AI Race Heats Up as OpenAI, Google, and Mistral Unveil New Multimodal Models
• OpenAI, Google, and Mistral release new frontier AI models within 12 hours of each other as industry ramps up for summer • Releases come as sector prepares for next major version of GPT, which powers ChatGPT • Models accept text, image, audio, and video input, showing increasing multimodality • Mistral takes open source approach, publishing model code for free unlike OpenAI and Google • Meta's LeCun says large language models hitting limits, "objective-driven" AI that can reason needed for AGI