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 H2O AIa company that helps enterprises develop AI systems today announced the launch of two fully open source products: a generative AI product called H2OGPT and a no-code development framework called LLM Studio.
The offering, available today, provides enterprises with an open, transparent ecosystem of tooling to build their own instruction-following chatbot applications, similar to ChatGPT.
It’s because more and more companies want to use generative AI models for business use, but remain wary of the challenges associated with sending sensitive data to a centralized Large Language Model (LLM) provider that runs a proprietary model behind a API serves.
Many companies also have specific model quality, cost, and desired behavior needs that closed offers do not meet.
How do H2OGPT and LLM Studio help?
As H2O explains, the no-code LLM studio provides enterprises with a fine-tuning framework where users can simply go in, choose from fully licensed, commercially usable code, data and models – ranging from 7 to 20 billion parameters, 512 tokens – and start building a GPT for their needs.
“One can take open assist-type datasets and start using the base model to build a GPT,” Sri Ambati, the co-founder and CEO of H2O AI, told VentureBeat. “They can then refine it for a specific use case using their own dataset, adding additional tuning filters such as specifying maximum prompt length and answer length or comparison to GPT.”
“Essentially,” he said, “you can build your own GPT with every click of a button and then publish it back to Hugging facewhich is open source, or internal to a repo.”
Meanwhile, H2OGPT is H2O’s own open-source LLM – refined to plug into commercial offerings. It’s just how Open AI ChatGPT offers, but in this case the GPT adds a much-needed layer of introspection and interpretability that allows users to ask “why” a certain answer is being given.
Users on H2OGPT can also choose from a variety of open models and datasets, including viewing response scores, highlighting issues, and adjusting length.
“Each company needs its own GPT. H2OGPT and H2O LLM Studio empower all of our customers and communities to create their own GPT to help improve their products and customer experiences,” said Ambati. “Open source is about freedom, not just free. LLMs are way too important to be owned by a few big technology giants and nations. With this important contribution, all our customers and community can work with us to make open-source AI and data the most accurate and powerful LLMs in the world.”
Currently, about half a dozen companies are splitting the H2OGPT core project to build their own GPTs. However, De Ambati was unwilling to disclose specific customer names at this time.
Open source or not: a point of discussion
The H2O supply comes more than a month later Databricksa well-known Lakehouse platform, took a similar step by releasing the code for an open-source large language model (LLM) called Dolly.
“With $30, one server and three hours, we can teach [Dolly] to start doing human-level interactivity,” said Ali Ghodsi, CEO of Databricks.
But as efforts to democratize generative AI in an open and transparent manner continue, many still favor the closed-loop approach, starting with OpenAI – which hasn’t even disclosed the contents of its training set for GPT-4 – doing so. referring to competitive landscape and security implications.
“We were wrong. Flat out, we were wrong. If you believe, as we do, that AI – AGI – will be extremely, incredibly powerful at some point, then there’s just no point in using open source,” said Ilya Sutskever, chief scientist and co-founder of OpenAI. Forget in an interview. “It’s a bad idea… I fully expect that in a few years it will be perfectly clear to everyone that open-source AI just doesn’t make sense.”
Ambati, for his part, agreed with the possibility of malicious use of AI, but also emphasized that more people are willing to do good with AI. The abuse, he said, could be addressed with precautions such as AI-driven curation or some sort of audit.
“We have plenty of people who want to do some good with open source AI. And that is why democratization in this way is a necessary force,” he noted.
VentureBeat’s mission is to become a digital city plaza where tech decision makers can learn about transformative business technology and execute transactions. Discover our Briefings.
