Taek

6 ms·

1.
by Taek 25d ago
What parts of the United States can you visit which have a >10% chance of being kidnapped or murdered? That means for every 100 visitors, 10 people don't come home. I honestly don't think that applies to a single neighborhood i
2.
by Taek 25d ago
I get that the political system could be better, but having been to places that actually are not governed by the rule of law, I can assure you that the US is doing quite well on that front. When there is actually no rule of law, consequence
3.
by Taek 26d ago
Definitely booting into a decoy system beats an automatic wipe.
4.
by Taek 26d ago
I do like the idea of forcing due process to access an encryption key. I'm not sure if that idea is compatible with current law, but it seems just at the very least.
5.
by Taek 1mo ago
I just wrote a utility to rip all comments out of the code. Now the code is fully uncommented and it has saved lots of input tokens and also lots of meandering because the model is no longer getting stuck on bad ideas it told itself about.
6.
by Taek 1mo ago
A job faire might serve you well!
7.
by Taek 1mo ago
I have a hard time looking at the world around me and thinking "close to no value"
8.
by Taek 1mo ago
Some sounds are genetically more intense than others, like shrill screeching sorts of noises. So not quite as bad as the euphemism treadmill, because there is biology that makes certain sounds fundamentally less terrifying than others.
9.
by Taek 1mo ago
Was on the 6th floor of a tower. Shaking lasted almost 2 minutes. There was no damage as far as I could tell (Medellín) but they still evacuated everyone to inspect the building. City is in a bit of chaos, people are fine but unsettled, and
10.
by Taek 1mo ago
This title is killing me. "The runtime stays open" is such a heavy, attention grabbing sentence, it's all I can stare at on the front page. It's the type of writing that makes me hate working with LLMs, because it's
11.
by Taek 2mo ago
That's a different bad UX pattern. If a button has already rendered in a certain location, a new button shouldn't replace it without first giving the user ample warning that a material change is about to happen.
12.
by Taek 2mo ago
I wish software apps had "tape-out rules" the way that computer chips do. Basically, when you design a computer chip, a program reviews the design and compares it against something like 300 pages of rules with stuff like "wir
13.
by Taek 3mo ago
I don't think it's that bad, if I recall correctly it's about 8 kilobytes per token, and a token can be 3-4 characters so you're talking ~2 kilobytes per character. An image token I recall is something like 16x16, so you
14.
by Taek 3mo ago
There's also Chinese models, which aren't trying to self-limit capabilities.
15.
by Taek 3mo ago
Is anyone in favor of this?
16.
by Taek 3mo ago
What is Krea's approach to content such as pornography and gore? It's been frustrating to see all of the leading models take a very hard line on excluding vice content, even when it is perfectly legal, in the name of safety.
17.
by Taek 3mo ago
I hope you don't mind if I get super pedantic here. But what qualifies as "a problem", and what counts as "has not encountered before" and what counts as "adapted solutions"? Because "a problem"
18.
by Taek 3mo ago
Can you define a clear boundary for me somewhere between a rock and a human brain at which consciousness is definitely not possible on one side of the boundary, but is maybe possible on the other side of the boundary?
19.
by Taek 3mo ago
Yes I agree. We have such a poor understanding of consciousness that really we can't even rule out whether a simple set of if-else statements is enough to create consciousness. Common sense says "surely not"... But where is t
20.
by Taek 3mo ago
If an LLM, which abstractly is just a stack of transformers, is able to reason about itself, why wouldn't a chess engine (also a stack of transformers) also be able to reason about itself? That reasoning may not manifest as English tha
21.
by Taek 3mo ago
That's where things get interesting. As soon as you start asking what counts as communication... what about signals passing between cells? What about heat passing between atoms? As soon as try and draw a firm line between "X count
22.
by Taek 3mo ago
On what grounds would someone establish than an LLM could be conscious but a sufficiently large/complex transformer model aimed at chess would not be conscious?
23.
by Taek 3mo ago
"we" is the set of people who believe that machines can potentially be conscious.
24.
by Taek 3mo ago
Another weird thing that keeps coming up - "people don't think that image models or chess models are conscious"... yes we do, and we have for many years. Or rather, we aren't *certain* that those things are conscious. Bu
25.
by Taek 3mo ago
Your comment is conflating a few things here, the philosophical problem of induction is one I am very familiar with, and it's also unrelated to what we are talking about. "If every piece of knowledge is created by combining previo
26.
by Taek 3mo ago
I think this argument is fragile, especially when you consider agents like Claws. Any LLM that is training on its own previous interactions with users has indeed directly experienced them. But also, what's the difference between having
27.
by Taek 3mo ago
It's not intuitive to me at all that the two are connected. Evolution adapted us to suffer because it is a powerful motivator for growth. It's not clear whether evolution adapted us to be conscious, or if it just happened.
28.
by Taek 4mo ago
I've gone back and forth with AI on this stuff quite a bit, and there are many, many theories of consciousness, which is why when you were vague about the "more core" concepts, I asked for which ones specifically. And I broad
29.
by Taek 4mo ago
Are you sure that AI-consciousness implies a responsibility to not make them suffer? Suffering is an evolutionary invention that motivates living things to improve themselves. But also, what qualifies as suffering to token prediction engine
30.
by Taek 4mo ago
Is it really so easy to assert that an AI doesn't have emotion? Lots of the AI models are capable of getting pissed off at the users, especially the earlier ones. And sure, their emotion is merely a bias that the context window generat

More ›