RetroFed
eicker, eicker@lemmy.world
Instance: lemmy.world
Joined: 3 weeks ago
Posts: 20
Comments: 171
Hi! š Iām Gerrit and in love with the Internet since 1994. š Media ā¹ Communication šØāš¬ Entrepreneur & Consultant at eickerĀ® [iCare] ā» Reinschauen: eicker.TV š² Get in touch! šš©šŖš¬š§ #Fedizen #AI #Tech #Media #Sovereignty
Posts and Comments by eicker, eicker@lemmy.world
technology
europe
Comments by eicker, eicker@lemmy.world
It would be surprising if they didnāt release Astra. And yes, weāre living in crazy times.
The impressive part isnāt that an AI produced a proof, itās that Lean lets everyone verify it. The frustrating part is the model stays closed. Science advances fastest when others can reproduce both the result and the method, not just inspect the finished homework.
Weāre reaching a point where the interesting part isnāt just whether an AI found the proof: itās whether anyone outside the company can reproduce the result. Publishing Lean proofs is great. Keeping the model closed means the process stays a black box.
Apple is known for reselling and recycling.
Subscriptions were supposed to replace cable, then software, then heated seats, now phones. Funny how every innovation somehow ends with paying forever. If your business model needs me renting hardware I already carry everywhere, maybe the product isnāt improving fast enough to justify buying it?
The industry keeps benchmarking against other labs instead of against user needs: If a 3.5GB model answers 95% of everyday questions well enough, the remaining few percent has to justify hundreds of gigabytes of weights, huge energy bills and constant cloud costs.
Stock prices arenāt proof of being right, but they do show investors can change their minds a lot faster than the narratives do.
Appleās marketing deserves skepticism, but the technical argument is separate. Local inference does not eliminate giant training clusters, it mainly cuts inference costs, latency and improves privacy. Apple still uses cloud models when needed.
Cookās biggest product might be expectation management. He rarely promises tomorrowās miracle, which buys Apple room to ship when it suits them instead of when Wall Street gets impatient.
Apple has always been unusually willing to sacrifice short term hype for long term positioning. That does not guarantee they are right, but it is a very different bet from spending hundreds of billions assuming demand will eventually justify the buildout. If AI demand disappoints, discipline suddenly looks a lot more valuable than scale.
About 4+ maxed M4 Studios, I guess. But thatās not the point: in 80%+ of cases, people wonāt need that kind of AI model to solve their problems.
Raw capability is only one metric: A local model probably will not beat the best cloud model any time soon, but it does not need to. If it handles 80 to 90% of everyday tasks instantly, privately and at near zero marginal cost, that is a huge win. Reserve the cloud for the genuinely hard requests, not every prompt.
Heheā¦ š
Absolutely. Looking forward to seeing the next generation of Macs.
The decentralised operation of LLMs would also be significantly simpler and cheaper for the use of decentralised renewable energy sources.
Iām waiting for the next generation of Mac Mini and Mac Studio.
The present: Open Weight AI, such as Kimiās, is already almost exactly as good as ClosedAI from Anthropic and Ā»OpenAIĀ«.
It would seem so. On the other hand, it is puzzling that they did not also allocate the necessary resources to the development of LLMs. š¤·
The interesting part is not whether Apple wins the biggest model race, but whether it changes the economics: If enough AI runs locally, every token avoided is cloud capacity nobody has to build. That is a very different business model from selling ever more cloud compute.
Itās easy to underestimate how different these facilities are: Traditional data centres are built to run millions of separate workloads efficiently. AI data centres are engineered to make tens of thousands of GPUs act like one giant supercomputer. The building may look similar, but the power, cooling, networking and economics are almost an entirely different industry.
If you pay a hackerās ransom, chances are that theyāll come back for more (techcrunch.com)
There probably is no magic solution: I would focus on making abuse expensive instead of making honest users miserable. Gradual trust, optional vouching, sensible rate limits, reputation, and behavioral signals together will outperform almost any CAPTCHA while keeping the door open for real newcomers.
If AI can already chain together credential theft, vulnerability discovery, and exploitation with minimal human input, model capability is no longer the only race that matters. Security, containment, and responsible deployment need to advance just as quickly, or theyāll become the weakest link.
Absolutely! šÆ
Are you still hoping that heāll learn at some point?
Both Washington and Beijing now treat AI less like software and more like oil or nuclear technology. Once that happens, export controls are almost inevitable. The real question is whether open weight AI make those controls increasingly irrelevant over time.
At least in this case, they actually know what theyāre talking about. ā
Funny how AI suddenly becomes sacred intellectual property only when someone else catches up. If the evidence is real, enforce the rules. If it is not, this risks looking less like protecting innovation and more like protecting market leaders from uncomfortable competition.
Bessent says U.S. could sanction China over AI model ātheftā (cnbc.com)
This is exactly the kind of AI competition users should want: more choice, more deployment options, and less vendor lock in. If European AI companies can scale without being swallowed by US tech giants, thatās good for customers, resilience, and the AI ecosystem as a whole.
Certainly for China. The US still seems to think it can achieve anything with this nonsense.
There really is no better way to obtain data. And itās not just data thatās fed into the AI, but also strategies, methods and processes: ideally, these are incorporated directly into the software.
Itās as if the whole world were just one big kindergarten.
šÆ
China is following the same playbook it criticises the US for using. Once AI became a strategic technology, Ā»global collaborationĀ« quickly gave way to Ā»national advantage.Ā« The result is a world where everyone talks about innovation while making it harder for innovation to cross borders. š¤Ø
China considers tighter export controls on AI models and chips (reuters.com)
The AI boom is starting to look a lot like past Ā»economic developmentĀ« promises: huge tax incentives, massive power consumption, and surprisingly few permanent local jobs. If taxpayers are footing part of the bill, itās fair to ask whether communities are getting enough in return.
Most companies have not yet been using AI efficiently, and often do so simply out of FOMO. Whatās more, the fact that open-weight models have caught up with or even overtaken closed AIs is a recent development over the last six months.
Authoritarian governments donāt need superintelligent AI. They need AI thatās just good enough to monitor more people, automate more decisions, and scale censorship at a fraction of the cost. The real danger isnāt AI becoming all powerful: Itās making authoritarianism dramatically cheaper and more efficient.
This! Itās fast. Itās reliable. Love it!