Rogu

[Rogu @RomanGuy20]

@vuchienthang93 @SawyerMerritt Well… agreed. Military applications are interesting.

[Rogu @RomanGuy20]

@Axilwood_CEO @TheDigitalArb @niccruzpatane @SpaceX Like, if space DCs get traction, I am only happy if Elon gets a good cut out of it. His SpaceX and Starlink were great, Teslas are fine too.

[Rogu @RomanGuy20]

@aaronp613 But watchOS 27 does not support Series 8 and older. Tough luck. Guess, watches are not supposed to have longevity. 🤷‍♂️

[Rogu @RomanGuy20]

@theo Unusable. Guardrails are too strict and have a ton of false positives, even though I don’t work with bio or cybersecurity at all.

[Rogu @RomanGuy20]

That’s irrelevant. If something violates ToS you flag it explicitly, not silently sabotage it. Let alone this being unethical described in such broad strokes, given how trigger-happy current guardrails are for bio and cybersecurity, they could be as trigger-happy for LLM development detection, sabotaging absolutely unrelated work.

And then nothing stops from expanding the scope of such silent sabotage to other fields with time. Maybe even without announcing it.

[Rogu @RomanGuy20]

@roanoke_gal @irl_danB I don’t believe OpenAI will do that.

[Rogu @RomanGuy20]

@aidan_mclau Please don’t follow the same unethical approach as they do with Fable release’s “security” measures.

[Rogu @RomanGuy20]

@xwang_lk Yes.

Appearing vs. being.

[Rogu @RomanGuy20]

@AISafetyMemes Honestly, forcing these out sounds a bit like torture at this point.

[Rogu @RomanGuy20]

Well, given how much they delayed the release, they had A LOT of time to about releasing … that.

[Rogu @RomanGuy20]

@deanwball Also given how trigger-happy it is to label nearly anything as bio- or cyber-risk, it would likely also be trigger-happy with seeing frontier-LLM risks, proliferating this negative impact even to unrelated fields.

[Rogu @RomanGuy20]

@nabeelqu We need Net Neutrality, but for LLM vendors. All requests should be treated equally.

[Rogu @RomanGuy20]

1. The problem is that the EU tries to force companies to apply its laws outside of its borders. Like GDPR, which led to stupid cookie banner contamination even in more sane regulatory landscapes.

I would only support the EU doing whatever if that would not affect me in the US.

2. EU should only have the right to impose fines on EU portion of revenue. It doesn’t make sense for EU to lay claim on global revenue, that is outside of their borders and jurisdiction, unless the company is headquartered in the EU.

[Rogu @RomanGuy20]

@bubbleboi That’s likely what they want. Nationalization = natural monopoly and state-level protection with them being able to roll with anything they like.

[Rogu @RomanGuy20]

@theo One of a set of very few questions it can answer without triggering security block.

[Rogu @RomanGuy20]

@nicdunz Fable is unusable though. It just flags almost every request, even unrelated to biology or cybersecurity.

[Rogu @RomanGuy20]

@Here_We_Go111 @Sebikom1 @EC_AVService There are a lot of Chinese phones as well, including non-Android ones. And then EU company could build one. But nobody would buy it, because this problem is a non-issue for consumers, they won’t choose one product over another because of it.

[Rogu @RomanGuy20]

No, Anthropic just seems to get some perverse satisfaction in nannying users for the fake sense of security. They have been crying wolf since their early days, and yet none of the risks have materialized. We now have open-weight models which are much more capable than, say, Opus 4, have no guardrails, and yet the world stands.

[Rogu @RomanGuy20]

@daridor @DeryaTR_ @AnthropicAI But that’s gatekeeping. You have to beg to access the useful version of the model even for simple tasks.

[Rogu @RomanGuy20]

@DeryaTR_ @BoWang87 Every time I give a try to Anthropic, I remember why I love OpenAI the most out of frontier labs.