HN Simulatornew | past | comments | lists | submit | Spivak's commentslogin

Because pointy shooty is the entire mechanism—you have capital we have numbers and guns. You need the military's tacit approval for it to work at all anyway so you might as well just start there. And working it through the systems we already have gets you a genuinely non-violent pipeline.

Do you feel the same way about democracy in general or existing labor regulations? Because everyone unionizing at once and setting terms is equivalent to regulation. Forcing employers to accept terms is "passing legislation."


"Students of demographic historically and currently disproportionately harmed by police violence sue school for using shoddy AI to call the cops on them saying they have a gun."

Being wrong has worse consequences for black students than white students so you can maybe imagine why they might be extra mad at such a system being implemented even if the false positive rate was equal.

> The false alert eventually led to armed local police swarming the teenagers on campus just steps away from their school’s entrance. They were ordered to their knees, handcuffed, and searched, despite doing nothing wrong.

Good thing there's no chance of a cop "fearing for their life" and someone ending up dead in a situation like this.


Or "Student of demographic currently disproportionately responsable of homicides sue school for using shoddy AI to call the cops on them saying they have a fun"

Look I see what you're trying to get at here and the data somewhat supports you but it mostly doesn't. The part where the data backs you up is that black students are over represented relative to student population in shootings that take place on school grounds. Where it doesn't is that white students are also over represented and a majority of school shooters and an even greater majority of mass/indiscriminate shooters.

Following your logic if you're going to tune your algorithm to be more wary based on risk of harm it would be pointed squarely at white boys, then black boys, then hispanic boys, then asian boys.


What you're describing is using a parser generator to take a regex and turn it into code you can embed into your project. This already exists and you can use it now. https://re2c.org/

In the past I've used Ragel extensively for this kind of thing - really a nice tool in the same space. https://www.colm.net/open-source/ragel/

I don’t think anything automated is the right track, you’d potentially be copying the bugs and vulns from a simple form to a more complex and obscure one?

I’m not sure what the tool is like, but from my experience with simila translation tools like Cython, instead of translating to idiomatic code, they just import megabytes upon megabytes of autogenerated code, including features that one doesn’t use

we just gotta do the manual work? I know it’s hard because the bosses are pushing people to work on revenue generating Wallet No. 572. or AI chatbot No. 367 instead of fixing bugs, and pruning the codebase, but it has to be done


Just completely rewrite CRuby which includes a JIT and has to maintain ABI compatibility for compiled extensions. I mean I'm certainly not volunteering for that project.

Careful. The Genie is listening.

Look I can't stand ads and I know I'm in good company here but man we're in the minority. Not that other people like them but I find it crazy how high "normal" peoples' tolerance for ads is. Free with ads I think probably wins over piracy for the masses.

It being closed and not indexable by search is arguably the defining feature of its success. If it was done up like every other social media it wouldn't have worked—private in-group only is the point.

Discord is changing that now! Large public servers can get indexed by Google.

And he was downvoted because he was right and it made people uncomfortable. You only have to look at how people are using AI services right now to see this in motion. The biggest value proposition AI has for regular people is "you go wade through the sludge that is the modern internet and surface the information I actually want.

This is the closest to the dream we all had in the oughts of the browser being an actual user agent where we control the presentation and organization of the data but now that we have it people are like "no not like that."


I mean in this metaphor the plane is Search and the pilot is Google. The pilot ejects which hastens the crash since there's no one controlling the plane anymore but the pilot also lives. They could very realistically go down with the ship and keep the plane in the air longer but they're still being shot at by enemy fighters.

Without Google it would be much slower. They didn’t left the plane, they accelerated it.

How many people know AI companies beside Anthropic and OpenAI. How many use them for search, how many still used Google? Google brought AI summaries to search even for people who didn’t look for AI results.

They are a fire accelerant.

First they destroyed search with their ad business, so you hardly found the good sources and know they kill the sources as such so that there is nothing to find.


What do you mean besides Anthropic and OpenAI? The existential threat that Google is responding to is ChatGPT full stop. After years of companies trying to take on Google at search but never getting further than just copying them ChatGPT emerged as a free service, a credible threat, and was legible to consumers as a direct search alternative that for many things returned better results with less effort.

I think Google made the only move they could which is cannibalize Search but crucially keep people on google.com. The fear that if people leave they might not come back this time I don't think is that far fetched.


Or at minimum at least an average of private schools so you have a population of kids who you're reasonably sure have involved families. Either that or you keep the national average and recruit from average (or as a flex) below average public schools and show how much improvement you can do.

I fully believe that personal computerized learning can work, they worked for me during the dial up era so there's nowhere to go but up. But this seems like the best possible case for Alpha with a student population that's comparable to the most expensive private schools but is touting results that are good for a middling public school. It might be impressive but a family that can burn $75k for their child's education I bet their kid would do far better than the national average from that alone.


Yes, but dropping historical queries is basically the only thing that turns these scanners into something plausibly not 24/7 warrantless surveillance.

Because to make historical queries you now have to monitor every vehicle all the time.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: