Cerebras Systems, an artificial intelligence chip startup based in Silicon Valley, has released ChatGPT-like open-source models for the research and business communities to use for free in a bid to encourage more collaboration. The announcement was made on Tuesday, March 28th, and included seven models that are all trained on the company’s AI supercomputer called Andromeda.
The models range from smaller 111 million parameter language models to a larger 13 billion parameter model. Models with more parameters can perform more complex generative functions. For instance, OpenAI’s chatbot, ChatGPT, launched late last year, has 175 billion parameters and can produce poetry and research, generating broader interest and funding for AI.
According to Cerebras, the smaller models can be used on phones or smart speakers, while the larger ones run on PCs or servers, although complex tasks like summarizing large passages require larger models. Cerebras’ founder and CEO, Andrew Feldman, stated that “There’s a big movement to shut down what was open source in AI… it’s not surprising since there’s a lot of money in it now. The excitement in the community, the progress that we’ve made, was in large part because she was so open.”
However, Karl Freund, a chip consultant at Cambrian AI, said bigger isn’t always better. “Some interesting articles have been published showing that (a smaller model) can be accurate if you train it more,” Freund said. “So there’s a trade-off between bigger and better educated.”
Feldman said his largest model took just over a week to train, a job that usually takes several months thanks to the architecture of the Cerebras system, which contains a dinner plate-sized chip built for AI training that can last.
Most AI models today are trained on Nvidia Corp (NVDA.O) chips, but more and more startups like Cerebras are trying to partake in this market. The models trained on Cerebras machines can also be used on Nvidia systems for further training or customization, Feldman said.
This move by Cerebras is significant, as it challenges the trend towards proprietary models in the AI community, particularly in light of recent concerns about the concentration of AI power among a few dominant players. Cerebras’ release of ChatGPT-like models is a significant contribution to the open-source AI community, and may help to democratize AI technology by making it accessible to a broader range of researchers and businesses.
In conclusion, Cerebras Systems’ release of open-source ChatGPT-like models for AI computing is a welcome development for the research and business communities. The models’ range in size and capability means that they can be used for a variety of purposes, from smaller models used on mobile devices to larger models for complex generative functions. This move by Cerebras is an important step towards a more open and democratic AI community.
