Join C-suite executives in San Francisco July 11-12 to hear how leaders are integrating and optimizing their AI investments for success.. learn more
H2O AI, a California-based company that helps companies 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. bottom.
Available today, the product provides enterprises with an open and transparent ecosystem of tools for building proprietary, command-and-return chatbot applications similar to ChatGPT.
This has occurred as more companies look to deploy generative AI models for their business use cases, but they have been forced to leave sensitive data to centralized large language model (LLM) providers who provide their own models behind APIs. I am still wary of the challenges that come with sending a .
Also, many companies have specific needs for model quality, cost, and desired behavior that a closed product cannot meet.
event
transform 2023
Join us July 11-12 in San Francisco. There, he shares how management integrated and optimized his AI investments to drive success and avoid common pitfalls.
Register now
How can H2OGPT and LLM Studio help me?
H2O describes the no-code LLM Studio as providing a fine-tuning framework for enterprises, allowing users to create fully acceptable and commercially available code, data and models (ranging from 7 to 20 billion parameters, 512 tokens). ) to select from. For their needs he starts building GPT.
“You can get an Open Assist type dataset and start building GPT using the base model,” H2O AI co-founder and CEO Sri Ambati told VentureBeat. “You can then use your own dataset to fine-tune it for your specific use case, or add additional tuning filters such as specifying maximum prompt and answer length or comparing to his GPT .”
“Essentially, you can click a button to build your own GPT and publish it to the open source Hugging Face or to an internal repository,” he said.
H2OGPT, on the other hand, is H2O’s own open-source LLM, fine-tuned to connect to commercial products. This is exactly how OpenAI offers ChatGPT, but in this case GPT is a much needed introspection and interpretability that allows users to ask “why” they get certain answers. add more layers.
H2OGPT users can also choose from a variety of open models and datasets, review response scores, flag issues, and adjust length.
“Every company needs its own GPT. H2OGPT and H2O LLM Studio will enable every customer and community to create their own GPT to improve the product and customer experience,” said Ambati. rice field. “Open source is more than just free, it means freedom. LLM is too important to be owned by a few tech giants or nation states. The community will be able to partner with us to make open source AI and data the world’s most accurate and powerful LLM.”
Currently, about six companies have forked the core H2OGPT project and built their own GPT. However, Ambati was reluctant to name specific customers at this time.
Open Source or Not: Debatable
The H2O product offering comes more than a month after Databricks, known as the Lakehouse Platform, made a similar move to release code for an open-source Large Language Model (LLM) called Dolly. .
“For $30, a server, and 3 hours, we can teach.” [Dolly] It’s about initiating human-level interactivity,” said Databricks CEO Ali Ghodsi.
However, as efforts to democratize generative AI in an open and transparent manner continue, many, including OpenAI, have not even made public the content of their GPT-4 training sets, citing the impact on the competitive environment and safety. of people still vouch for a closed approach. .
“We were wrong. Quite frankly, we were wrong. If so, it makes no sense to open source it,” OpenAI chief scientist and co-founder Ilya Sutskever told The Verge. in an interview. “That’s a bad idea…I fully expect that in a few years it will be completely clear to everyone that open sourcing AI is unwise.”
Ambati agreed that AI could be abused, but also emphasized that more people are using AI to do good deeds. He said the abuse could be addressed with safeguards like AI-driven curation and a kind of check.
“There are a lot of people who want to make good use of AI in open source, and that’s sort of why democratization is a necessary force in this way,” he pointed out.
VentureBeat Mission will be the digital town square for technical decision makers to gain knowledge and transact on transformative enterprise technologies. Watch the briefing.
