needs.
“One can take open assist type data sets and start using the base model to build a GPT,” Sri Ambati, co-founder and CEO of H2O AI, told VentureBeat. “Then they can tune it for a specific use case using their own dataset, as well as add additional tuning filters, such as specifying the maximum request length and response length or comparing to GPT.”
“Essentially,” he said, “with each click of a button, you can create your own GPT and then republish it to hug facewhich is open source, or internally in a repository.”
Meanwhile, H2OGPT is H2O’s own open source LLM, perfected for incorporation into business offerings. it’s like like open AI ChatGPT provides but, in this case, GPT adds a much-needed layer of introspection and interpretation that allows users to ask “why” a certain answer is given.
H2OGPT users can also choose from a variety of open datasets and models, view response scores, flag problems, and adjust length, among other things.
“Every company needs its own GPT. H2OGPT and H2O LLM Studio will allow all of our clients and communities to create their own GPT to help improve their products and customer experiences,” Ambati said. “Open source is about freedom, not just free. LLMs are too important to be owned by just a few tech giants and nations. With this important contribution, all of our clients and the community will be able to partner with us to make AI and open source data the most accurate and powerful LLM in the world.”
Currently, about half a dozen companies are forking off the core H2OGPT project to build their own GPTs. However, Ambati was not willing to reveal specific client names at this time.
Open source or not: a matter of debate
H2O deals arrive more than a month later data bricksa well-known Lakehouse platform, made a similar move by releasing the code for an open source large language model (LLM) called Dolly.
“With $30, a server, and three hours, we can teach [Dolly] to start doing interactivity on a human level,” said Databricks CEO Ali Ghodsi.
But as efforts to democratize generative AI in an open and transparent way continue, many still support the closed approach, starting with OpenAI, which hasn’t even declared the content of its training suite for GPT-4, citing the big picture. competitive and security implications. .
“We were wrong. We were flat out wrong. If you believe, as we do, that at some point, AI (AGI) is going to be extremely, incredibly powerful, then there is simply no point in open source,” said Ilya Sutskever, chief scientist and co-founder of OpenAI, at Edge In an interview. “That’s a bad idea… I hope that in a few years it’s completely obvious to everyone that open source AI isn’t smart.”
Ambati, for his part, agreed with the possibility of AI being misused but also stressed that there are more people willing to do good with AI. Misuse, he said, could be handled with safeguards like AI-powered curation or some kind of verification.
“We have enough humans who want to do good with open source AI. And that is why democratization is a necessary force in this way,” he noted.
VentureBeat’s mission is to be a digital public square for technical decision makers to gain insights into transformative business technology and transact. Discover our informative sessions.
