> My experience of using AI as a search engine has surprised me. I never expected an overgrown pile of matrices to work that well.
The first version of Google was also surprising. Mind blowing use of linear algebra (also a pile of matrices, but this time sparse matrices mostly) to rank websites
So maybe the search business was always meant to use pile of matrices
> I'm not sure I'd call it an alignment issue, because, in all cases I've seen where it does this (usually what I've seen is writing a python script to get around the harness permissions blocking something), it's trying to do the thing I just told it directly to do, and it's overcoming obstacles to accomplishing that.
The paperclip factory problem is definitively a misalignment issue. That's because we expect agents to be aligned not only to your immediate prompt, but to shared, implicit values
It's weird because I am sure that a large number of established apps in flathub, such as.. Chromium https://flathub.org/en/apps/org.chromium.Chromium has AI-assisted code, and this bans all AI-assisted code
The policy as it stands is contradictory with flathub practice, unless it specifies that already accepted apps are grandfathered or something like that.
After some time it's looking like AI is becoming just another tool in the developer toolbox. So we can confidently say that, over time, only projects with specific anti-AI policies will be free of AI code (and actually, even such projects I'm not sure)
So it's basically a ban on most new projects, made by inexperienced developers. Which is okay I guess, if this means they can better review projects.
> I’m surprised to see such a reaction to a left-aligned content column. I’d have said it was pretty common, though probably not as common as it used to be.
When it was common, columns weren't so thin I think
Left aligned content was much less of a problem on the old displays with a 4:3 aspect.
Nowadays, with wide displays, the browser windows may also be wide, especially when using a lot of open tabs, and left aligned content becomes much more annoying when it is thrown far away from your center of vision.
well, ripgrep is not a company.. what means to "buy" an open source project?
companies either "fund" open source developers (usually a pittance), "contribute" code, or if all fails "fork" them, but straight buying, that's something I never see
In slave states without proper civil rights, like the US, you can buy them. Your contract forbids everything. Eg every SW you produce in private belongs to the company, not to you. Or everything you say in private will affect your contract.
Well, the background checks in the US at least have to follow laws.
Then theres reference checks, less formal but generally still follow some kind of rigor or process and you generally know who they’re going to talk to.
Then there’s the third type of investigation, where they actually don’t tell you they’re doing it, may or may not admit to doing it, and you have no way of knowing whether you were singled out for some reason, assuming you are able to find out it happened in the first place.
And good luck searching for any recourse if ethics or even laws (such as HIPAA) are ignored in that situation, even if it’s for example a b2c healthcare company that is expected to set an example for health privacy standards, rather than calling everyone you knew to probe their memories of how healthy you appeared to be at the time, then using that information to evaluate your performance with different criteria than other employees.
Sure, it’s not a typical scenario, but if it happens to you, you’ll probably never be able to scoff at a comment like this again. You might never get over it.
> The function of sleep in animals is largely obscure.
This just obscures the conversation IMO. We know a great deal of functions that sleeping performs. We don't know everything. For some reason this prevents us for using this word in a computing context? What do you think about sleep(1) in Unix?
Also..
> Does their LLM "die" if it can't perform the function described?
what "death" means in the context of a computer program?
they already have enough data to make this correlation almost 100% accurate
they know things like your git author line, your github handle, and the exact codebase you were working on
the general shape of commits
even if you change extensively, they will probably be able to match this with claude code sessions
sure the atribution at the end of commits is a signal, but I doubt it's much valuable
if anything it's more valuable to anthropic competitors, that don't have claude code session data to match to open source contributors, and will have to guess if any given code is AI generated, and by how much
reply