AI race heats up: OpenAI, Google, Mistral release new models in one day

Open source vs Closed Door: Debate rages over sharing powerful new AI models

By
Web Desk
Frontier of AI? Experts question if language-focused models represent true intelligence. — Reuters
Frontier of AI? Experts question if language-focused models represent true intelligence. — Reuters

The AI industry witnessed unprecedented competition on Wednesday with OpenAI, Google and French startup Mistral, all unveiling their most advanced AI models, within a mere 12 hours of each other, The Guardian reported.

Such a progression came at a time before Meta was to release Llama 3 and GPT-5, both August inbound, the latter being a successor to GPT-4.

The models now introduced have stunning capabilities. GPT-4 Turbo from OpenAI and Gemini Pro 1.5 by Google are multimodal chatbots that can accept both text and image data as input.

But Gemini goes a step ahead, doing not only the audio but also video. However, it's worth noting that, among all these helpful apps, Mixtral 8x22B, which was free, open-source and can be downloaded, is the only one. That stands out to Meta as well. Such an approach enables anyone to create and use a pre-trained model.

However, this has generated discussions about the possible dangers of bioengineering, as geneticists face less control over possible uses. Critics warn that all the issuers of this technology can use it for bad purposes or do nothing if the vulnerabilities or sides are revealed.

The Information has sources saying that Meta will roll out the lite, less adept versions of the Llama 3 models for early release and will explicitly display the top-of-the-class model in the summer. 

Everyone thinks that OpenAI is also moving along the same path with GPT-5 as the COO of the company Brad Lightcap hinted about the upcoming release in an interview with the Financial Times.