Anthropic and OpenAI announce more powerful (and cheaper) AI models

Researchers have been exploring the idea that AI models could be more powerful and cheaper to run if they disclosed less about how they worked, but this trade-off could make it harder for humans to check the systems. https://www. japantimes.co.jp/commentary/20 26/09/20/world/opaque-ai-security-dilemma/?utm_medium=Social&utm_source=mastodon # commentary # worldnews # openai # astra # ai # artificialintelligence # anthropic # llms # samaltman # chatgpt # cybersecurity

Suggestion: 40% of your quota left at the end of the week gets transferred as bonus quota for next week. look i personally do want anthropic to have more compute available, because i really like them over openai. but they make their product EXPENSIVE AF. if im paying 100$ a month to be able to use fable, you bet your ass im gonna use every last drop of the money i gave them. why am i paying them rather than openai the company that actually gives timed resets that you can use whenever you like? i wanna support anthropic but if they wanna extract my money, then why should i not at least be allowed to extract my money's worth? there is 0 incentive for a user like me to leave any quota at the weekly reset time. so why doesnt anthropic do this, if i have some quota left at the end of the week,

I've been thinking about the whole AI slowdown discussion, and something doesn't quite add up to me. Anthropic's CEO has been calling for the industry to slow the development of increasingly capable AI models so there's more time for safety and evaluation. But now Anthropic is reportedly considering another new model as it faces growing pressure from OpenAI. And I don't know honestly, I don't think that's necessarily contradictory. That's probably the problem. If you're an AI company competing in a market where everyone else is still improving their models, how do you actually slow down without giving up ground? Maybe “slow down AI” doesn't really mean stopping the race. Maybe it means slowing down the release of capabilities while the underlying AI development keeps moving. What an actual

Sam Altman: AI models need their own device so they can stay with you and understand the context of your life. “The iPhone is currently the greatest piece of consumer hardware ever made.” “But it was not meant for a world where you needed hardware that could absorb all the context of your life.” “My phone is in my pocket, and it’s not gonna understand us.” “I would like a device that can participate, understand, and know about this conversation.” “And OpenAI is committed to building something along those lines.” #AI #Anthropic #AISafety #AGI #ArtificialIntelligence #CNN #TechNews

Anthropic and OpenAI both shipped new models today, GPT 6 Sol and Opus 5.5, and I wanted to pull up a proper head-to-head. On raw scores, Opus wins every benchmark in the table. The one that surprised me is GDPval, the "real office work" eval. Sol actually scores about 100 Elo lower than the GPT-5.6 Sol it replaces, and Artificial Analysis says the drop mostly came from weaker presentation and deliverables that skipped required parts of the task, not reasoning failures. Sol's advantage is cost. It's half the price per token, and Opus uses a lot of tokens. At max effort, AA measured Opus 5.5 at around 119K output tokens per task, while Sol used about 31K and cost $1.06 per task to run the whole index. Genuinely curious to see how these perform out in the wild and what feedback we get over t