This robot moves like Cristiano Ronaldo

Hello everyone and welcome back! Hope you had a wonderful time in the weekend! Here’s what you need to know about AI today:

👉 Sam Altman admits that AI benefits are not widely distributed

👉 Humanoid robots that move like athletes

👉 France is so back in the AI game with attracting billion of euros in investment

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: In a recent essay on his personal blog, OpenAI CEO Sam Altman has admitted that the benefits of AI may not be evenly distributed, at least initially. Altman suggests that OpenAI is considering unconventional solutions like a "compute budget" to ensure that AI benefits are accessible to everyone on Earth.

What does it mean: This concern is relevant as AI is disrupting the balance of power between capital and labor. Altman's vision includes running AI systems on a massive scale, potentially involving thousands or even millions of highly capable AI units across various knowledge work fields. He points out that while the development of advanced AI will require significant financial investment, the cost of using AI at certain levels decreases dramatically, approximately by 10 times every 12 months.

More details: 

• Altman has described AGI as “An AI system that can tackle increasingly complex problems, at human level, in many fields” which is different from his former definition, describing AGI as a system that can generate $100b in profits for OpenAI and Microsoft.

• OpenAI does not plan to end its relationship with close partner and investor Microsoft anytime soon by using the term AGI.

• There’s evidence that AI training and development costs are coming down, however massive investments will be needed to achieve AGI-level intelligence.

What’s going on: Carnegie Mellon University and Nvidia have collaborated and developed a system called Aligning Simulation and Real Physics (ASAP), enabling humanoid robots to perform (mimic) complex athletic maneuvers such as Cristiano Ronaldo's famous mid-air spin and Kobe Bryant's renowned fadeaway jump shot. (Watch full demos here)

What does it mean: Researchers have tried for years to create humanoid robots that move like humans, but their work has mainly concentrated on walking and locomotion rather than athletic actions. Recent research has added expressiveness to the upper body of robots, but making them “agile” is still tough because of hardware issues and differences between computer simulations and real-life physics.

More details:

• To tackle the issue of discrepancies between simulated and real-world physics, three main strategies have been developed: System Identification (SysID), domain randomization (DR), and learned dynamics methods.

• SysID estimates key physical parameters but relies on predefined spaces and torque measurements, which limits its adaptability.

• DR trains policies in simulation with randomized parameters but can result in overly cautious movements.

• Learned dynamics models use real-world data to enhance accuracy but remain unexplored for humanoid robots.

• ASAP method utilizes a two-stage approach that tackles the issue of dynamics mismatch by first training motion tracking in a simulated environment and then fine-tuning these policies with actual data from the real world.

• If you are interested to learn more and get into the details, read this.

👋 Do you enjoy reading so far? Be an AI ambassador by sharing the link below with your friends. You can also forward this email to them! Thanks!
the79newsletter.com/subscribe

🔓 DeepSeek-R1 AI model is reportedly more susceptible to jailbreaking attempts compared to other AI models.

🖼 Google has introduced a new AI protection feature for Google Photos, using invisible SynthID watermarks to indicate when images have been edited with AI tools.

💰 Brookfield, Canadian investment firm, plans to invest €20 billion by 2030 in AI projects in France. The total investments in French AI ecosystem is now close to $85 billion.

🧬 Anthropic's CEO stated in a recent interview that DeepSeek models performed the worst on a critical bioweapons data safety test. Anthropic, which is backed by tech giants such as Amazon and Google, has a strong reputation for its influential research in the field of AI safety.

🏢 OpenAI wants to open an office in Munich, Germany, to expand its European presence. They have stated Germany's technical expertise, academic excellence, and industrial innovation as key reasons for this strategic move.

🤖 The UAE plans to invest between €30 to €50 billion to construct a massive AI data center in France, with a capacity of up to 1GW, as part of a new AI campus initiative.

🧠 Andrew Ng, the founder of Google Brain and one of the most influential figures in AI, supported Google's decision to drop its pledge against building AI systems for weapons.

💸 Safe Superintelligence, an AI company started by ex-OpenAI chief scientist Ilya Sutskever, is negotiating to secure investment with a valuation of around $20 billion.

AI + learning a new language

Capgemini - Machine Learning Engineer

Meta - AI Research Scientist, Language - Wearables

Torc - Senior, Software Engineer - Developer Environments

Box AI - Machine Learning 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!