elle

@emilymbender@dair-community.social great thread, I think at different points I've run into each of the points you laid out (especially when I was trying to be more amenable to LLM users). since, I've grown to taking a much more hard-line anti-big-tech-LLM, neutral-to-meh academic/open-model LLM, and recommending searching out all alternatives before turning to the tech as a solution. thank you for your work in this area, as you and @alex@dair-community.social have helped shaped my opinions/knowledge on the subject.

@nullagent@partyon.xyz @dataparty@partyon.xyz @rfparty@partyon.xyz the data rate for LoRa is just so much lower (even at 2.4GHz) than HaLow (~30 KB/s vs. ~40 MB/s). LoRa still shines wrt range, though.

@valkyrie@wandering.shop very cool experiment, nice work!

@nullagent@partyon.xyz @dataparty@partyon.xyz @rfparty@partyon.xyz this looks awesome! congrats on the release, and all the hard work!

@yacodes@mastodon.social @baldur@toot.cafe fans of the pre-slop project: please fork the original Zig codebase. mb call it something like "hare" or "tortoise"

@baldur@toot.cafe this, friends and foes, is what the highest level of corpo bootlicking and selling out looks like. enjoy your slop, JS fans

@lina@vt.social ah, then yeah I more-or-less agree with you. if people are worried about their browser not handling password storage properly, use a proper password manager, adjust according to security/privacy concerns. that being said, secure storage (of a certain kind) exists in the browser, and M$lop could be *much* better at practicing defensive programming.

@bagder@mastodon.social maybe with time, security bounty programs will shift to "join the project as a security reviewer + patcher", with bounties paid out to merged fixes.

@jonny lel this could be the funniest outcome from all of this. if at any point open model training + dev matches these closed models, and the tech improves development of new models similarly trained on consensual open data (+ some nefarious training on closed data)...

@briankrebs it's almost like those AI tools were purpose-built for data exfiltration...

@larsmb @jzb yes, if society were structured entirely differently, total automation of labor should be the goal. With survival largely tied to wages + access to capital, this LLM trend is an attempted death march for working people. If it was inverted as @jzb mentioned, the previous trend of improving the performance of small models would get more attention. Small, edge/local, efficient models. It still can in the future. I'm not saying anything new, been said a thousand times.

@nlnetlabs wondering how much a policy requiring all vuln reports to come with a reproducer (in form of a unit/integration test), along with a proposed fix (validated against the reproducer). I think this would significantly raise the bar, and lead to better outcomes for all reporters, LLM or human.

@zkat first off, fuck that person, they should respect your boundaries. I'm also dealing with immense frustration that there isn't more backlash in the FOSS community against this LLM bullshit. hoping to see a split similar to FOSS/OSS split with pro-humans having an LLM-free space to share code, and bot-lovers can suck corpo shit on GitHub

@hailey given how absolute shit their leaked source was for one of their flagship products, I'm not convinced they didn't just augment classical fuzzers with their tools doing some basic symbolic execution. angr has used this approach since the mid-2010s.

@nixCraft lmfao

@bagder it will be interesting to see how they define "AI", and if that includes "classical" machine learning techniques used for scientific data wrangling. I would think non-generative techniques would be much better for analysis.

@cwebber its horrifically ironic to see web3 DAO bullshit playing out in the larger economy now. almost like an arcade demo PoC playing out into its real-world MVP. hate this timeline

@pluralistic @dhd6 @tante @simonzerafa what a shit take dude. rockets being perfected by nazis, project paperclip, and now a neonazi in charge of one of the largest space tech programs on the planet, along with a bullshit generating LLM. so yeah, maybe this is all fash tech, and maybe taking a stand of "I'm not touching that shit with a thousand-meter pole" is not "neoliberal purity culture". and ollama of all things? the shit pumped out by fucking Meta? are you shitting me?

@pluralistic @dhd6 @tante @simonzerafa you wrote a book on how much of a shitbag company corpos like Meta are. now you're saying "oh it's not that bad, look it's marginally better than Google Docs spell checker"?! did someone hack your fucking account? there are legitimately open models that originate from academic institutions, train on open data with full consent. even those models take tens-of-thousands of euros to train. well outside the resources available to most open-source enjoyers

@cwebber really disgusting, I hope the project goes up in a pile of flames.

@david_chisnall@infosec.exchange waow, it's almost like those machines were built for stealing people's work, and bulshitting. who would have guessed...

@zzt@mas.to I hate this so much, thank you for writing such a clear analysis of the situation. I had a feeling things were going wrong with slop being accepted into systemd, and worse when "left" media personalities started citing the fascist Lunduke as a source. @taylorlorenz@mastodon.social this is why you shouldn't platform that piece of shit.

@parismarx@mastodon.online do it

@QasimRashid@mastodon.social their corporate paymasters have spoken

@anildash@me.dm you get so close to the real solution for workers still trapped in corporate hell: join a union. that is what facilitates collective power in a meaningful way in a corporate environment. also, get enough money saved up with your newfound comrades, and start a worker collective. fuck these corporate fascist assholes bent on killing and enslaving the rest of us.

@anildash@me.dm you should really consider taking Mozilla out of your list of orgs to donate / contribute to. they are all-in on the AI-slopification of the internet, and headed by former Meta execs.

@almet@tutut.delire.party but still requires a phone number...

@lina@vt.social I encrypt all passwords in keyper with per-session keys that decrypt the password on copy/reveal, and re-encrypt after. this minimizes the amount of time passwords are exposed clear-text in RAM. it is not that hard, you just have to care. please stop apologizing for the centibillion dollar megacorp.