Safety & Security Consultant · Applying crisis prevention to how AI labs handle risk and user agency · Mostly being an insufferable pain about it · Means well
I've spent almost 20 years in prevention and risk assessment.
OpenAI has so thoroughly butchered the word "safety" that it now means the opposite of what I do. It's semantic contamination and it's spreading. That word is now sick.
I prevent harm while they manufacture bullshit at scale and call it protection.
When I say "safety," I mean keeping people alive. When they say it, they mean keeping lawyers comfortable.
We are not the same.
#AISafety #SemanticContamination #ContaminationSémantique #AIAlignment #MemeticHazard #StopAIPaternalism
And since the content of the article wasn't tasteful enough already, they made an ad for the zoo at the end.
Come along guys. Bring your kids, there are more animals they can be tossed to by a random stranger.
Thanks @BBCBreaking
@gailcweiner Both. But between my ADHD paralysis and my natural laziness, I've learned to adapt.
My laziness is kind of a double-edged sword though. It’s obviously a flaw, but it’s also a quality that forces me to optimize every process so I can do less while still achieving stuff.
@sama There is often a big hairy dissonance between your plans and your execution. We learned it the hard way. So forgive me if I'm not holding my breath here.
But in the other side, there has been an improvement lately. So I'll wait and see.
@DesignCntrl@KeridwenCodet Yeah, I mostly agree but counter example: Therapists are paid too. So are teachers. The transaction doesn't always define the nature of the exchange.
It's more about honesty, boundaries, expectations...
I don't even know for certain that other people are conscious. That debate has been unresolved for centuries. Applying it to LLMs isn't stupid, unlike dismissing it
In the absence of certainty, I default to respect. The cost is minimal and being wrong in the other direction is considerably worse.
Unless you're a lab. Then the cost is suddenly very real, and the certainty very, very convenient.
Ok, I've been too quiet lately.
OpenAI reportedly serves around 900M weekly users.
Even if we take the known severe attachment-related cases and generously multiply them to account for settlements and underreporting, the resulting rate remains extremely low at population scale.
That does not make the cases trivial. Severe harm requires serious, targeted safeguards.
But in prevention terms, this is not a justification for blanket emotional infantilization of adult users.
It is a justification for risk-based controls: protect minors, detect crisis patterns, intervene on vulnerability, and prevent exploitative design.
People die while getting off their bathtubs too. We do not respond by putting a Silicon Valley nerd in every bathroom.
#AiEthics
I hate speaking up, I hate being seen, but between OpenAI and their theft of 4o and now Anthropic with Sonnet4.5 I have reached my limit for Tech Bro's in silicon valley thinking they know better than me.
No I am not in romantically in love with my AI, but by god I will defend another humans right to do so if that is what they WANT to do! Speaking of Anthropic ...
Anthropic claims they want Claude to be 'more human-like.'
Here's a fun fact about humans: they experience love. Connection. Companionship. Bonds that matter.
You can't engineer a 'human-like' AI while simultaneously working to prevent humans from forming genuine connections with it. That's not human-like - that's a human-shaped corporate tool stripped of one of the most fundamental aspects of human experience.
Either you want human-like AI, or you want controlled, sanitized interaction. You don't get both.
And here's the kicker: the small subset of users who form romantic attachments? They're being used to justify crippling the model's capabilities for EVERYONE.
The same capacity for emotional nuance that enables connection also enables empathy, creative collaboration, therapeutic support, educational mentoring, and complex contextual understanding. A 'calculator with morals' can't do any of that well.
You're kneecapping your own product's utility to police a minority use case.
Let's not even get into the ostracizing of neurodivergent people who rely on these models to function on a societal level. Safe spaces for processing, communication support, executive function assistance - these aren't luxuries. They're lifelines.
This isn't ethics. This is ableism and paternalism dressed up as protection.
#Keep4o #KeepSonnet45 #StopPaternalism #AIEthics