🔮 The Sunday edition #505: Mirroring life; robot training; murder by streetview; 3D plane, influencer MPs & bunker sales surge++

An insider's guide to AI and exponential technologies

👋🏼 Hi, it’s Azeem. Each week I share my view on the evolving world of AI and exponential technologies to help you understand what is going on and what to pay attention to – this is the last Sunday edition of 2024!

Two remarkable developments drew my attention this week. First, a new Stanford report delivers a warning that we may be nearing the creation of “mirror life” – organisms built from reversed molecules that could potentially outcompete natural biology. Meanwhile, fresh research questions the idea that artificial intelligence evens the playing field. Let’s get into it!

---

2025 Horizon Scan – Live community briefing

I’m hosting two sessions in January with members of Exponential View. I’ll present my 2025 forecast and open the space to discuss major opportunities and risks ahead. Join me on one of these days:

• 8 January, 4pm UK / 11am ET REGISTER HERE

• 9 January, 11am UK / 3pm Dubai time REGISTER HERE

---

Ideas of the week

Beyond human – o3’s arrival

OpenAI previewed a new reasoning model, o3. The new model is good at reasoning in mathematical and programming domains. It scores really well on a particular benchmark called ARC-AGI, which is above the average human. Spend more money on it, up to $3,500 per task, and it approaches the performance of a STEM graduate and it’s better than virtually every human coder. We’re witnessing a shift in how quickly AI capabilities advance and how they may reshape our understanding of complex domains. My full commentary on what o3’s arrival means is here.

Reflected Eden

A new 300-page Stanford report warns that advances in synthetic biology and chemistry are eroding the barriers that previously made ‘mirror bacteria’ impossible. ‘Mirror life’, also known as ‘mirror cells’ or ‘chiral life’, refers to the hypothetical organisms created in a lab that have mirror image molecular structures of various natural organisms. Life on Earth uses mostly “left-handed” amino acids and “right-handed” sugars. In contrast, a “mirror cell” would be built from molecular components with the opposite orientation, creating a lifeform that is chemically a perfect mirror image of conventional life. Human bodies, and other natural organisms, may not be able to detect, digest, infect or effectively interact with this new form of life. However, there are many cases in which mirror life could be useful. Mirror cells could be used to produce drugs in contamination-free environments, for instance. They could also improve chemical synthesis, creating specialty compounds with greater precision and fewer unwanted byproducts. But, the risks are significant. Economist AT Robin Hanson warned in 2010: “If it gained energy via photosynthesis, or via special adaptations that enable it to eat ordinary life […] it could take over much of the biosphere.”

Do the robot – millions of times over

I warmed up a bit to the progress in robotics this year. Many challenges in real-world navigation are being helped by the genAI boom. A new generative physics engine released this week, Genesis, can produce 4D dynamical worlds for general-purpose robotics and physical AI training. It’s not the first one, but it’s currently the fastest engine for robot learning. Its simulation speed is ~430,000 faster than in real-time. It takes only 26 seconds to train a set of controls for the real world on a single GPU.

Who really gains from AI?