Those systems based on LLMs really create crazy security issues as soon as they're allowed to interact with other systems.
There's clearly an issue with the security and privacy practice of those companies...
Not what I signed up for years ago, but it's the political and social landscape we inherited... Surveillance is being on steroids now, so definitely need to fight it at every turn.
The web scraper situation isn't getting better... How long can the open web still hold?
This keeps being a very interesting TTS model. Looks like it's getting simpler to deploy too.
Really this kind of AI push is a bad move for employers, especially when interacting with FOSS communities so much. This forces people to pass the ethical issues onto volunteers...
The data set is rather small but the trend is really bad. So much reviewer time wasted due to AI slop... this time on the Flathub side.
How is going this social experiment at scale? Not well I'd say... And some in those cohorts will end up in positions of power, that's when it'll become really "interesting" I guess.
You sure you can trust those systems? It's proprietary software and they're clearly on a slippery slop. For something so security sensitive this is concerning.
Interesting paper (go to the full one for all the details) which shows that with the current architecture it's really hard if not impossible to make safe systems with LLMs. This gives interesting insights in the weird form of proto-cognition those models exhibit.
More reasons why the whole "data centers in space idea" is stupidly dangerous and likely unreachable.
Remind me how your built your models in the first place? Yeah right...
Very sobering opinion piece. For all the talks about a China / USA race, it feels more like two flavors of the same dystopia. The race is just here to justify acting against their own population interest. The result is then the increase in illiberal fixations and nihilistic world views. This can't end well.
Some news from the Apertus project, they released smaller models. Interesting work.
This is an interesting proposal, let's hope it gets picked up and appear in more licenses.
A very balanced set of recommendations from the SFC around LLM uses. It's just the beginning and still lacks a bit in details. It's very welcome though and I look forward to their updates.
What happens when targeted scams become cheap to run? This covers it fairly well, and we need to change our heuristics and trust model.
Badly apparently, looks like it makes for prose people avoid. Now the thing is... with the widespread suspicion, some people might be wrongfully flagged as using LLM to write their posts.
What a surprise... It turns out it's very easy to manipulate AI "search". Something which operates of statistical similarity to queries, who knew it could be manipulated. 🙄
The latest move by the US government treating LLMs like dangerous weapons tells something about the geopolitical moment. Can we collectively raise to the challenge and build on cooperation instead? It'd be a much better position than assuming governments or big companies will make the right choices for everyone else in isolation.