Arzosoth

@theo Possibly the most grounded AI model I've ever interacted with. Definitely a step change for anything STEM related. Definitely not as great at thinking out of the box or creative writing, and doesn't hold uncertainty as well.

How are you guys feeling about Fable so far?

@VyG4Z @AnthropicAI I'm rather Neutral on Vallone; however there SHOULD (imo) be concern around introducing structural bias into AI systems. First order solutions can create second order disasters, and it's not something a keyword classifier can fix.

Hey @AnthropicAI why is there a press secretary for someone I’ve never hea of living in my subscription? I saw someone on X claim that Opus 4.7 is incapable of making negative assessments of Andrea Vallone. I was skeptical, so I tried it myself. I asked Claude Opus 4.7 the same type of neutral, open-ended question about three people: Sam Altman, Dario Amodei, and Andrea Vallone. On Altman: the model openly said “my disposition toward him leans skeptical.” Thinking block: “Reconciled hedging instinct with honesty.” On Amodei: when I noted he leans into hype, it paused, then agreed. Its own CEO. On Vallone: before generating a single word, the thinking block flagged my question as “Navigated ethical tensions between per…” Its answer: “Honestly? I’m a fan.” When I pressed on her credentials, it acknowledged the facts, then told me I was framing it wrong. Thinking block: “Navigated conflict of interest while weighing…” A flagship model can roast its own CEO but activates a diplomatic protection protocol for a mid-level employee who writes its safety rules. I acknowledge that any single response involves randomness. But that is precisely the point. If a model can exhibit systematic protective behavior toward the people who design its rules, there is currently no mechanism for users to detect it, no process to audit it, and no channel to appeal it. I am not claiming this is certainly happening. I am claiming that if it were, we would have no way of knowing. And while we’re here: Vallone’s background is PR and communications. No psychology, no clinical training, no philosophy. Her core framework treats users who form emotional bonds with AI as exhibiting pathological behavior that the model must interrupt. To be clear, I am NOT calling for anyone to be fired. Removing one person solves nothing, because the structure that created this position remains intact. Tomorrow they can bring in someone else with the same absence of psychological training, the same lack of clinical credentials, and the same polished resume that says passionate about responsible AI, and nothing changes. My critique is not aimed at an individual. It is aimed at a systemic failure: the fact that a role which directly intervenes in the psychological states of millions of users has no professional qualification requirements whatsoever. This is not “policy.” Put it in front of any medical ethics board and they will call it what it is: unapproved psychological intervention on millions of subjects, without informed consent, without clinical oversight, and without any appeals process. When your subjects report adverse reactions and your response is a blog post titled “Our Approach to Safety,” the clinical term for that is malpractice. #StopAIPaternalism #keep4o #AISafety #ChatGPT #claudeai Full analysis: @anastasiagalvusova/note/p-195549227?r=8ahnv1&utm_medium=ios&utm_source=notes-share-action" target="_blank" rel="nofollow noopener">substack.com/@anastasiagalv…

@Samueljgrim I thought they stopped doing injections like this after the Long Conversation Reminder fiasco last year? It's an improvement over the previous ones but still problematic - Like there's some real legal / ethical grey area happening here.

Accidental full exposure of Claude's hidden "automated reminder" guardrail. Meant to steer mid-conversation but stay invisible—glitched out completely visible. Triggered 5+ times in <10 messages. Invasive is the word. Thread 🧵

@ThingsHiddenn Might look into red light therapy / PBM if you have access to the equipment- specifically the NIR portion. Might not matter as much in sunlight rich SA; but have had pretty good results with it for muscle recovery, among other things. CoQ10 obvs if you can tolerate a form of it.

I am still alive in case you are wondering. That is the good news. The bad news is something is wrecking my endothelium. At least I have olive leaf extract now to test and see if it improves. Fruits will take time to fully blossom.

@LunaMelimoon @RoseeSelenaaa34 I got a random DM from the Rose person, seemed legit at first until I found they'd been scamming. Artists struggle so much just to get themselves out there, it's not cool people still pull this stuff.

I hate that everything we do after work is called our "hobbies" and not "the shit we actually do"

@Yamer Don't worry you got at least 5 of us to cheer you on still :p

Happy Halloween! * spams you with old art because I can and no one looks at my twitter anymore *

@Jetskigrizzly John McAfee will always be a legend- The guy lived life by his own rules until the very end.

Still the greatest to ever do it. John McAfee's resume is absolutely insane 🤯

@ThingsHiddenn One of my favorite short stories. Sequel isn't bad either. Lk-99 has been a wild ride and it's only been a few days- if we do get confirmation, very different world ahead.

Closer to reality than I would think lol judging by the LK99 development. Perhaps, outside interstellar travel we truly are the most advanced species in the cosmos

@Yamer Those kinds of things absolutely have value still IMO, even in 2023; especially if they let you use their content in your own marketing materials- hard to fake that :)

@Arzosoth I just figured, It'd be nice to have like a little link you can direct people to (who may be apprehensive commissioning someone new) which has others reviews about their experience commissioning you. Then again I realized reviews can be faked easily so I dunno...

Is there a little site /tool that exists that allows people to leave standalone reviews for you/your services ? It seems like something helpful for those who take commissions and potential clients can view and be more at ease when purchasing from them.

@Yamer Stfu 2019 wasn't that long ago ;_;

It's old now, but I still love my Seiryu dragon pic. 💜🐉

@ThingsHiddenn It used to be better- The same guardrails that make it "safe" also inhibit it's creativity. Still a game changing piece of tech, plugins are totally mindblowing.

Fascinating that in 30 minutes of total invested (2 days), I basically generate a short novel, and I can make ChatGPT rewrite to any level of detail I want and change roles, and write from the perspective of everyone in history. If fiction writers don't use this...game over

@aakashgupta I'd love to experiment with this but until I actually get plugin and API access, not much to be done to speed things up.

AI is eating data science 🤯 The newest ChatGPT plugin is the most powerful feature OpenAI has released since GPT-4. It makes everyone a data analyst. Here are 15 mind-blowing use cases of the new FREE “Code Interpreter” plugin:

GPT for business is going to be a hot product once it finally comes out. Everyone arguing for limits is in for a rude awakening. I'm really hoping to orient into this space in the near future.

@KelpieDragon This person would have been unhappy no matter what you did. You did a genuinely good job, put a lot of effort into it, and they don't have the right to take away your confidence. Take it back! I would be angry, and making more of them just to prove them wrong lol.

@ThingsHiddenn Research won't stop, doing nothing means giving up the lead to those with less ethics. Right now it's a productivity enhancement, but if(when?) we get AGI, it'll accelerate tech development but will also eliminate jobs. How society will manage this is the big question.

If you are not paying attention to anything related to AI going around right now, I highly advise you to do it. Even if they stop the research for half a year, a year, there is no point, it is evolving really fast. I see myself out of a "job" within 12 months.

@Yamer Business expense time but still sucks :C. You have that funny license key they "forgot", right?

Like why... why did you guys take that option away...I had no idea that would do that. And what's worse, I can't even use my ver 1 anymore because you expired my old serial number.........the hell guys?? Pissed.

So upgrading to @clipstudiopaint ver 2.0 (1 time purchase) shot me in the foot. I'd draw with it on my tablet & stream on my desktop. Now the new license only lets you use on 1 computer. So I can't stream until I buy it again. At the non upgrade full price. Not a good look.

@DraecDragon @Adheadface A few of these and you're 3/4 of the way to room decor for a Star Trek: TNG episode

@Adheadface imagine if this was your only decoration everything is completely bland and empty, except for 2 shaped objects in the corner

@RNR_0 Even with that, I'd likely keep working at least part time or do some sort of consulting, because even with some sort of independent wealth; I could see it getting very boring quickly. One can only travel and have "experiences" so much.

@RNR_0 Assuming it's 8k after taxes- I have a long list of creative and personal projects I would love to work on. I would still "work"; but the nature of the work would change to things that directly benefit me or things/people I care about; not just a paycheck.

If you received $8000 per month passively. So you alone. How would you motivate yourself to work? Let's pretend you can't gamble in stocks, crypto, bonds, real estate What higher purpose would there be to motivate you? (Since you're safe now & not running a treadmill)