Cerebras Systems introduces its series of seven GPT-based large language models that can be used by research or commercial organizations to build ChatGPT-like ones.
Heaptalk, Jakarta — Artificial intelligence (AI) computing startup, Cerebras Systems, announced the release of open-source ChatGPT-like models (03/28). These models can be used for research or commercial ventures without royalties as it was released under the industry standard Apache 2.0 license.
The open-source models include a series of seven GPT-based large language models (LLMs). Cerebras has trained its LLMs, starting from 111 million to 13 billion parameters, on the 16 CS-2 systems in its AI supercomputer, namely Andromeda.
According to Cerebras, the smaller models can be deployed on phones or smart speakers while the bigger ones run on PCs or servers, although complex tasks like large passage summarization require larger models, as reported by Reuters.
Meanwhile, ChatGPT developed by OpenAI still has a larger parameter of 175 billion. The AI chatbot can generate poetry and research, resulting in large attention and capital to AI extensively.
Keeping AI as an open technology
Founder and CEO of Cerebras, Andrew Feldman, saw efforts to block open sources in artificial intelligence. The launch of these open-source LLM models demonstrates Cerebras’ commitment to keeping AI as an open technology.
“There is a big movement to close what has been open-sourced in AI. It is not surprising as there is now huge money in it. The excitement in the community, the progress we have made, has been in large part because it is been so open,” said Andrew as quoted by Reuters.
According to Co-founder and Chief Software Architect at Cerebras, Sean Lie, several organizations can train large-scale models, but only a few of them is capable to do it on dedicated AI hardware.
“Releasing seven fully trained GPT models into the open-source community shows just how efficient clusters of Cerebras CS-2 systems can be and how they can rapidly solve the largest scale AI problems – problems that typically require hundreds or thousands of GPUs. We are very excited to share these models and our learnings with the AI community,” dconcluded Sean.
