AI's puzzle mastery and an AI-guided voyage to Alpha Centauri
The newsletter highlights how AI models have rapidly improved at solving puzzles like the New York Times Connections, though humans still outperform them in some cases. It also reports on the Fermi Explorer Mission's plan to launch a spacecraft toward Alpha Centauri by 2029, using a trajectory discovered by an AI system from PSI. Additionally, it touches on evolving perspectives on human space exploration as discussed in recent books.
The newsletter’s puzzle test highlights a rapid leap in AI capability: by early 2025, top models solved New York Times Connections puzzles nearly perfectly, up from just 18% in late 2024. Yet humans still outperform AI on certain tasks, and the curated set of seven puzzles aims to reveal where models stumble. Separately, the Fermi Explorer Mission plans a 2029 launch toward Alpha Centauri, following a trajectory generated by PSI’s AI system. The journey could take up to 80,000 years, and the mission may also probe questions about alien life. The piece also notes that recent books frame human spaceflight as an extension of our innate drive to explore, with astronauts and anthropologists offering varied justifications for leaving Earth.
This story suggests AI’s rapid puzzle-solving gains may shift public expectations of machine intelligence, but its uneven performance underscores that human cognitive strengths remain relevant. The Alpha Centauri mission, while extremely long-term, could inspire renewed interest in interstellar science and AI-driven discovery. However, the reported restriction of OpenAI’s next model due to “critical” cyber risk highlights a growing tension: as AI becomes more capable, its potential for harm may require tighter controls, affecting developers, users, and regulators who must balance innovation with safety.