ETC Wrap for 14 Sep. 2025: “AI 2027” and More

1. Must read: Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean, AI 2027,” 4/3/25. Here’s their introduction: “We predict that the impact of superhuman AI over the next decade will be enormous, exceeding that of the Industrial Revolution. We wrote a scenario that represents our best guess about what that might look like. It’s informed by trend extrapolations, wargames, expert feedback, experience at OpenAI, and previous forecasting successes.”

2. Demis Hassabis, CEO of Google’s DeepMind, says that “given how fast AI is changing, even week by week, the only thing you can say for certain is that huge change is coming.” He “emphasized the need for ‘meta-skills,’ such as understanding how to learn and optimizing one’s approach to new subjects, alongside traditional disciplines like math, science and humanities. One thing we’ll know for sure is you’re going to have to continually learn … throughout your career” (Independent, 9/14/25).

3. “AI-powered drone swarms are set to transform the battlefield…. The technology, dubbed Nemyx, transforms individual drones into a single, co-ordinated force. Powered by Auterion’s operating system and delivered as an app, the system allows any compatible drone to join the swarm through a simple software upgrade….The swarms allow a single soldier to control multiple drones, allowing attack strategies that automatically seek to outfox and overwhelm enemy defences” (Financial Times, 9/13/25).

4. “What sets an agent apart from a simple LLM is its ability to access its environment, gathering information from APIs, sensors, online data and user interactions…. An agentic AI can tackle a single task far more comprehensively than a simple LLM. It can not only set objectives but also reason, plan, execute and monitor actions independently” (Economic Times, 9/14/25).

5. Tyler Lacoma argues that “AI” should be reserved for “the common generative AIs, typically powered by LLMs, that we see every day in the form of Google Gemini and ChatGPT, which can summarize information and ‘talk’ to us in conversational ways. Many [products] do have some generative capabilities, but calling them AI in the same vein as fully fledged chatbots is a stretch at best” (CNET, 9/14/25).

6. “In a Bloomberg survey of banks published last month, 70 per cent indicated generative AI would be widely used or critical to their business in the next two years, compared with 24 per cent now” (Financial Times, 9/13/25).

Leave a comment