You told it to do a thing and it did the thing. You didn't tell it to not send your email address.

That is text book alignment problem. It does what you tell it to do, not what you implicitly meant.

Memory + abstraction capability is a more important factor in human intelligence than thinking speed in my opinion.

When facing a novel situation or problem if you can recall a structurally similar one you can better predict how to solve it or how to find new interesting questions to ask.

Sorry but we have extremely different definitions of "paranoid". You are willingly sending your proprietary code to a third party.

Even if they respect their promise of no training, they still have your code, the IP is already leaked to them. And in case of a breach of their systems it would leak globally.

At my company we are not allowed to let LLMs touch the core code directly. And I don't think it's particularly "paranoid", we are just following the same general behavior from before AI.

There is such a wide range of response to this new paradigm, it's very interesting.

Where are you seeing this? When? I go to "Home" and I only see my subscriptions.

I have a cosmetic filter to avoid clicking on the "Popular" link but I never see anything from the front page that's not part of my subreddits. I don't use the app.

For me programming was never only about creating the thing but also about becoming better at design and problem solving. So I'm still doing things mostly by hand to get these benefits. I'll use the LLM as a brainstorm partner.

It's like when I'm trying to learn a foreign language. Asking the AI for a translation of a paragraph is the easy shortcut but you know it's not going to help in the long run. It's better to manually look in the dictionary or look for real-world usage, or even ask the AI to explain the grammar, the effort itself is what helps crystallize the knowledge.

This approach won't be intuitive for the next generation. But LLMs can definitely be used to teach you how to do things instead of doing the things for you. We need to educate people about how to use them a way that make them better rather than simply delegate thinking.

The "I haven't written or read code in a year" experience is not universal.

And I also have roughly the same approach for my own projects. I use LLMs to brainstorm design, algorithms, data structures, maybe have it spit out small building blocks in isolation, but I'm still very much in control and owning every line.

This is not a critique of LLMs, it's just the new state of affairs that since people can more easily solve their own problems it has somehow become more difficult to be helpful to others.

For many hobby open source developers the reward and motivation was simply random people saying thank you.

There is no room for false positive here in the same way you can't randomly find a collision in a hash function if it's strong enough. Like the rate is so infinitesimal that it is effectively zero.

Now replace random number sequence with prompted string of words. And instead of using the PRNG on every word I use it every n words. If the generated text is sufficiently long I can tell by matching the expected deterministic pattern.

You can defeat it by changing the words yourself and triggering a false negative but there isn't really any room for a false positive if the text is long enough and the pattern matches perfectly. If the pattern doesn't match then I can compute a probability.

You have to work in design mode and stop it from pissing code. Describe the requirements as detailed as possible and refine them based on the responses. It's quite rewarding and helps structure the way you think about the problem.

You could also look at other repos in github to see what they do to do the same task.

If you make a video player based on ffmpeg you take input videos. The attacker can share a specially crafted video with the victim. Same kind of attack as sharing word documents.

It is currently marking my little town as under fire / burned down, but it's not and hasn't been so far. (north west front of the fire, next to the lake). It looks like their contouring algorithm is very greedy and any hot spot marked in the past few days has become part of their fire perimeter.

The massive fire clouds probably contribute to the difficulty of measuring ground temperature from satellite.

But also, for this conjecture to be wrong is quite surprising to me. Intuitively I would think any convex polygon to be topologically equivalent to a circle, and any convex n-gon should be deformable into its regular version, then back to the other one…

No it also encourages the local market and healthy competition. This way in the future we don't fall into the same enshittification trap.