Replying to
@huronbikes@cyberplace.social @Viss@mastodon.social The data is the instructions! Just like the worm and the spice.
Remote
Pier Hegeman
@pier@infosec.exchange
mastodon 4.8.0-alpha.3+glitchPrivacy and information security advocate. Formerly enterprise security architect. Currently CTO at MadeDaily - Combining branded marketing with end to end encrypted data collection tools for Lawyers.
#infosec #security #privacy #cybersecurity #developer
11 Followers
32 Following
40 Posts
Joined December 17, 2022
Website:
Replying to
@Slyence@mastodo.neoliber.al @Strandjunker@mstdn.social Balancing a budget is easy when you don’t let the ultra wealthy socialize all the risk and privatize all the profit.
Replying to
@jcrabapple@dmv.community Especially not in Anoka MN, where the Sheriff just signed an agreement with Flock for $174,999. Exactly $1 less than the amount that would have required County board approval.
https://app.govly.com/public/signals/150426
Replying to
@reverseics@infosec.exchange That first balance adjustment is pretty amazing though. It won’t be long before the murder bots are following you up all the stairs.
Replying to
@buckfiftyseven@mastodon.social From every project that had their work swiped without consent or compensation. Google didn’t magic up that script out of thin air, they stole it. From the people of the future who are going to live with constant droughts and catastrophic climate disasters. And from yourself, through deskilling.
Replying to
@WindOfChange@mastodon.online @Natasha_Jay@tech.lgbt Feels more like they want the Altered Carbon universe to me.
Replying to
@simon_brooke@mastodon.scot @david_chisnall@infosec.exchange 237 zero days just dropped in v4.01p9, I would recommend updating.
Replying to
@menelion@dragonscave.space @cstross@wandering.shop New to you does not mean new. LLMs cannot suggest new anything, they return what is statistically likely to be a response to any given input based on their training data. With a bit of randomness thrown in just for that extra spicy level of “you can never trust this for any real work.”
Replying to
@AlisonW @JustinMac84 @cstross Except that apparently Chrome doesn’t even use the local model except for a couple of features that most people don’t want or use.
Replying to
@Daojoan@mastodon.social
We are unquestionably wiser*
The fact that a relatively small number of bad actors have used the tools of information to build obscene wealth and power for themselves doesn’t negate our wisdom, but it does expose the vulnerability of our social structure which, I think, is based on an assumption that the many good people can/will stop the few bad.
So wiser, or better informed, doesn’t necessarily mean we automatically have a better society.
*In my opinion. Not claiming to be the sole arbiter of wisdom, etc.
Replying to
@dalias @Tijn @baldur Yeah I was meaning to reply further up the chain; I find threading in Ice Cube a bit goofy. I agree with everything you said. I’d even take that further though and suggest that major software platforms that impact significant portions of society should be way more heavily regulated, along the lines of utilities or banks. Relying on an organization to enforce quality when all of their incentives push towards “move fast and break things” is going to be a losing play.
Replying to
@dalias @Tijn @baldur I think this framing is a sort of victim blaming. It is not the individual developers who must love and care about the craft. It is the organizations who must mandate, validate, and ultimately incentivize quality. When all funding for architecture, design, QA and maintenance are cut and everything is left to the developer it is an organizational failure, not an individual one. I don’t need my plumber to love their craft, I just need them to follow code and pass inspection. I do agree that this was a problem long before LLMs came along.
Replying to
Replying to
@shredder7579 @martin @jk LLMs churn out crap incessantly. These “tools” are in no way better than knowing how your code works. As for the “boilerplate/migration” argument, I see this assertion all the time but never any specifics. Like, how much boilerplate is everyone writing that they need an LLM to do it for them? Not to mention that every line of code generated by LLM is stolen without attribution, and almost certainly against the license of many of the projects that were lifted. Also, you know, boiling the planet just so some tech bros can avoid learning how to do their job (or paying someone who knows).
Replying to
@mttaggart Well, since this is objective reality at the moment I think the answer is “we deal with it.” For me personally, it means I’ve basically stopped reading anything on the internet. Fedi still seems mostly real so that’s nice. And I still read blogs from before the dark times by people I know actually write. It’s going to get harder to tell signal from noise, but I think we will develop techniques to identify real voices. Hopefully this doesn’t last forever.
Replying to
@GossiTheDog@cyberplace.social Who has two thumbs and wouldn’t touch this steaming pile with a 10 meter cattle prod?
Replying to
Replying to
@jeffowski But AI taking (and then failing to do) your job is progress and inevitable and get used to it this is just the way it is now.
Replying to
@gregalotl@c.im @aj@gts.sadauskas.id.au @adavid@aus.social @spriebsch@phpc.social @preinheimer@phpc.social @pluralistic@mamot.fr Believe me, we have that already. The vast majority of all code in the world is complete shit.
@HardCoreLinuxUser @bielsubob@infosec.exchange That’s okay, we’re banning passwords. And accounts.
@samlitzinger@journa.host It’s a grift. The whole thing is a grift on both sides. Iran makes billions, Trump and his cronies make billions. We get $10 gas now and food shortages next fall.
Replying to
@akareilly@hachyderm.io “If politicians gave a shit about…” pretty much only works when the next word is “money”.
Replying to
@menos @dangoodin Is this perhaps Google trying to get ahead of the next hype cycle, given the (hopefully) impending AI collapse?
Replying to
Replying to
@martin @shredder7579 @bob_zim @jk That’s another part of the problem and proves yet again that LLMs are toys, not suitable for actual work. Real production systems, and actual software development requires deterministic outputs for a given input. The fact that the same prompt will return different results day to day, or even between queries with the same exact model is a huge problem. Even if there is “enormous maturation” (citation needed), that problem will never go away. It’s baked in to the design of the system. Also, you know, stealing everything on the internet, ruining the economy, and melting the planet.
Replying to
@buckfiftyseven@mastodon.social Right, because Gemini is trained exclusively on bash man pages and Stack Overflow 🙄
Replying to
@shredder7579 @bob_zim @martin @jk That’s my experience as well. Beyond the enormous ethical and environmental problems, it just doesn’t work well. It’s very impressive if you’re not paying attention, but to do anything useful it falls over really fast. Even the single line autocomplete is usually more annoying than useful. It just keeps wanting to generate more and more and gets less and less relevant to the context.
RE: @MostlyHarmless@thecanadian.social
Oh, I get it I get it! It’s whatever we think of! If we think of J. Edgar Boozer, J. Edgar Boozer appears and destroys us.
Replying to
@KimCrayton1@dair-community.social Do you see that number continuing to go up? I hope it does because I feel like 71% of companies NOT re-hiring workers laid off for “AI” is… not great.
Replying to
@cstross@wandering.shop I think this is a major reason why LLMs are being pushed so heavily and against all financial logic by these massive corporations and billionaires. Though it always has to some extent, in the age of “AI” money literally equals truth. If fact is just the statistically most likely combination of words in a training set, then those with the most resources, who can buy the most domains and publish their propaganda the fastest/most completely, will always be “right.” And since we have ceded all control of our attention to the same billionaires, they can simultaneously strangle all real content creators by denying them any traffic.
Replying to
Replying to
@mochabeau @molly0xfff A lot of Minnesota is actually pretty red. It’s mostly the Twin Cities that keep the state blue, and that could change pretty quickly. In 2024 I saw a lot more Trump signs in St. Paul than I ever thought was possible.
RE: https://dmv.community/@jcrabapple/116986476987475930
This is illegal, definitely do not follow these steps
Replying to
Replying to
@codinghorror@infosec.exchange Yes, a new kind of sentience… in the same way that Monsanto is bringing a new kind of sugar into existence. As in, it’s not. It might trick some people who aren’t paying attention, but you can always tell. And consuming more than a tiny bit will make you sick.