Mistral CEO confirms ‘leak’ of new open source AI model nearing GPT-4 performance

🤖 Mistral CEO confirms ‘leak’ of new open source AI model nearing GPT-4 performance LINK

• Mistral's CEO Arthur Mensch confirmed that an 'over-enthusiastic employee' from an early access customer leaked a quantized and watermarked version of an old model, hinting at Mistral's ongoing development of a new AI model nearing GPT-4's performance.

• The leaked model, labeled "miqu-1-70b," was shared on HuggingFace and 4chan, attracting attention for its high performance on common language model benchmarks, leading to speculation it might be a new Mistral model.

• Despite the leak, Mensch hinted at further advancements with Mistral's AI models, suggesting the company is close to matching or even exceeding GPT-4's performance with upcoming versions.

🧪 OpenAI says GPT-4 poses little risk of helping create bioweapons LINK

• OpenAI released a study indicating that GPT-4 poses at most slight risk in assisting in the creation of a bioweapon, according to their conducted research involving biology experts and students.

• The study, motivated by concerns highlighted in President Biden’s AI Executive Order, aimed to reassure that while GPT-4 may slightly facilitate the creation of bioweapons, the impact is not statistically significant.

• In experiments with 100 participants, GPT-4 marginally improved the ability to plan a bioweapon, with biology experts showing an 8.8% increase in plan accuracy, underscoring the need for further research on AI's potential risks.

💥 Meta's plans to crush Google and Microsoft in AI LINK

• Mark Zuckerberg announced Meta's intent to aggressively enter the AI market, aiming to outpace Microsoft and Google by leveraging the vast amount of data on its platforms.

• Meta plans to make an ambitious long-term investment in AI, estimated to cost over $30 billion yearly, on top of its existing expenses.

• The company's strategy includes building advanced AI products and services for users of Instagram and WhatsApp, focusing on achieving general intelligence (AGI).

🍎 Tim Cook says big Apple AI announcement is coming later this year LINK

• Apple CEO Tim Cook confirmed that generative AI software features are expected to be released to customers later this year, during Apple's quarterly earnings call.

• The upcoming generative AI features are anticipated to be part of what could be the "biggest update" in iOS history, according to Bloomberg's Mark Gurman.

• Tim Cook emphasized Apple's commitment to not disclose too much before the actual release but hinted at significant advancements in AI, including applications in iOS, iPadOS, and macOS.