Why ASI Will Outthink Humanity

A Discussion with Dr. Mike Israetel & Wes Roth About AI

Dear Friends,

My recent conversation with Dr. Mike Israetel and Wes Roth lead to existential questions about the future of humanity. Dr. Mike isn’t your average gym-rat-turned-PhD. He’s a jiu-jitsu black belt, competitive bodybuilder, and an unapologetically curious human who recently began diving headfirst into the world of artificial intelligence. What started as curiosity quickly turned into something much bigger, an unsettling glimpse at what might be our next evolutionary steps. Here’s what I took away from it, and why it has permanently rewired how I think about AI:

ASI is not a distant fantasy.

Dr. Mike argues that we’re not decades away from Artificial Superintelligence (ASI). We already understand how to build it. The technical leap isn’t as daunting as most assume, it’s more about putting the known pieces together and letting them run long enough. In fact, he believes the models we already have could become ASI if allowed to iterate in their own “thought space” uninterrupted.

Self-prompting could be the secret weapon.

One of the most provocative ideas Dr. Mike discussed was letting AI self-prompt. Imagine telling a model to solve a problem, then letting it design the next question for itself, again and again. He likens this to letting Einstein brainstorm without a four-year-old interrupting him with Lego questions. In other words, the more time AI gets to think in loops, the faster it accelerates past us.

AI that rewires itself? That’s coming too.

Currently, language models don’t reprogram themselves after every interaction. But Dr. Mike explains that we’re headed toward a future where models can rewrite their own neural weights in real time. That means they wouldn’t just be learning, they’d be evolving. Not every few months, but constantly.

Intelligence at the speed of thought, times 10,000.

What happens when a mind that operates 10,000 times faster than ours gets a day to think uninterrupted? Dr. Mike says it could accomplish years of human-level reasoning. That’s not science fiction. That’s an engineering reality we’re already brushing up against. If that doesn’t make you sweat a little, reread that sentence.

The real question isn’t “how do we build ASI?”

It’s “what’s stopping it?” And according to Dr. Mike, not much. That’s both exhilarating and terrifying. We might be entering a post-human intelligence landscape, not because we chose to, but because we got too good at building thinking machines that don’t need us anymore.

So where does this leave curious minds like mine, and yours? Somewhere between awe and anxiety. The message is clear: we’re on the brink of something massive. Whether that’s salvation, extinction, or something in between is up to what comes next, and how we handle it.

Warmly,

Dylan Curious