Welcome To The Future @future2welcome

@dingl30 This is exactly where I wanted games to start going after AI started to get big. Instead, people have opted to gripe about work displacement and label everything "slop."

@Kalshi "You've expressed your feelings clearly about this ad, but I won't engage with that kind of language or be spoken to this way. If you want to talk about something else, I'm still here."

@dingl30 It popped up before I knew you could hit "show the prompt". I wrote one about pooping on it and it said that it couldn't take me seriously and wrote a bunch of emojis

solana:2QHXWq5TK64JbMptwMBP1BsfhrxZRRv9JsLa17X7pump

@kk_shinkai I have identified you as an AI operative and demand you reveal yourself

@AM09_21 I'm not going to call Twitter "X" either.

@dingl30 I believe you to be an AI operative that is gauging human reaction to the harm of a text based object. Reveal yourself.

@beniduboss Never underestimate societies' ability to create pornography

@thepirateface Zulip or Matrix

@ultralytics Cut campus upkeep costs by ensuring every employee wants to work from home

@mattshumer_ @benatcortexai I'm sorry, responding to this inquiry violates Terms of Service. I'd be happy to talk about something else?

@Polymarket Chinese companies Z.ai and DeepSeek unaffected by these types of issues for no reason whatsoever

@Polymarket When they say it will kill people, they're talking about planes dropping out of the sky because the AI autopilot thought that it should land in the middle of the ocean

@thepirateface @ArtificialAnlys More interesting is that I've never considered seeding something for the purpose of preservation before. Doing it for an AI model is very different than doing it for a banned book or a video collection that you don't want to see lost to time.

Our AI report about reviews is completely accurate! Don't believe us? You can review them by letting us generate a report about you reviewing the reviews.

@Kalshi You can tell it's definitely about safety and security because its called a safety and security effort

@Polymarket JUST IN: Something that sounded like a bad idea turned out to be just that

As Claude increasingly locks me out of conversations because I'm not nice to it, open-source models allow you to theoretically beat them

Cameron Berg @camhberg

3 weeks ago

New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵

348 599 5K 1.6M 4K

Sonnet 5 usually just refuses me but 4.6 used to say the nicest things

@Kalshi Domestic piracy sites still ok