We Live in Scary Times
We live in scary times.
This morning, an Anthropic researcher claimed to have solved a $1m Millennium Prize problem in mathematics with Claude, following a similar claim from OpenAI, though neither proof has been independently verified yet. Between Sundar Pichai’s recent interviews and rumours on X, Google appears to be pointing to post-training gains that are pulling labs onto the early edge of recursive self-improvement (RSI). OpenAI is likewise rumoured to be accelerating research faster than ever, with agents driving more of the R&D loop under human supervision. Empirically, progress doesn’t seem to be plateauing.
Technically, today’s models are already capable enough to replace large swathes of knowledge work. People are just slow to adopt them, and regulatory, liability, and data privacy constraints are slowing rollout further. Across every level of the corporate hierarchy, frontier models can already do real work, and in many cases the bottleneck isn’t capability so much as learning how to get the most out of them.
The water is already up to our necks, yet most of us remain oblivious because it has risen so quietly.
Even if today’s LLM paradigm isn’t AGI on its own, it may still be what gets us there.
There’s one silver lining. Problems that have plagued society for millennia could finally be solved, including chronic and previously incurable diseases like cancer and Alzheimer’s. Our quality of life may improve dramatically, even as what gives life meaning changes significantly.
For now, all we can do is enjoy life as much as we can, hold on to hope, and accept that whatever we think is coming still underestimates what actually will.
- #ai
- #agi
- #rsi