🔮 Sunday edition #503: AI barriers fall; computational war; autonomous microbes; TikTok election, med-tattoos & black hole puzzles ++

An insider's perspective on AI and exponential technologies

Hi, it’s Azeem. The future of warfare and intelligence is unfolding in concert. Silicon Valley’s most advanced AI models are being integrated into military systems. Meanwhile, scientists have created autonomous bacterial sensors that can detect TNT in soil – that’s the synth bio escalation point I’ve been waiting for. Let’s dive in!

---

Ideas of the week

Solving gnarliness

At $200 per month, OpenAI’s new Pro tier for ChatGPT’s o1 reasoning system sets a new standard for an AI assistant. By one benchmark, which measures PhD-level science understanding, o1 pro is substantially better than a typical human expert. I have been using it and found it impressive on harder research questions and a good companion to Claude. Would most people find it helpful? For now, likely not. In many situations, most people are likely satisfied by a cheaper, less capable model. But, I stress “for now”, because a powerful system like this can help with gnarly problems that we don’t even bother trying to address because of their gnarliness. If you are an entrepreneur or work with highly complicated issues, this more expensive tier is probably worth looking into.

This week, we looked at the evolution from AI assistants to agents. We foresee that billions of agents will serve us in the background in the near future and this will change what we do and how we work.

To work effectively with an AI infrastructure we all need to become managers. Just as executives must learn to delegate and verify work, interfacing with an AI chief-of-staff will demand clear judgment about tasks, goals and overall intelligent oversight. This isn’t about outsourcing our thinking; we’ll need to elevate how we manage our AI and our human coworkers.

Putting microbes to work

The most exciting thing about synthetic biology is that it’s only just getting started. The development of an autonomous microbial sensor capable of detecting TNT in natural soil is described in a new Nature paper this week. This innovation combines tools like computationally designed riboswitches and genetic memory circuits that enable the bacteria to detect and respond to trace explosives over extended periods of time. Unlike traditional detection methods, these biosensors are self-sustaining, scalable and adaptable. This could enable a future where biotechnological solutions dynamically adapt and mitigate negative human impact on the environment. As the technology evolves, one can imagine it being used to monitor levels of toxic chemicals in our waters and soil or to predict hazardous events such as gas or oil leaks.

Two sides of a transformation

The AI landscape is undergoing a shift that both lower barriers to entry and redefine where value lies.