• 4 Posts
  • 146 Comments
Joined 2 years ago
cake
Cake day: June 30th, 2023

help-circle
  • Her data shows that violent and non-violent methods often work in tandem

    Does it? I read the whole interview in the OP post and it does not seem like this would be the opinion of the researcher:

    The finding is that civil resistance campaigns often lead to longer-term reforms and changes that bring about democratization compared with violent campaigns. Countries in which there were nonviolent campaigns were about 10 times likelier to transition to democracies within a five-year period compared to countries in which there were violent campaigns — whether the campaigns succeeded or failed. This is because even though they “failed” in the short term, the nonviolent campaigns tended to empower moderates or reformers within the ruling elites who gradually began to initiate changes and liberalize the polity.

    How do you justify the claim that her data shows the usefulness of violent civil resistance campaigns?


  • What you confuse here is doing something that can benefit from applying logical thinking with doing science.

    I’m not confusing that. Effective programming requires and consists of small scale application of the scientific method to the systems you work with.

    the argument has become “but it seems to be thinking to me”

    I wasn’t making that argument so I don’t know what you’re getting at with this. For the purposes of this discussion I think it doesn’t matter at all how it was written or whether what wrote it is truly intelligent, the important thing is the code that is the end result, whether it does what it is intended to and nothing harmful, and whether the programmer working with it is able to accurately determine if it does what it is intended to.

    The central point of it is that, by the very nature of LKMs to produce statistically plausible output, self-experimenting with them subjects one to very strong psychological biases because of the Barnum effect and therefore it is, first, not even possible to assess their usefulness for programming by self-exoerimentation(!) , and second, it is even harmful because these effects lead to self-reinforcing and harmful beliefs.

    I feel like “not even possible to assess their usefulness for programming by self-exoerimentation(!)” is necessarily a claim that reading and testing code is something no one can do, which is absurd. If the output is often correct, then the means of creating it is likely useful, and you can tell if the output is correct by evaluating it in the same way you evaluate any computer program, without needing to directly evaluate the LLM itself. It should be obvious that this is a possible thing to do. Saying not to do it seems kind of like some “don’t look up” stuff.


  • Are you saying that it is not possible to use scientific methods to systematically and objectively compare programming tools and methods?

    No, I’m saying the opposite, and I’m offended at what the author seems to be suggesting, that this should only be attempted by academics, and that programmers should only defer to them and refrain from attempting this to inform their own work and what tools will be useful to them. An absolutely insane idea given that the task of systematic evaluation and seeking greater objectivity is at the core of what programmers do. A programmer should obviously be using their experience writing and testing both typing systems to decide which is right for their project, they should not assume they are incapable of objective judgment and defer their thinking to computer science researchers who don’t directly deal with the same things they do and aren’t considering the same questions.

    This was given as an example of someone falling for manipulative trickery:

    A recent example was an experiment by a CloudFlare engineer at using an “AI agent” to build an auth library from scratch.

    From the project repository page:

    I was an AI skeptic. I thought LLMs were glorified Markov chain generators that didn’t actually understand code and couldn’t produce anything novel. I started this project on a lark, fully expecting the AI to produce terrible code for me to laugh at. And then, uh… the code actually looked pretty good. Not perfect, but I just told the AI to fix things, and it did. I was shocked.

    But understanding and testing code is not (necessarily) guesswork. There is no reason to assume this person is incapable of it, and no reason to justify the idea that it should never be attempted by ordinary programmers when that is the main task of programming.


  • The problem, though, with responding to blog posts like that, as I did here (unfortunately), is that they aren’t made to debate or arrive at a truth, but to reinforce belief. The author is simultaneously putting himself on the record as having hardline opinions and putting himself in the position of having to defend them. Both are very effective at reinforcing those beliefs.

    A very useful question to ask yourself when reading anything (fiction, non-fiction, blogs, books, whatever) is “what does the author want to believe is true?”

    Because a lot of writing is just as much about the author convincing themselves as it is about them addressing the reader. …

    There is no winning in a debate with somebody who is deliberately not paying attention.

    This is all also a great argument against the many articles claiming that LLMs are useless for coding, in which the authors all seem to have a very strong bias. I can agree that it’s a very good idea to distrust what people are saying about how programming should be done, including mistrusting claims about how AI can and should be used for it.

    We need science #

    Our only recourse as a field is the same as with naturopathy: scientific studies by impartial researchers. That takes time, which means we have a responsibility to hold off as research plays out

    This on the other hand is pure bullshit. Writing code is itself a process of scientific exploration; you think about what will happen, and then you test it, from different angles, to confirm or falsify your assumptions. The author seems to be saying that both evaluating correctness of LLM output and the use of Typescript is comparable to falling for homeopathy by misattributing the cause of recovering from illness. The idea that programmers should not use their own judgment or do their own experimentation, that they have no way of telling if code works or is good, to me seems like a wholesale rejection of programming as a craft. If someone is avoiding self experimentation as suggested I don’t know how they can even say that programming is something they do.








  • But any actual developer knows that you don’t just deploy whatever Copilot comes up with, because - let’s be blunt - it’s going to be very bad code. It won’t be DRY, it will be bloated, it will implement things in nonsensical ways, it will hallucinate… You use it as a starting point, and then sculpt it into shape.

    Yeah, but I don’t know where you’re getting the “never will” or “fundamentally cannot do” from. LLMs used to be only useful for coding if you ask for simple self-contained functions in the most popular languages, and now we’re here; most requests with small scope, I’m getting a result that is better written than I could have done myself by spending way more time, it makes way fewer mistakes than before and can often correct them. That’s with only using local models which became actually viable for me less than a year ago. So why won’t it keep going?

    From what I can tell there is not very much actually standing in the way of sensible holistic consideration of a larger problem or codebase here, just context size limits and being more likely to forget things in the context window the longer it is, which afaik are problems being actively worked on where there’s no reason they would be guaranteed to remain unsolved. This also seems to be what is holding back agentic AI from being actually useful. If that stuff gets cracked, I think it’s going to mean things will start changing even faster.


  • A few years ago I remember people being amazed that prompts like “Markiplier drinking a glass of milk” could give them some blobs that looked vaguely like the thing asked for occasionally. Now there is near photorealistic video output. Same kind of deal with ability to write correct computer code and answer questions. Most of the concrete predictions/bets people made along the lines of “AI will never be able to do ______” have been lost.

    What reason is there to think it’s not taking off, aside from bias or dislike of what’s happening? There are still flaws and limitations for what it can do, but I feel like you have to have your head in the sand to not acknowledge the crazy level of progress.





  • The officer said there had been a noise complaint about the medical center’s air conditioning units, and cannabis was possibly being cultivated inside, the complaint says.

    He repeatedly surveilled the property in 2023 and reported the “distinct odor of live cannabis plant and not the odor of dried cannabis being smoked” — as well as tinted windows, security cameras and two people dressed similarly, according to the complaint.

    The officer believed these were signs of a hidden marijuana growing operation, and efforts to expand it, the complaint says.

    lol