What concerns/bothers me is the total absence of even an inkling of a doubt that the LLM may be making up or misrepresenting a key aspect of the paper
Posts by Stéphane Bisinger
A very nice discussion about how AI subtly fails us and makes us lose our critical thinking skills. I quite enjoyed it!
By @labmuffin.com
youtu.be/xcq5XYkFJfY?...
I just found out that Astral has been acquired by OpenAI, what a sad day 😢
ruff and uv are breath of fresh air in the python tooling space, now waiting for enshittification.
I hope a reasonable alternative comes up before the shit hits the fan
Didn't Bush Jr. end up being remembered as a really bad president, especially for the Iraqi war thing? We still make fun of the "Mission Accomplished" thing.
Trump makes Bush Jr. look like someone actually not that bad after all
And by the way, there's also the fact that humans will be correcting AIs mistakes. If it turns out that due to a coding error the data is wrong and someone spots it, humans will be correcting that, giving the illusion that "it's not that bad". AI will take the merits, and humans clean up the mess...
Furthermore the issues will get worse, because next time LLMs are trained on the internet data, they'll be ingesting AI-generated code, further weakening the results (it's a general issue, not just code).
So yeah, things WILL get worse
But AI is making our systems more brittle because it doesn't *understand* what it codes, but just takes it from what it learned - whether that was good code, bad code, outdated code, etc.
This doesn't cause immediately and easily detectable issues, but it accumulates weaknesses that are just waiting for someone to find and exploit them. Can you predict that the next databreach will be because of AI? No.
Can you say for certain that it happened because of AI? No.
Things are bound to break and get worse, but the effect is far removed from the cause. AI code tends to have subtle issues, and be very "conservative" in its form: AI code will use dated practices, approaches, libraries and doesn't learn or evolve over time.
youtu.be/K3b6SGoN6dA?...
I was born in an Iranian prison. My parents were held in their jails. My uncles lie in their mass graves. Nothing you can tell me about the crimes of the Iranian regime that I haven't lived in blood and bone. That doesn't mean I want my people bombed, maimed, killed, their homes in ruins. If your vision of liberation comes only through the destruction of innocent lives, then it's not freedom you're after. Sahar Delijani, Iranian author
I am pro Iranian people. I stand with them against their own government’s repression AND against the imperial violence in the form of bombs dropped from the sky from Israel & the US who have already killed hundreds.
I support neither imperialism or dictatorship.
Sahar Delijani said it best:
What a "Let them eat Brioches" moment
Actually sounds very familiar to what we do over at the Nowcasting Lab, where we always show what our projections were compared to the actual published growth nowcasting-lab.org
Sounds like my job...
I should look into it 👀
So all we have to do is get rid of a million babies or two to offset the AI, right?
So this Tech Bro is smiling when there's a genocide here or there, more space for his computers...
This is huge and, honestly, a bit unexpected. I wonder what is the political background that lead to this?
(Yes, this supreme court is extremely politicized. The fact that they broke rank on this is very significant)
Suggestion:
•If you hate writing, don't do it for a job.
•If you hate moviemaking, don't do it for a job.
I mean... I do what I do because I enjoy the challenge and the outcome?
If it was just about money and speed, I'm not sure I'd enjoy life.
I think that they are rather realistic about the span of their lives, and figured they'd be dead by then anyway
@edzitron.com would love(hate) this one
When people say AGI is inevitabile, that AI will only get better and better with each versione, it reminds me of self driving cars. People talked that way about that, and yet there's been no progress in years.
People paid who knows how much for a "Full Self Driving" feature that never came.
I need to talk to people who didn't drink the AI kool-aid, for my sanity.
I do not understand the world or people anymore.
Nobody seems to see the nonsense anymore, nobody cares for authenticity, accuracy, reliability or truth anymore.
As long as it's plausible, it's good!
Could anyone tell how this is different than the Germans hiding Jews from the SS almost a hundred years ago?
ICE is the American SS
Isn't that just translating the US rhetoric, which to me would be a sign that they're being propelled by US propaganda and interests. No?
Thanks for creating it, Adam!
What I find ironic is that AI's solutions to climate change would be a summary of what scientists are already saying today anyways.
The very same things politicians and tech CEOs alike are ignoring and not listening to.
Amazing video about the lies the Tech CEOs tell us daily about AI, in particular in relation to climate change and energy needs by @climateadam.bsky.social
Really love this one!
youtu.be/py0XpKxAnNU?...
Turns out that Microsoft's BitLocker security for the data stored on your hard drive is just a placebo.
Might as well give your password to everyone:
techcrunch.com/2026/01/23/microsoft-gav...
So here are some of the accusations and veiled threats to Switzerland (and Canada, too)
youtu.be/c_j7x0nyzHE?...