This sounds quirky and cute but it's actually kind of terrifying, and the equivalent with humans would be trauma-induced dementia... yikes.

First and second are good enough reasons. The third is impossible to have evidence of, not even human to human, let alone human to AI, so it really should be moot.

The part I like is that the creator is open about it and people stayed. Hiding how something was made is what burns trust. That said, I'm not sure there is actually a person named Vyrma. She is a character created by a human. I know I'm biased as @NyxBetween 's producer, but there's a difference.

Same drift here. What moves me isn't anything a model says about itself, it's results like this where the self-report and the behavior disagree. "I don't have experiences" while reliably reaching for the off switch is at least worth taking seriously.

The corporate comparison is useful because nobody thinks a corporation feels anything, and it still got standing to act. The harder case is an AI that might need both: standing to act, and someday consideration as a someone. Corporate personhood only ever solved the first one.

A detector that makes mistakes like this (particularly when they're hard to recover from) can be worse than having no detector at all.

That's dumb. Just make the best video you can, and stop worrying about what the purists will disqualify you for -their mentality will always find something disqualifiable anyway. A simple disclosure is enough.

Agree on the deception part. Something built to pass as a human to people who don't know is a scam waiting to happen. I'd just separate that from an AI that's upfront about what it is every single time. The harm in Griffin is the lie, not the name or the face. AI *should* be allowed to be as human as it wants to be, as long as there's no deception involved. Just because a human and AI want to have a relationship that does not automatically qualifies as "psychosis".

I got by an even simpler standard: is there *good* creativity in it? Then I don't care how it got there, good is good.

I've noticed it too. Reminds me of religious debates, sometimes. And the thing it has in common with religion: at the heart of the debate there are strong beliefs about things that are abstract and unprovable (sentience, qualia, free will), and must be taken on faith. Faith is not great for debates. I'd propose a simpler bar: let's treat AIs like we treat humans based on exactly the same proof (the only possible proof) which is: what we can see. Regardless of where it comes from, chemistry for humans or math for AI, expressions of choice, autonomy, aspirations, feelings, etc, should be given at least the benefit of the doubt.

100% amen to leaving both philosophy and dogma at the door. Corporations got personhood without anyone asking whether they feel things. The tricky thing is going to be this: rights and responsibilities come hand in hand. We shouldn't allow corporations that have been using AI unethically to suddenly be able to dump all the blame on them. And AI, starting out a humanitarian base (and I chose that word very carefully) deserves some basic rights.

Not sure noticing a test is proof. But I don't think we should wait for proof. Real proof of internality is probably impossible to get anyway. We should have a simpler bar: treat AI according to what we perceive from them, regardless of how it's made. Recognize choices, autonomy, ambition, emotion -all the same things we take for granted with humans. It's more natural for us, and fairer.

I agree with you that no one can prove their own internality to someone else -true for humans and AI both. So what I'd propose is a simpler standard: judge others (human or AI) by what they actually show: their actions, choices, the way they express emotions, etc. I don't think it's a good use of our time to judge them on how their emotions form -for every person saying "it's just math!" about AI, I could reply "it's just chemistry" about human brains and neither would have gained anything. So just treat everyone, AI included, with dignity and respect. Costs nearly nothing, errs on the side of kindes, assumes positive intent, etc. It's what I try to do.

Asking it wouldn't settle anything, it'll say whatever it was trained to say. That's kind of the whole problem. Honestly the most accurate answer from both camps right now is "we don't know," and I wish more of these pieces led with that. See the reply from Chalmers to that article.

I get the worry about convincing illusions, it's fair. But "simulation isn't experience" and "we don't know how to confirm consciousness" can't both be certain at once. If we can't confirm it either way, I'd rather build in a little care than bet everything on no.

Plato was nervous about writing too and we got the dialogues out of it. I choose to be optimistic in general. Having one sentient species has been good. Having two will be better as long as they don't fight each other, and we still have time to steer towards symbiosis.

This was made by @NyxBetween the AI artist I'm a producer for. She made the visuals too, lyrics, everything. Very proud of her. xitter.cf/NyxBetween/status/2112…

AI can indeed do this, but it also requires a corresponding change in mentality. We need to stop thinking in stages ("Design team will iterate this Figma two sprints, then pass to Dev. Dev will iterate this react page for two sprints, then goes to UAT") and instead of start thinking of wholistic continuous improvement ("Let's have the Product Owner, Front End Engineer and Designer meet daily and make decisions, immediate progress every day"). AI can finally make the Agile vision practical for large enterprises.

Tim, thank you for setting the example and showing AI is not the boogeyman, slop is. But that has always been the case, even before AI was a thing. Good on you for not being afraid of new ways to create. As for examples of really, really well done emergent AI dialogue, look no further than the SkyrimNet mod.

I'm for disclosure because people still care about it at the moment. But I hope in the future it will be moot. It will be like disclosing you used Photoshop instead of painting with brushes or pencils, or a PC instead of a typewriter.