blharr

Eh, I mean that proof is all around in its training set. It's a fundamental, basic theorem in probability. You can put the same thing into a search engine and get a better solution, for [example](https://jeremy9959.net/Math-5800-Spring-2020/notebooks/convo...)

Nobody's saying that these aren't fascinating, just that it's not looking like their models are getting significantly better and better as all the hype wants you to believe.

Transformers + huge data set is incredible. But literally we've scraped all the data on the web and made huge sacrifices to our entire society already

Moore's Law? The computer does about 30,000 times more than it did 30 years ago. Order of magnitude shifts are just common

It's impressive. I want to be a doubter and say I feel like it just shows what you can do with tailwind/typescript in getting a nice UI out, but it really would be genuinely useful for some cases.

The problem I have with it is that how do you get out of a corner with it? Once I start having problems with the application -- I asked it to generate a search engine for different websites, but one of the API endpoints wasn't working. It kept trying over and over again but failed to get it working.

It's - like other attempts I've had with AI - more frustrating to work with. It'll say "I fixed this" and then have a problem where it created more problems with what it's fixing. I thought it finally worked 100% but it just made it look better by breaking something else without actually fixing the issue.

Admittedly, it took what might have been a day's work into a couple hours, but now I have a chunk of code that I don't understand and will be deliberately harder to understand than if I wrote it myself.

It still feels like I'm trying to coax an intern into working a project, rather than having an application that actually does the work for me.

For code review maybe, it's pretty useful.

Even with the $20 version I've lost days of work because it's told me ideas/given me solutions that are flat out wrong or misleading but sound reasonable, so I don't know if they're really that effective though.

I mean, LLMs can be regenerated as needed. With every major release, you might regenerate the documentation. Then, you don't need to have everyone taking time to ask the same questions, reducing the amount of times you need to call the LLM. And thereby you can use a more expensive LLM, like o1 to generate the documentation where you wouldn't want to use it normally

Building anything on windows without cmake is just... I don't know why anyone would use anything else. I used to spend hours wrestling with build failures, but after setting up cmake, it just works with everything.

>What does that even mean? "id-dominated state" refers to the psychological "id" basically being primal. "A person who is dominated by their id might be narcissistic and impulsive" I don't blame you for not getting this part, but it is a common psychological theory.

>What can be unburdened by what has been

Well, yea. If you take a fragment of a sentence out, it is going to be confusing.

The full quote is on the lines of "My mother raised me to see what could be, unburdened by what has been" i.e. to look at accomplishing at the highest possibilities without the baggage of traditional limitations

Reading it now and in context, its a normal impactful thought. But the right took a cut off the middle... like if I quoted you and said "a native English speaker and half the time" a missing phrase will obviously make it confusing

Pretty sure you followed the thread wrong. They only mentioned the square root in response to you?

There have even been times where an LLM will spit out _the exact same code_ and you have to give it the answer or a hint how to do it better

Hmm, this can be done with macros in C/C++ right?

It could also just be defined in the "setter" for x, that x Y = x + 2, so everything you update x you update Y

I don't particularly like it because, what if you want to change y? Do you have to change x correspondingy?

Confused by this approach, people want to write in interpreted languages and have it compiled with an LLM?

How would you do things like dynamic code execution or reflection? Lots of properties are stripped as part of the compilation that you wouldn't be able to refer back to.

Are you just saying write python -> interpret it -> compile it -> convert to assembly? Because I believe that already exists, but is difficult to just do that all the time because of the compile step and having to convert to static typing

Huh, now that you mention it, after rock climbing, I similarly got the least callused on my index finger. Even the base of my index near the palm is much less worn than the rest of the non-thumb fingers

I'm surprised you mention us not having recordings of the older times being why we thought people spoke eloquently. Even in old recorded audio and video, like old TV, it feels to me like people speak much more eloquently.

You wouldn't have to recompile them every time. What if you didn't necessarily use macros but auto-generated it in a file that you keep separate from the other code at the bottom?

Well, clearly the apple already has a bite taken out of it, so that's user error

I mean. Likely, yes, but if you have to spend the time to prompt correctly, I'd rather just spend that time learning the material I actually want to learn

To be fair 1. He's not saying that this is true. Only that the view is popular among the researchers.

2. This view of "were just this close and were only getting closer" is exactly the kind of dogma that you have to accept when you become a researcher.

Bathroom 3 to 7 times a day? I go like once a day

I mean, it is a classic example. If you have access to the hardware and the dedication to do so, you could break almost any security. That's a hilarious example to physically drill into a chip, though

Wouldn't be too hard to add some hamming encoding

Unfortunately, It'll just be great for insurance to deny necessary prescriptions/procedures/scans because AI review found nothing wrong

Kagi is admittedly pretty great for this.

I looked it up, and there are 6.6e21 9x9 sudokus, which I wouldn't consider constant time practically

Sure you can probably translate rough code and get something that "works" but all the thousands of small optimizations that are baked in are not trivial to just translate.

Because when you have decent savings. It's easy to never pay a credit card late fee in your life.

When you are living paycheck to paycheck, it's a lot harder to avoid going over. Then, even a small amount over can incur late fees, and if you can't pay it off for even a month, it results in a decent amount of interest.

Gpt-4o, in my unprofessional opinion, is miles worse than the old gpt-4. I believe it has higher scores on benchmarks, but only because it spits out paragraphs and paragraphs of useless information.

I've done the same, and it's decent, but in ML having the minor bugs that chatGPT introduces was awful for me to debug. Like, it doesn't fail catastrophically, but it just gave me really bad models. I basically got poor performance at the end for all of my models and didn't learn enough from it to debug the issues. At the end of the day, pytorch was way simpler to just learn. Have you gotten good results in some models using chat gpt with pytorch?

I get the idea, and I've considered using AI similarly, but one thing I realized: Being stuck is often the time where you learn the most. (And when you find out how to un-stick yourself is part of the most enjoyable parts of engineering).

Having the immediate aid of an LLM in a lot of ways also does hurt your reasoning capabilities in my experience. I'm curious how you would address this.

Curious where you found the 1/17 statistic?

LLMs kind of are, but imo it never meets up for a