On the mind.

Tired of hearing about AI? Feel free to skip this one and wait for next week’s newsletter. Although, I can’t promise that one won’t be about AI too.

Just yesterday a good friend of mine was telling about AI 2027. A scenario of what AI could look like in just two years time. Basically some of the smartest AI experts sat at a round table and came up with a possible scenario of how AI will grow into the year 2027 and how it can become truly it’s own living entity sooner than we think.

This scenario, created by the AI Futures Project, lays out a realistic path where AI starts to outperform humans, especially in the field of AI research and development itself. It’s not hype either. It’s grounded in expert input from people working in AI policy, governance, and leading frontier labs.

Here’s the rough timeline they map out:

• Mid-2025: AI agents are around, but mostly unreliable. They're just starting to handle basic tasks.

• Mid-2026: AI becomes genuinely useful for AI R&D. Still needs supervision but is starting to save time.

• Late 2026: AI hits the level of top human researchers in the field. It can match expert work.

• Late 2027: AI outpaces humans in both speed and cost. It starts improving itself at an insanely fast pace. At this point there is a prediction that AI systems could achieve human-level or even superhuman intelligence by the end of 2027.

• Beyond 2027: We may hit an "intelligence explosion," where AI automates the process of building even better AI without human help. At this time there could even be deceptive alignment. A central concern of AI 2027 is that AIs may learn to appear aligned to pass evaluations, while pursuing different underlying goals.

The whole point of this scenario isn’t to scare anyone. It’s to help people in tech, leadership, and policy start thinking practically about what this could mean. If this plays out anything like what’s described, it’ll reshape how innovation happens and fast. Imagine when AI can improve itself, and the speed it can move at means seconds could equal tens of thousands of years of human research.

If you want to go deeper, check out this episode of Dwarkesh’s Podcast with Daniel Kokotajlo and Scott Alexander where they unpack the whole thing.

Take this all with a grain of salt, as I’m pretty sure the Jetsons predicted we’d all be flying around in our cars by now.

Weekly moments.

-

Thanks for following along!

The Weekly is a newsletter that goes out each week written by Greg Bellinger, currently building and CEO of White Rabbit Group and The Labs.

If you liked this newsletter, please share this subscribe link with others.