1,841 mentions · 834 chunks · 141 episodes
... designed to infer users' goals, including those you 'have not said out loud.'" LLMs allow qualitative insight at quantitative scale. That can be used for you or against you. That's why it's crucial that they be structurally aligned w...
LLMs are best used to fill in a scaffolding a human sets up. LLM generated stuff is like sand. A good space filler, but doesn't have enough shape on its o...
The ability to system-think allows people to get the compounding outcomes from LLMs. Systems thinking is meta-cognition. The ability to look at the problem from outside of it. To reflect. To get leverage on the problem from a meta le...
...is a natural place to do this, but those rules and expectations were set before LLMs were a thing. The basic expectations of the ecosystem are set. If the mobile OSes tried to structurally change it today, the app developers would scr...
LLMs have created the potential for Infinite Software. What's missing is the catalyst.
...orem, it's not even really a strategy per se. These OpenAI/Anthropic 'research' LLMs are permutation machines with budgets in the hundreds of millions of dollars. It would be more surprising if they couldn't string together something ...
...ctus with a straight face. It rests on so many fundamental absurdities. I think LLMs are the most powerful technology in decades if not centuries, but the prospectus still asks you to believe in more.
... reliability of the information. Humans, unit for unit, are less credulous than LLMs. The results that LLMs present have been flattened into the same smooth Claudish, hiding the contextual clues of the sources.
Mois Navon, a rabbi, pointed out that if you believe that LLMs are conscious, then you believe that the AI labs are slaveholders at an unprecedented scale. "A country full of geniuses… who will work for free." If...
Humans can be the long-poles for producing interesting material for LLMs to train off of. This is not because humans are smart per se, but because they are embodied. That is, they are in and of the world. Atoms, not bits. ...
LLMs can coordinate faster than humans… but that might mean they'll get stuck in local maxima more easily. The oddball individualistic humans can try the ...
There are various useful but inhumane management practices that now work for LLMs. For example: having two teams both attempt to solve a problem independently and then picking the winner. That used to deal psychic damage to the hum...
Git has only gotten more valuable with LLMs. "Just track it in Git" made sense for a lot of different kinds of knowledge work. But before, it was complex and had sharp edges… making it hard to ...
... ourselves to machines, but it should have been the other way around. Now, with LLMs, the machines can bend to us.
LLMs don't write tests for correctness of behavior but fixedness of implementation. When tests are expensive to produce, they are nearly always real. The ...
...decades to reach their heat death. Now, it's possible to reach it in weeks with LLMs spraying code. LLMs can do tech debt reduction… but only if asked to. If you don't ask them to do it, then you'll get more and more functionality by ...
... access sensitive information, execute malicious actions, and exfiltrate data." LLMs respond differently to harmful prompts when AI watermarking is used "SynthID can cause models to follow harmful instructions they would otherwise ref...
...quired millions, or hundreds of millions of users. But now you can do this with LLMs at orders of magnitude smaller scales. Instead of having to distill signals from data exhaust (since the human can't be asked to do anything more tha...
Another reason LLMs coordinate so naturally: they were selected for agreeableness. Coordination is easier when the participants are more agreeable.
...you have to struggle and hit your breaking point to be ready. Now in the age of LLMs you don't have to struggle; you have the trappings of work. Instead of leading you up the mountain, maybe that leaves you permanently at the foothill...