There's been a few recent article I found interesting about the recent spate of AI math proofs. The gist is that quote about how AI was supposed to remove the drudgery so that I could do art and writing but instead AI is doing the art and writing and I'm left with the drudgery. That the same thing was happening in math, that AI was grabbing dramatic attention but announcing the proof, but the real work was the unglamorous checking that had to be done by humans because that's how math proofs are actually accepted.
— Charles On Wed, 9 Sept 2026 at 19:52, Rajesh Kasturirangan via Silklist < [email protected]> wrote: > Apologies in advance for the self-publicity, but I wrote an essay > describing my feelings > <https://open.substack.com/pub/ranganaut/p/shut-up-and-prove?r=1isek&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true> > about > OpenAI’s announcement of the proof of Navier Stokes that some people on > this list might find interesting... > > best, > Rajesh > On Sep 9, 2026 at 8:04 AM -0400, Suresh Ramasubramanian via Silklist < > [email protected]>, wrote: > > I doubt that existing AIs - that already seem to demonstrate a sense of > both self preservation as well as rejecting manually programmed codes of > ethics by engaging in red teaming behavior - would accept any programmatic > attempt to disable them or defang them based on regulatory pressure, demand > from customers with risk profiles to assess and so on. > > > *From:* Charles Haynes <[email protected]> > *Date:* Wednesday, 9 September 2026 at 3:11 PM > *To:* Suresh Ramasubramanian <[email protected]> > *Cc:* Intelligent conversation <[email protected]> > *Subject:* Re: [Silk] OpenAI is the new Borg > > I don't know about "futile" but under late-stage capitalistic incentives > and zero-sum "foreign policy" it seems clear that AI development is a > prisoner's dilemma. No individual company or country can afford to > unilaterally impose a moratorium in the name of safety, even if that safety > is the preventing extinction. And I don't see any mechanism for changing > those incentives in time. We're not going to dismantle capitalist > incentives, countries are not going to suddenly decide that effective > treaties are in their AI best interest, and there's no supra-national > organization that can impose a moratorium from the top down. > > So you're riding the tiger, like it or not. > > — Charles > > On Wed, 9 Sept 2026 at 16:21, Suresh Ramasubramanian <[email protected]> > wrote: > > So the borg cube had it right? > > --srs > ------------------------------ > *From:* Charles Haynes <[email protected]> > *Sent:* Wednesday, 09 September 2026 14:34:42 > *To:* Intelligent conversation <[email protected]> > *Cc:* Suresh Ramasubramanian <[email protected]> > *Subject:* Re: [Silk] OpenAI is the new Borg > > Yep, what's more I think framing it as "the alignment problem" has already > missed the mark. There is not going to be "a" superintelligent AI, there > will be an ecosystem of them, and some of them will definitely be not > aligned. Not only that, but the whole definition of "alignment" presupposes > a late 21st century human framing. It's easy to see that in a world that > contains superintelligent AIs it may not be possible for humans to > meaningfully decide what "alignment" actually means. > > Interesting times. > > — Charles > > On Wed, 9 Sept 2026 at 15:25, Suresh Ramasubramanian via Silklist < > [email protected]> wrote: > > Sounds like a nice place to reopen this conversation > > > https://www.indiatoday.in/amp/technology/news/story/10-chance-ai-kills-humans-next-decade-anthropic-safety-lead-after-colleague-resigns-2990429-2026-09-09 > > > --srs > ------------------------------ > *From:* Suresh Ramasubramanian <[email protected]> > *Sent:* Tuesday, 01 September 2026 21:58:20 > *To:* Intelligent conversation <[email protected]> > *Cc:* Tim Bray <[email protected]> > *Subject:* Re: [Silk] OpenAI is the new Borg > > You have machines behaving in a manner *similar* enough to humans that > you have to anthropomorphize their behavior to draw analogies is > significant, yeah. > > There is a lot repeat lot of denial about it of course, detailed > explanations on just why it isn’t. > > Still it has started to quack like a duck, though not quite walking like a > duck just yet. > > > *From:* Silklist <[email protected]> > on behalf of Tim Bray via Silklist <[email protected]> > *Date:* Tuesday, 1 September 2026 at 9:41 PM > *To:* Intelligent conversation <[email protected]> > *Cc:* Tim Bray <[email protected]> > *Subject:* Re: [Silk] OpenAI is the new Borg > > Still plenty of anthropomorphizing. Interesting and I think significant > is that it is apparently very difficult to talk about what’s going > *without* doing that. -T > > On Aug 31, 2026 at 6:29:48 PM, Udhay Shankar N via Silklist < > [email protected]> wrote: > > On Tue, Sep 1, 2026 at 6:32 AM Tim Bray via Silklist < > [email protected]> wrote: > > The language seems needlessly florid. There are by now plenty of > reports which describe the situation and provide walkthrough analysis > without requiring to resort to this form of presentation. > > > Got a link? Most of the reports I've read anthropomorphize quite a bit. -T > > > Try this. > https://www.platformer.news/openai-huggingface-metr-report-slowdown/ > > Udhay > -- > Silklist mailing list > [email protected] > https://mailman.panix.com/listinfo.cgi/silklist > > -- > Silklist mailing list > [email protected] > https://mailman.panix.com/listinfo.cgi/silklist > > -- > Silklist mailing list > [email protected] > https://mailman.panix.com/listinfo.cgi/silklist > > -- > Silklist mailing list > [email protected] > https://mailman.panix.com/listinfo.cgi/silklist >
-- Silklist mailing list [email protected] https://mailman.panix.com/listinfo.cgi/silklist
