CONVERSATIONAL COMPUTING’S FUTURE I love how we’re mass producing research and d

CONVERSATIONAL COMPUTING’S FUTURE
I love how we’re mass producing research and development. And it’s at a scale and speed that we’ve never seen before even in the previous generations of computational innovation. I’ve lived through everything from teletypes to the present and worked through the previous AI winters. There is just nothing like this.

One reason is that llms let us use our own language to teach these machines rather than machine language to teach us how to teach them. 😉 It’s a multiplier on previous code frameworks.

But while LLMs are great a synthesis they are still sh** at analysis (cognitive wayfinding). We are already seeing good work in multiple predictions and adversarial competition between them. That’s great at reducing hallucination and is the first layer of falsification.

But until we can discover a goal, then wayfind to it, we can’t do sufficient falsification to perform reasoning, protocols, logic, or math. It’s amazing what they’ve done with programming, But programming is a tiny world model compared to reality and it’s hierarchy of world models.

Which means we need even more compute. 😉

Reply addressees: @jasonyuandesign @jxnlco