OpenAI, Google, and French artificial intelligence startup Mistral released new versions of their frontier AI models within 12 hours of each other, as the industry prepares to ramp up activity over the summer.
The unprecedented flurry of releases comes as the industry prepares for the anticipated launch of the next major version of GPT, the system that underpins OpenAI’s hit chatbot Chat-GPT.
The first came just hours after Nick Clegg appeared on stage at an event in London, where he confirmed that the third version of Meta’s own AI model, Llama, would be published in a few weeks.
Seven hours after Clegg left the stage, Google’s Gemini Pro 1.5, its competitor’s most advanced large-language model, was released to the general public, with a free tier limited to 50 requests a day. .
An hour later, OpenAI released the final version of its frontier model, GPT-4 Turbo. The GPT-4 Turbo and Gemini Pro 1.5 are “multimodal” systems, capable of accepting more than just text. Each can take image input, while the Gemini can also accept audio and video.
Earlier this morning in France, Mistral, an AI startup founded by several of Clegg’s former colleagues on Meta’s AI team, released its frontier model, the Mixtral 8x22B. Unlike its two American competitors, Mixtral was released via a simple download link of a 281GB file: the company, like Meta, takes an “open source” approach, making its AI systems available for anyone to download. Publishes free to load and build.
This approach has been criticized as potentially dangerous, as it leaves developers unable to intervene and prevent their systems from being used for malicious purposes, nor to pull models offline. If weaknesses or biases are detected and need to be corrected. Others, including Metta, defend it as ultimately leading to better outcomes than systems “held in secret by a small number of very, very large, well-heeled companies in California”.
The Miata’s Lama 3 is expected to be released initially in its smaller, less powerful version, in preparation for the release of the company’s latest Frontier model this summer. However, it may face stiff competition: OpenAI is believed to be planning a similar timeframe for its next GPT model, the GPT-5, according to the company’s chief operating officer, Brad Light. Cap told the Financial Times that it was coming “soon”.
However, experts have questioned whether the “big language model” approach shared by all frontier AI systems is reaching its limits. “We hear a lot of people say: ‘Oh my God, we’re going to get. [artificial general intelligence] Within the next year,” said Yann LeCun, Meta’s chief AI scientist, responding to xAI founder Elon Musk’s claim. “It’s just not happening. We have AI systems that can pass the bar exam, but they can’t clean your dinner table and load the dishwasher. We have systems that manipulate language, and fool us into thinking they’re smart, but can’t understand the world.”
After newsletter promotion
Instead, LeCun suggested, researchers need to work on “purpose-oriented” AI with the ability to reason and plan about the world, rather than just working on words.
LeCun said this approach could produce AI systems with truly superhuman capabilities. It’s “more of a vision than anything,” he added, but it’s progressing quickly, so I’m pretty sure we’ll get there.