The week's developments in AI, quantum, and biotech, explained | 28.02.24

If you’re enjoying this briefing, sign up here — and share it with a friend. Reading time: ~4 minutes

Just days after the buzz surrounding Google's Gemini 1.5 release, another European player, Mistral, steps into the ring with its latest flagship model, Mistral Large. Hosted on Microsoft’s Azure AI, this powerful new model boasts "top-tier" capabilities in areas like reasoning, code generation, and multilingual support. But how does it compare to OpenAI’s ChatGPT or Google's Gemini?

Meanwhile, the humanoid robot race intensifies as Bezos, Nvidia, OpenAI, and other tech giants invest heavily in Figure AI, a startup striving to create practical, human-like robots. (The AI robotics industry has been busy as of late, remember 1X Technologies raising $100 million?).

Plus, Reddit sparks controversy with its new lucrative revenue stream: licensing its vast user-generated content to fuel AI training models.

Keep reading for the latest updates on Google’s new open-sourced AI models, the world's first text-to-interaction foundational model and a new computational algorithm that lets you see behind objects — plus the coolest AI tools to create, search, and integrate.

— Charlie and the Research & Intelligence Team 

P.S. You can now register your interest for CogX Festival LA 6-9 May, where we’ll be talking all things AI, Impact, and Transformational Tech!

Innovation and Releases

🤖 Mistral partners with Microsoft to unveil a new LLM model to rival OpenAI's ChatGPT. Just months following its initial Mistral 7B release, the Paris-based startup is launching its new open-sourced flagship model for developers, "Mistral Large".

🦾 Bezos, Nvidia, OpenAI and other tech giants back humanoid robot Startup. The startup Figure AI Inc. is raising over $675 million in funding from various big technology names amidst an industry-wide scramble to develop new AI applications for real-world tasks.

💸 Reddit leverages AI in business strategy ahead of IPO. Reddit Inc. announced that AI will play a significant role in its business operations moving forward. Additionally, the company disclosed a £60 million annual partnership with Google, granting the tech giant real-time access to Reddit's data to train its AI systems.

💎 Google unveils open-source AI models Gemma 2B and 7B for Developers. Gemma comprises a series of open-source AI models designed for language processing tasks, providing developers with a more accessible alternative to other flagship models.

🧑‍🚀️ UK offers £15m to space tech projects for commercial astronaut mission. Teaming up with Axiom Space, the Swindon-based UK Space Agency is gearing up for a mission to the International Space Station (ISS) featuring an all-UK astronaut crew.

AI Tools of the Week

🎥Transform:.Fliki is an AI-powered software that helps content creators transform text into videos with realistic voices. It features a rich stock media library and customisable subtitles. 

📧 Optimise: SkimAI is an AI-powered email assistant designed to enhance productivity by automating email management. It features AI summaries, draft responses, and inbox sorting.

🎨 Design: Logomark AI leverages Stable Diffusion models to offer an AI-powered logo creation tool, enabling users to generate custom logos through prompts.

Latest Research

⚛️ DeepMind and Quantinuum unveil a major advancement in quantum computing, the AlphaTensor-Quantum algorithm. By leveraging deep reinforcement learning, this technique optimises quantum circuits, particularly focusing on minimising the use of T gates.

🧠 Researchers seek consensus on what constitutes AGI. A team of researchers working at Google DeepMind have developed a new framework for evaluating AGI. They propose using different "Levels of AGI" to classify and measure the progression, autonomy, and capabilities of future AGI models.

👁️‍🗨️ Science fiction becomes reality: Researchers have crafted a novel method to bypass obstructed views. They have created a computer algorithm that generates highly accurate, full-colour 3D reconstructions of spaces hidden behind obstacles.

👾 Google Deepmind announces Genie, the first generative interactive environment model. Trained on over 200,000 hours of publicly available Internet gaming videos, the experimental model can turn any image or handmade drawing into a 2D platformer.

In case you missed it

The introduction of Genie by Google DeepMind could revolutionise human digital interactions by transforming any image into an interactive, playable experience. Check out the full story here:

✍️ Enjoying this newsletter?  Subscribe to our free weekly briefings on Preparing for AI, Cinema & AI, The Race to Net Zero, AI & DeepTech and The Future of Work.

🚀 Remember to register your interest for CogX Festival LA 6-9 May!