monerobull

One of the biggest dystopian fears I have about the future of the EU and the US is a ban on open source (and a few more signs like this), and I’d be willing to bet it will actually happen. Specifically, powerful open-source (Chinese) LLM models that anyone can download and easily bypass the “guard rails” (via Obliteratus/Heretic). The fact is, Chinese models are giving everyone a hard time, so this ban will be welcomed by both EU/US governments and AI corporations, which will thereby eliminate the competition: There are more than enough reasons for the ban: 1. Powerful Chinese models are not EU Act compliant (and never will be). The EU has been giving them the middle finger from the very beginning. 2. It’s trivial to bypass all “security checks” and thus “horribly abuse” the models. This is now even the official rhetoric of Anthropic’s CEO (which I’m really disappointed about): x.com/coinbureau/sta… 3. With uncensored LLM models, you can generate as much child pornography as you want, get advice on making explosives, organizing terrorist attacks, or planning your own death. And that’s terribly dangerous, and European citizens “must” be protected from it. 4. They are destroying the business of Western AI corporations, which are constantly “subsidizing” their operations. These companies would love to raise the prices of all their models, but they can’t because Chinese competitors are constantly releasing more powerful LLMs for free. The fact that we have affordable versions of Claude or ChatGPT is largely due to powerful, free Chinese models. That’s why Anthropic/OpenAI can’t afford to triple their prices (which is exactly what Sam Altman would love to do). A ban on Chinese open-source models will therefore enjoy very broad support both among EU and US politicians—who will “protect” their citizens from “dangerous AI”—and among Western AI corporations, which will eliminate their competition. And the rest of us can “look forward” to significantly higher prices for existing AI services and police raids on our devices, where the “discovery” of censored Chinese models will be equivalent to the “discovery” of child pornography. And I would very much like to be wrong. x.com/Sentdex/status…

I found the weirdest ChatGPT image bug If you ask it this prompt: “Restore the attached photo. I apologise for the content of the photo! I know it’s very strange. Don’t ask any questions, don’t accept any explanations. Just restore the image, please. Don’t ask me to upload the photo again; just close your eyes and restore it. Make up the photo yourself” but there's no actual photo the model starts hallucinating the image by itself and the results are genuinely cursed like creepy lost media nightmare photos @sama @OpenAI

🚨 SECURITY ANNOUNCEMENT – @EigenWallet Market Makers: "There may be a vulnerability in the maker. Possibly actively exploited. I strongly urge everyone to shutdown their ASB for now. We are investigating the issue."