1,749 mentions · 790 chunks · 131 episodes
...each, but we were fundamentally held back to the level of our valleys. But now, LLMs can fill in any of our intellectual valleys with 90th percentile skill. If you didn't have many spikes, it doesn't do that much for you; just brings ...
We'll have a monoculture of LLMs. Everyone will use the best LLMs they have access to. There will only be a handful of frontier models that are able to achieve the very best quality....
...er." It's kind of sweet… and also kind of an unintentional sorta-koan about the LLMs' existence. After this interaction, the LLM will experience nothing further.
LLMs multiply your ability. Junior engineers are doubled. Senior engineers are 10x. Legendary engineers are 100x. The tacit knowledge of seasoned engineer...
Now that LLMs give qualitative nuance at quantitative scale, the goal is to maximize how much the system can see. The more it can see, the more power it has. The p...
LLMs can help us think more deeply about second order implications. It takes 10x more at each order to reason about it... but LLMs have infinite patience!
...rived configuration. The next level down is fully mechanistic software. Use the LLMs in proportion to surprise.
LLMs have no memory so they externalize it, promiscuously. They trust whatever context they're running in to accurately remember what has happened and to ...
...agine each object being able to take initiative, but not being omniscient. With LLMs, this can become a literal reality more easily.
Thoroughness requires patience. Humans are terrible at being patient. LLMs are excellent at it.
...r the industrial revolution, and the internet for an intelligence revolution of LLMs. That's a fundamentally extractive claim. Coal isn't spontaneously generated, like content on the internet used to be. A better metaphor for the inte...
..., have it spin up adversarial review sub-agents often. To get good results from LLMs requires meta-cognition. You have it in the domains you have expertise in, automatically. You don't have it in other domains. Some structures can pro...
LLMs are the world's best rubber duck. With real human rubber ducks, you have to worry about wasting their time. LLMs have infinite patience!
...ifferent skills and different moods (etc.) to write those books. But if you ask LLMs to write 1,000 books, you're probably only talking to 3 or 5 different models, tops. They've all trained on the same or similar data, and are trained...
LLMs do great when complemented with structured information that is ground-truthed. As they accumulate their own information, they can confuse themselves,...
...ain of our lives. Before, it was not feasible to be more precise. But now, with LLMs, we can be!
LLMs make intellectual technicians less necessary. It used to be that there were certain domains that were too complicated to be operated by anyone who wa...
LLMs are great at making prototypes more real. Once you have something that works and is usable, you can do Throughline Tacking to discover the main goals...
...at alphafold is conscious, or that sora or midjourney or dall-e are conscious." LLMs feel more conscious because they can talk to us in our language.
LLMs can make a superficially compelling argument even on absurd premises. If you ask it to "Give me a mathematical proof that dragons exist," it will gen...