gitaarik

17 ms·

331.
by gitaarik 1y ago
You're absolutely right! I also get this too often, when I sometimes say something like "would it be maybe better to do it like this?" and then it replies that I'm absolutely right, and starts writing new code. While I w
332.
by gitaarik 1y ago
That's why I wrote Lit State, a simple state management lib to use in combination with Lit, a simple web components lib. It works really simple, and it's simple to use. Much more intuitive than React. https://github.com
333.
by gitaarik 1y ago
I don't think it's a complete good comparison. In the past painting was the only way to depict real world events, but painting is also art, and it often doesn't necessarily depict reality, but the artist's interpretation
334.
by gitaarik 1y ago
As a independent dev, I haven't had the need to pay for any AI yet. When I run into my limit at one company, I switch to the next one. Not always the same experience, but the next day I can start fresh again.
335.
by gitaarik 1y ago
It's probably because you never did the Django tutorial. Once you understand what what does, it's actually very nice, because you don't have any boilerplate. Django takes care of the boring parts, and you can focus on making
336.
by gitaarik 1y ago
Yeah, I also see tons of comments in there loading up, and then at some point the page "crashes" and you get the "unable to load" page
337.
by gitaarik 1y ago
This Indian gov YouTube channel has all kinds of videos for that: https://www.youtube.com/watch?v=tXWPf0HQd5U
338.
by gitaarik 1y ago
It's used for coloring, because it makes it apparently gives it the same color as good real tumeric
339.
by gitaarik 1y ago
Yes indeed. Here's also an instruction video for it: https://www.youtube.com/watch?v=tXWPf0HQd5U
340.
by gitaarik 1y ago
Actually the recent fails with Grok remind me of the early fails with Gemini, where it would put colored people in all images it generated, even in positions they historically never were in, like German second world war soldiers. So in that
341.
by gitaarik 1y ago
Seems logical to me. 1 = break the current level, 2 = break 1 level up, 3 = break 2 levels up, etc. If you would do it the other way around, and then if you would add another for loop around the others, the breaks will break. You wouldn&#x2
342.
by gitaarik 1y ago
Or you don't longer realise you're seeing and being influenced by them. You just become part of them.
343.
by gitaarik 1y ago
19th century France is AI generated
344.
by gitaarik 1y ago
I use a Bluetooth audio receiver in my car. It has a Jack output that you can plug into your car's amp. You connect your phone to it through Bluetooth, and then you can play music in your car. It's very convenient, because your ph
345.
by gitaarik 1y ago
I use a Oneplus 9 pro with Iodé OS (based on LineageOS). It comes with Aurora store and MicroG pre-installed. It works very good and I'm very happy with. It is regularly synced with security updates from LineageOS.
346.
by gitaarik 1y ago
Then use Session: https://getsession.org/
347.
by gitaarik 1y ago
No they have apparent empathy because of the reward models humans trained them with. To be Mark you have to experience real existential fear, and a need to control other people to compensate the fear. And LLMs can't do that indeed. But
348.
by gitaarik 1y ago
Aren't halluciantions enough proof for you that they don't think/understand? At least not in the same way as humans? If a student was on a regular basis hallucinating and giving complete nonsense as an answer, I don't th
349.
by gitaarik 1y ago
So how would you explain how LLMs work to a layman?
350.
by gitaarik 1y ago
So you mean because we don't know how the human mind works, LLMs won't be far off?
351.
by gitaarik 1y ago
Well, I think they're implementing it because they are the mad men!
352.
by gitaarik 1y ago
The thing what I wonder here is how do they make the benchmark testing environment? If that needs to be curated by humans, then the self-improving AI can only improve as far as the human curated test environment can take them.
353.
by gitaarik 1y ago
Every new major technology always endangeres the status quo. https://chatgpt.com/share/683f3932-fce0-8012-a108-4b70c3e5fd... Things change and it's scary, but it usually works out. Or at least we just get used to
354.
by gitaarik 1y ago
Well chess has a very limited set of rules and playing field. And the way to win in chess is to be able to think forward, how all the moves could play out, and pick the best one. This is relatively easy to create an algorithm for that surpa
355.
by gitaarik 1y ago
You come up with very interesting points, and I'm thankful for that. But I also think you're missing the crux or my message. LLMs don't experience the world the same way humans do. And they also don't think in the same w
356.
by gitaarik 1y ago
Ok, this is interesting indeed and I'll investigate more into it. But I think my points still stand. Let me elaborate. An LLM only learns through input text. It doesn't have a first-person 3D experience of the world. So it can&#x2
357.
by gitaarik 1y ago
Ok, so you claim that LLMs can get smarter without human validation. So why do they hallucinate at all? And why are all reward models currently curated by humans? Or are you claiming they aren't?
358.
by gitaarik 1y ago
Sorry but I don't follow your logic. Are you claiming that reward models that aren't curated by humans perform as well as ones that are? Then what is a reward model's function according to you?
359.
by gitaarik 1y ago
But reward models are always curated by humans. If you generate a reward model with an LLM, it will contain hallucinations that need to be corrected by humans. But that is what a reward model is for. To correct the hallucinations of LLMs. S
360.
by gitaarik 1y ago
So to clarify, it could potentially come up with (something close to) C, but if you want it to get to D, E, F etc, it will become less and less accurate for each consequentive step, because it lacks the human curated reward models up to tha

More ›