Global AI Sector Heats Up with Major Releases from OpenAI, Google, and Mistral

OpenAI, Google, and the French startup Mistral have all unveiled new versions of their advanced AI models, signaling a surge of innovation within the technology sector.

This wave of releases occurred within a span of 12 hours, setting the stage for what is anticipated to be an active summer for the AI industry. This series of launches precedes the expected introduction of the latest iteration of GPT, OpenAI’s foundational system behind its popular Chat-GPT chatbot.

The sequence of announcements began shortly after Nick Clegg’s presentation in London, where he announced the upcoming release of Meta’s third version of its AI model, Llama. Google swiftly followed with the release of Gemini Pro 1.5, its latest large language model, offering limited free access. OpenAI then introduced GPT-4 Turbo, a multimodal system capable of processing image inputs, closely mirroring Google’s Gemini Pro 1.5’s capabilities, which also handles audio and video inputs.

In France, Mistral launched Mixtral 8x22B, adopting an open-source approach by making it freely downloadable, a strategy shared by Meta but viewed by some as risky due to potential misuse or the discovery of biases that cannot be quickly mitigated.

Meta plans to release Llama 3 in stages, starting with less powerful versions before unveiling its most advanced model later in the summer. This approach may position it as a direct competitor to OpenAI’s forthcoming GPT-5, anticipated to be released in a similar timeframe.

As the industry progresses, experts, including Meta’s chief AI scientist Yann LeCun, caution against the limits of current “large language model” strategies. LeCun advocates for the development of “objective-driven” AI that can reason and plan, moving beyond mere language manipulation to achieve truly superhuman capabilities.