naasking

29 ms·

511.
by naasking 1y ago
* Humans are not deterministic. * Humans that make mistakes are still considered to be reasoning. * Deterministic algorithms have limitations, like Goedel incompleteness, which humans seem able to overcome, so presumably, we expect reasonin
512.
by naasking 1y ago
I get it, but it's not trivial to be precise enough at this point to avoid all false positives and false negatives.
513.
by naasking 1y ago
No you can't. You're just not proving what you think you're proving.
514.
by naasking 1y ago
> The valid chains of reasoning you speak of (assuming you are talking about text you see in a “thinking model” as it is preparing its answer) are narratives, not the actual reasoning that leads to the answer you get. It's funny tha
515.
by naasking 1y ago
> Do not confuse the mathematical description of physical processes as the world being made of math. Again, this doesn't apply to information. A simulation of a computation really is equivalent to that computation. > One is sim
516.
by naasking 1y ago
> They're not thinking, it's doing math, mathematical transformations of data Whatever thinking fundamentally is, it also has an equivalence as a mathematical transformation of data. You're assuming the conclusion by sayin
517.
by naasking 1y ago
> But it's not the LLM that makes modifications in those databases - it just retrieves data which is already there. So what? > I'm not following you here. If you're time travelling, you're resetting the state of th
518.
by naasking 1y ago
It too isn't rigourously defined. We're very much at the hand-waving "I know it when I see it" [1] stage for all of these terms. [1] https://en.wikipedia.org/wiki/I_know_it_when_I_see_it
519.
by naasking 1y ago
> In any case LLMs most likely are not minds due to the simple fact that most of their internal state is static. This is not a compelling argument. Firstly, you can add external state to LLMs via RAG and vector databases, or various othe
520.
by naasking 1y ago
> But a simulated mind is not a mind. This was already debated years ago with the aid of the Chinese Room thought experiment. Yes, debated and refuted. There are many well known and accepted rebuttals of the Chinese Room. The Chinese Roo
521.
by naasking 1y ago
> I don't need to know how the brain works in detail to know claims that humans are doing the same things as LLMs to be similarly silly. Yes you do. It's all computation in the end, and isomorphisms can often be surprising.
522.
by naasking 1y ago
> LLMs do a good job of teasing it out that resembles reasoning/thinking but really isn't. Given the fact that "thinking" still hasn't been defined rigourously, I don't understand how people are so confident
523.
by naasking 1y ago
> don't get why you would say that. it's just auto-completing. https://en.wikipedia.org/wiki/Predictive_coding > If it thought and reasoned, it would know to ignore those and focus on the variables and
524.
by naasking 1y ago
You're just assuming that mimicry of a thing is not equivalent to the thing itself. This isn't true of physical systems (simulated water doesn't get you wet!) but it is true of information systems (simulated intelligence i
525.
by naasking 1y ago
> To simply disregard that entire conceptual history and say, “well it’s doing a thing that looks like thinking, ergo it’s thinking” is the lazy move. What’s really needed is an analysis of what thinking actually means, as a word. Unfort
526.
by naasking 1y ago
I'm not on the moderation bandwagon to begin with per the above, but if an organization invents a bunch of fake reasons that they find convincing, then any system they come up with is going to have its flaws. Ultimately, the goal is to
527.
by naasking 1y ago
> But HATEOAS really isn’t about storing state in URLs. I think saying they are unrelated isn't correct either. In order for hypermedia to be the engine of application state, the continuations of your application must be reified as
528.
by naasking 1y ago
Wasn't IE6 the worst offender here? I think URL lengths can be longer now, although I doubt 4096 chars is portable.
529.
by naasking 1y ago
I'm skeptical: use two different AIs which don't share the same weaknesses + random sample of manual reviews + blacklisting users that submit adversarial inputs for X years as a deterrent.
530.
by naasking 1y ago
Setting aside the wisdom of moderation, instead of banning AI, use it to accelerate review.
531.
by naasking 1y ago
So what? People are experimenting with novel tools for review and publication. These restrictions are dumb, people can just ignore reviews and position papers if they start proving to be less useful, and the good ones will eventually spread
532.
by naasking 1y ago
There has been work that has pushed performance too, like tiny recursive models. Applying LLMs in recursive loops also improves output, so efficiency improvements make this viable, which can count as improvements in performance.
533.
by naasking 1y ago
At some threshold, efficiency gains let models move out of the data center though.
534.
by naasking 1y ago
> How likely are we to NOT see the AI data center apocalypse through better algorithms? Near certain IMO. Algorithmic improvements have outpaced hardware improvements for decades. We're already seeing the rise of small models and ho
535.
by naasking 1y ago
"The astronomer may speak to you of his understanding of space, but he cannot give you his understanding." ~ Kahlil Gibran ~
536.
by naasking 1y ago
> Yet, if the rest of the market does not react to the signal, the one lowering their prices hurts their profits and possibly kicks themselves out of the market. The reason this doesn't happen is because the ones lowering their pric
537.
by naasking 1y ago
Using pixels is still tokenizing. What's needed is something more like "Byte Latent Transformers", which has dynamically sized patches based on information content rather than tokens.
538.
by naasking 1y ago
> Well it turns 100k inserts into tens of millions. I don't think "number of inserts" is the right metric because the total amount of bytes being written is almost the same, it's just written in different areas and st
539.
by naasking 1y ago
> The last thing I want is a database that returns different columns per row. Actually that's exactly what you'd want, because it saves you from running two different queries in those cases with properly normalized disjoint dat
540.
by naasking 1y ago
> They also have a probably 30-40 year head start. Holy exaggeration Batman!

More ›