Details of Elon's $97b offer to buy OpenAI will surprise you

Hi AI lovers! Here’s what you need to know about AI today:

👉 Details of Musk’s bid to buy OpenAI is now revealed

👉 Researcher have designed life-like enzymes with AI from scratch

👉 Anthropic might release Claude 4 in weeks

and many more!

📧 Did someone forward you this email? Subscribe here for free to get the latest AI news everyday!

Read time: 5 minutes

What’s going on: As you know, Elon Musk and a group of investors, have made an offer to acquire OpenAI for $97.4 billion. However, OpenAI CEO Sam Altman has publicly dismissed the offer. Despite these public statements, OpenAI's board is legally obligated to consider such offers. The details of Musk's offer are now revealed through legal filings and indicate that there is a clear deadline for this offer, the payment is supposed to be all-cash, Musk wants to have access to full financial records of OpenAI as a due-diligence routine, and the bid contradicts Musk’s legal claims that the OpenAI’s assets can’t be “transferred away” for “private again”.

What does it mean: The timing of Musk's offer raises questions about its true intent. OpenAI's legal team suggest the offer is a strategic maneuver to undermine a competitor. Musk's team, however, maintains the offer is serious and that the cash infusion would benefit OpenAI's non-profit mission. Adding another layer of complexity, Musk has indicated he would withdraw the offer if OpenAI abandons its plans to convert to a for-profit model. This condition suggests that the offer might be a tactic to influence the valuation of OpenAI as it moves towards privatization.

More details: 

• The offer letter specifies a deadline of May 10, 2025, for acceptance.

• The proposed acquisition would be an all-cash transaction. However, Musk is known for debt financing for major acquisitions and he won’t just use his own money in these kinds of deals, as he previously used the same tactic to buy Twitter back in 2022.

What’s going on: Scientists at the University of Illinois and the University of Washington have achieved a significant breakthrough in enzyme engineering by using AI to design entirely new, life-like enzymes from scratch. These AI-designed enzymes are capable of performing multi-step reactions, a crucial characteristic of natural enzymes that has been a challenge to replicate artificially.

What does it mean: This achievement marks a major advancement towards the practical application of engineered enzymes like plastic recycling. In previous attempts, we were only able to create enzymes that could only manage a single step of a reaction. The researchers overcame this limitation by combining several machine-learning approaches. This innovative approach has yielded enzymes that are 60,000 times more effective at accelerating the target reaction compared to enzymes designed using previous methods that new enzymes were often created by relying on modifying existing structures.

More details:

• In this study, the researchers focused on designing enzymes that could perform serine hydrolysis, a 4-step chemical reaction crucial to many biological processes.

• They employed a combination of AI tools, including RFdiffusion, which generates new enzyme structures, and PLACER, a deep neural network that refines these structures by modeling the interactions between the enzyme and the molecules involved in each reaction step.

• These newly designed enzymes are not yet as efficient as their natural counterparts though.

• Are you interested in getting into the details of this study? (Read their published paper)

🚀 Anthropic is reportedly set to launch a new hybrid AI model with a potential name, “Claude 4“, in the coming weeks, designed to offer variable reasoning capabilities.

🧠 Google's Gemini now features a recall capability that allows it to tailor responses based on previous conversations.

🤖 Apptronik secures $350 million in funding from Google and others to develop humanoid robots for various applications.

⚡ A recent study suggests that by implementing demand-response strategies, data centers in the US could unlock an additional 76 gigawatts of power capacity by temporarily reducing their power draw during peak times. (Read the paper)

🖥 Arm is set to launch its first in-house designed chip this year, with Meta as one of its initial customers. It is a significant shift as Arm moves from just licensing designs to manufacturing its own semiconductors.

📽 YouTube Shorts has integrated its Veo 2 video model, enabling creators to generate AI video clips with enhanced realism and speed. It is now available initially in selected countries (US, Canada, Australia, and New Zealand) with plans for wider rollout.

AI + Regex design

Motive - Senior Data Scientist

AECOM - Data Scientist II

Zefr - Senior Machine Learning Engineer

Arine - Senior AI Engineer

Thank you for staying with us like always! If you are not subscribed, subscribe here for free to get more of these emails in your inbox! Cheers!