lelanthran

4 ms·

31.
by lelanthran 20d ago
> But we already have examples that break that rule (us). No, we don't. Where are you reading your research papers?
32.
by lelanthran 20d ago
> Aren't the new models thinking in neuralese? Where did you read that?
33.
by lelanthran 20d ago
> If you go this route, the amount of things you just don't know is innumerable. Maybe so, but there are things we do know. For example, LLMs are only in a state that looks like consciousness when they are fed input. IOW, their &q
34.
by lelanthran 20d ago
> On a related note, I’ve seen videos of astra getting depressed when a creeper blew up its chest full of precious items. The interesting question here is, would it still get depressed if depression was not in any of its training materia
35.
by lelanthran 20d ago
>> A conscious machine that always answer the very exact same thing, formulated the exact same way, bit for bit, to a query is, well, quite a weird kind of "consciousness". > You'll need to explain why. Where are you
36.
by lelanthran 20d ago
> I don't know how well it is studied, but I suspect it is possible there is language-related complexity constraints to the effectiveness of the LLM algorithms. I believe that's obvious - humans don't think in words. Neith
37.
by lelanthran 20d ago
> Solving cancer also has an unusual level of specification. Where did you read that? "Cancer" is not just a single disease, even though we layman use the term that way. Cancer is a family of diseases, each probably having thei
38.
by lelanthran 20d ago
> The navier-stokes shows us what mathematical problem can be solved when $10m worth of compute is thrown at something. > It also makes one wonder: What could AI solve if we managed to orchestrate billions worth of agents to take on
39.
by lelanthran 20d ago
> We're not talking about learning the rules of chess here, but playing a competent game from just being shown the rules. Okay, lets go with that: it's the "shown the rules" bit that we are arguing about. The argument
40.
by lelanthran 20d ago
> And how many years of direct play and study does it take for a human to get good at chess or any other game? Time is irrelevant to training; the more relevant comparison is "how many games does a human need to play to get diminish
41.
by lelanthran 20d ago
> Why should that matter? Because we want to use this as a replacement for humans, and the average human can learn the rules of chess without needing to see the rules explained hundreds of thousands of times in millions of games. So, yea
42.
by lelanthran 21d ago
> The LLM's tests passed, which is fine and dandy, but how do I know the tests are truly valid if I do not understand what is fundamentally being tested? Do what I do sometimes: use opaque pointers (in C, or C++), and don't gi
43.
by lelanthran 21d ago
> In a few years we'll look at this the same way as people who unicycle or blacksmith. It's equally likely that in a few years we see everyone who doesn't have skill the same way we view people who follow the Kardashians
44.
Nike's End of Men(houseofstrauss.com)
11 points by lelanthran 22d ago|1 comments
45.
by lelanthran 23d ago
> Assuming the LLM never got anything wrong or otherwise had to be re-prompted, that means your devs were reviewing 130 SLOC per hour, It's possible, I think, but humans cannot do repetitive mental work (like reviewing code) constan
46.
by lelanthran 23d ago
> We are going from the era of manual, line-by-line mental model transcription to one where software engineers can focus on data structures, software architecture and algorithms. Why? What makes you think your data structure design, arch
47.
by lelanthran 23d ago
> There are two choices, go to the slaughterhouse (probably losing your job due to inability to compete internationally and getting nothing in return), or take the different option, which is less bad, and might even be good (probably los
48.
by lelanthran 24d ago
> The idea of a distributed botnet of AI using the compute of its victims to continue its inference is pure science fiction given how LLMs actually work. An attack like this doesn't really need the exploited computers to run inferen
49.
by lelanthran 24d ago
> "durrr it's just matrix multiplications" mfer so is your brain. Where did you read this?
50.
by lelanthran 24d ago
> How can they be higher quality than models they were distilled from? The word "distillation" is not specific to AIs, has been in use for 100s of years, and does not mean the same thing as "dilution". It means "
51.
by lelanthran 24d ago
TBH it doesn't need to completely solve a problem. I can use GLM to do in 2 hours what would have previously taken me a week.[1] Switch to Astra or Fable might take that down to 1 hour, maybe. The difference is not so big when you lo
52.
by lelanthran 24d ago
> I don't think this adds up. The AGI blocks will have AGI kill bots and those that don't will have nukes. Doesn't matter how much "intelligence" you have stockpiled when nukes start raining down on you.
53.
by lelanthran 24d ago
> I wonder when they will give up on the gaming market because that could take down several publishers and developers. It really shouldn't: the most money in games are those games that don't require high end GPUs.
54.
by lelanthran 24d ago
All the models converged, in every single generation since 2022.
55.
by lelanthran 24d ago
Problem is conflict of interest: the studies are mostly from the providers of the biggest models, or someone who received free tokens to do the research.
56.
by lelanthran 24d ago
> I must have been unclear. My point is 3. Okay, in that case you're still being irrational and illogical, expecting people to go to the slaughterhouse quietly. Of course they are going to protest. Whether it makes a difference be
57.
by lelanthran 24d ago
> I spent hundreds of hours on this, so there is still some friction. I'm afraid your hundreds of hours of prompting will reduce to "clone this product, here are 200 screenshots of every screen." I don't understand wh
58.
by lelanthran 24d ago
> I always see these LinkedIn or X guys making $500k ARR or $1m ARR within months; how the fkk do they get that many customers without spending on ads? First, as others have pointed out, that ARR is an amplification number. Second, if i
59.
by lelanthran 24d ago
> If you make 200 connections on linkedin every week and send them a message, some % of that will convert to meetings and sales (the exact % depends on your audience, message, and how urgent the pain you're solving is) Doesn't
60.
by lelanthran 24d ago
It doesn't matter when you post it; some kid with a $200 claude sub (or cheaper) is going to clone it for far cheaper than you, and sell it for far cheaper than you. There's no friction to what you produced, so why do you think th

More ›