OpenAI to Release New Open-Source AI Model: Can it Compete with GPT?

Friday - May 19 2023, 07:17 UTC - 1 year ago

OpenAI is set to release a new open-source AI model soon in response to the growing competition and increased interest in generative AI. The move marks a shift towards accessible AI development and raises concerns over security and misinformation as OpenAI CEO Sam Altman appeared before a U.S. Senate panel to discuss the risks and limitations of AI.

In what seems like a response to the growing competition in the open-source large language model (LLM) space, OpenAI will soon release a new open-source AI model to the public, reported "The Information".

OpenAI hasn’t come up with an open-source model since 2019, and although the news is exciting, it might not be as sophisticated or in direct competition with its proprietary model GPT.

The report further said that OpenAI’s $27 billion private valuation depends on a future in which the most sophisticated AI for commercial purposes isn’t open source.

--- The move marks a shift towards accessible AI development --- .

Pressure is also mounting from its rival Meta, which released a bunch of open-source models in February. OpenAI’s move marks a significant shift towards a more democratic development in the world of AI, as more developers choose to opt for free models.

Ever since the launch of the AI chatbot ChatGPT, a deep learning model which allows one to have human-like conversations with a bot, there has been an explosion of interest in generative AI. Microsoft invested billions of dollars in OpenAI, the developer of ChatGPT, which paved the way for more competition with its rival, Google, which recently released a plethora of AI advancements in its i/o conference.

In the long run, it would be interesting to see how open-source models fare in comparison to proprietary models. However, the proponents and critics of AI are discussing a much larger question today.

--- Concerns over AI – misinformation and security --- .

Yesterday, OpenAI CEO Sam Altman appeared before a U.S. Senate panel to discuss the risks and limitations posed by AI in a larger context. Altman was asked to give his views on if AI should be subject to licensing, and he said a model that can persuade or manipulate a person's beliefs would be an example of a "great threshold." .

In addition, he said that firms should have the freedom to say that they do not want their data used for AI training. However, material on the public web would be fair game, added the executive.

OpenAI has previously published open-source models like Point-E, Whisper, Jukebox, and CLIP. Whether its new open-source model will be better than its competitors remains to be seen.