When people try to sell you on the idea that the future is already settled, it’s because it is deeply unsettled. If they can get us to accept that the future is already settled — #AI is already here, and AGI is just behind the corner — then we will create that for them. Which isn’t very dissimilar than the American plantations elites in the 18th and 19th century trying to convince everyone that slavery was a preordained fact of life. Like many other AI professionals, I’ve been banging the drum for a long time that the LLM approach won’t scale, isn’t bringing us any closer to “real” reasoning, there’s no “reasoning scaffoling” that can “emerge” from it (just statistical cheating), it is burning the planet in the process, and it’s actually holding back real progress. I’ve advocated for a long time that the stubborn push for LLMs everywhere is just happening because the technically illiterate sociopaths who invested a lot of money into this just want to see their return on investment without having to wait on new progress. After reading this article I can clearly see a second reason. A vision of AI that is only accessible to a dozen of actors in the world because hardware constraints are so high that entry barriers are huge is a feudalistic vision of AI, where those who arrived first and can afford to invest billions into it see their returns come back 100x, with very little probabilities for everyone else to even get a chance at being competitive. Because if we invest into graph-based models that are orders of magnitude more efficient then everyone with a decent CPU can run models that can compete with ChatGPT. And making sure that nobody can compete against them for a long time is exactly their purpose.
New #Mozilla CEO: “Blocking ad-blockers in #Firefox could bring in an additional $150M in revenue“. Sure, even extracting and selling one of my kidneys could bring me a big revenue, but it doesn’t mean that it’s a good idea. Firefox by now has less than 3% of the browsers’ market share, and the only folks who still use it regularly are exactly those who don’t want ads constantly shoved in their faces, nor trackers to suck up every valuable detail about them. And they are willing to go the extra step and use a browser that is less supported than the overwhelmingly Chromium-based crowd out there. Take ad blockers away (and, most importantly, take away Manifest V2 extensions), and those few remaining users will just dwindle away to Brave, Vivaldi, Chromium/Cromite/Vanadium, or to support Ladybird/Servo, before you even finish saying the words “profit margins”. And of course their collective hallucination with AI has now crossed into the realms of ridicule. Nobody knows exactly what kind of new useful features will be added nor how those will solve problems that users want to solve. All we know is that “AI is coming”, embedded in a lot of low-value features that nobody asked, and that some product managers rushed to release just because someone at the top of the food chain asked to “put AI in their product”, and with literally zero chances of competing against the AI features provided by the browsers of companies like Google and Microsoft who can afford to invest into AI 4-5 orders of magnitude more resources than Mozilla. I’m wondering why Mozilla seems so committed to this mission of firing dumb CEOs just to hire someone who is even dumber and detached from reality than their predecessor.
The #PVV may soon have to modify its governance model. From a one-man party ruled by a single member (a guy who wins votes by vomiting hate towards foreigners), to a proper party with members and internal voting on things like candidate lists and electoral programs. Which is quite sensitive after all. If you want to participate in a democracy, then the democratic principles should percolate all the way through your governance model. Wilders’ reaction (“we must call the government accountable, not the other way around”, and “forcing governance models on political parties is a reflex of a totalitarian regime”) is instead indicative of contradictions that the civil society ought to call out: He and his party until recently were the government. His party got the most votes at the 2023 general election. He was part of a government, and he triggered its collapse (for the second time) when he realized that he had no idea of how of govern besides shouting “kutmarokkaan”, and that being a barking dog was electorally more remunerative than rolling up his sleeves. The dichotomy “we the barking party vs. the government” should be dead because populism itself is dead (if it even ever existed as a coherent political ideology). The contrast between different versions of “freedom” must be finally called out. Wilders’ definition of freedom is “my party is only about me, I get to decide everything with no accountability, and I can still participate in the democratic discourse even if I don’t believe in democratic governance and I violate many other freedoms (from the freedom of my supporters to pick candidates and ideas, to the freedom of any foreigner who lives in my country to live a dignified life without feeling treating like a walking target and constant scapegoat)”. The definition of “freedom” as “freedom to do whatever the fuck I want with little regards for others” must instead be openly called out as a reflex of totalitarianism.
If you want a rigorous analysis of why statistical #AI models collapse when continuously trained on their own data without external supervision and constraints, read this amazing paper from last year. If you want to get a visual intuition of how model collapse looks like, look at this video. When AI stares at its own reflection for too long, and its inference is purely rooted on statistics rather than reasoning, this becomes statistically inevitable. Keep this in mind whenever you hear someone talking about “AI models learning from their own outputs” without addressing the statistical parrot issue.
Michael Glasheen: #Antifa is the most immediate violent threat we’re facing on the domestic side Bennie Thompson: So where is antifa headquartered? MG: What we’re doing right now with the organization… BT: Where in the United States does antifa exist, if it’s a terrorist organization and you’ve identified it as number one? MG: We’re building out the infrastructure right now BT: So what does that mean? You said antifa is a terrorist organization. Tell us as a committee, how did you come to that? Where do they exist? How many members do they have in the United States as of right now? MG: The answers to these questions are fluid, it’s ongoing for us to understand this organization, the same no different than Al Qaeda and ISIS. BT: Sir, you wouldn’t come to this committee and say something you can’t prove, I know. I know you wouldn’t do that. But you did. The difference with Al Qaeda and ISIS is that in those cases American intelligence knew quite well who was at the head, they knew their chains of command, they knew their funding networks, they knew of terrorist attacks claimed by these associations as a whole, they knew where they were headquartered. Anti-fascists, on the other hand, don’t tick any of these boxes. The only glimpse of truth in the FBI’s operations director came in this sentence: The answers to these questions are fluid. Yes, they are. That’s the case every time you try to forge the image of a universal scapegoat, of a common enemy, and gradually stick that label on top of anyone you don’t like, so you can either get rid of them or deflect the blame of everything that doesn’t work on them. It could be the pro-Palestine activist. It could be your black or LGBTQIA+ neighbour. It could be the academic who reads Gramsci or Bertrand Russell. It could be the liberal IT professional in NY or SF. It could be the guy who plays Bella Ciao on the street. It could be anyone. It could even be you. That’s what “fluidity” means, in this case. The most brutally fascist meaning of that word.
I am compiling a comprehensive list of names and faces to hand out to my kid when he’s old enough to ask me who fucked up the planet, despite a century of collected scientific evidence about the upcoming climate catastrophe.
Linus Tech Tips: There was a recent thing from a major tech company where developers were asked to say how many lines of code they wrote, and if it wasn’t enough, they were terminated. And there was someone here who was extremely upset about that approach to measuring productivity because it means nothing. Linus Torvalds: Oh yeah, no, you shouldn’t even be upset at that point. That’s just incompetence. Anybody who thinks that’s a valid metric is too stupid to work at a tech company. LTT: You do know who you just said that about, right? Linus Torvalds: No. LTT: Oh. Uh… he was a prominent figure in the improved efficiency of the U.S. Government recently. Linus Torvalds: Oh, apparently I was spot-on. https://www.neowin.net/news/linux-creator-torvalds-thinks-elon-musk-is-too-stupid-to-be-working-at-a-tech-company/
After the European Broadcast Union (#EBU) decided not to put the participation of #Israel to #Eurovision 2026 to the votes, #BoycottEurovision is finally becoming a thing. Boycotting countries so far: Ireland Spain Netherlands Slovenia To be decided: Belgium Iceland Eurovision took a brave stance to support the rights of LGBTQ+ minorities in times when they were still denied in their own home countries. They also took a brave stance by excluding Russia from the festival after it attacked Ukraine. But they’re failing to take a clear stance against this #genocide, proving their partisanship in the minorities that they decide to support and the aggressors they decide to condemn. Hence it’s everybody’s duty to ensure that Eurovision 2026 will be about the Israeli music act alone, with no audiences to watch and cheer for them. image