ADHD got me fucked

It's talked about like being obsessively neat or repeating pointless tasks is the main part of it, when really those are just potential symptoms.

The main thing behind OCD is not being about to turn off a thought. There's a thing where most people can just stop thinking about something. If it's over, it's not relevant, it doesn't matter anymore, people can turn their attention away. For OCD, that mechanism can get stuck. And some thought that was supposed to just temporarily pass through your head just stays there. An image of an object rotating. An anxiety about something bad happening. A wish that you made on a dandelion. These are all things that have at some point gotten stuck in my head, sometimes for years at a time.

The compulsions, the rituals, are the person trying to address the thought so it can go away. After all, if you're worried about the door not being locked you can check the lock. But for someone with OCD, that doesn't make the stuck thought go away. So they check it again. And again. And they made a ritual, maybe if I check it exactly five times, I'll know that it's locked and I can let this worry go.

It helps a little. It feels like you're doing something. But it doesn't solve the problem. Actual therapy for OCD involves not doing the compulsion. Instead, you ignore the thought, move around it, try not to give it space in your life. Your mind won't let the thought go normally, so instead you fill yourself with other thoughts. Other parts of your life.

At least people are no complaining that the prices are out of date because the graphic was made in 2019. I had to reblog it several times with links to the artist, Dani Donovan, and disclaimers, and addendums, and amendments, and people still misinterpreted everything somehow. The Tumblr userbase is really really good at misinterpreting and complaining when their exact experiences aren't featured.

There's something gallows humor funny about that right there. Priceless.

Here she is again, the person who made that graphic before the pandemic.

I dont care if 9/10 of the people who use the wheelchair ramp arent actually in wheelchairs. As long as the 1 person who needs it has access to it.

I dont care if 9/10 people who use the automatic push button on the library door can actually push the door open themselves. As long as the 1 person who the door is too heavy for gets to use it.

I dont care if 9/10 people who buy the can tab opener, or the little guitar clamp that holds the chords for you, or the hand grip that helps you hold chop sticks, don't need any of it and just get it to "be lazy". As long as the one disabled person who needs it gets access to it.

I do not care. Oh my GOD I do not care. As long as there's a disabled person on this planet who the accessibility device will benefit, the accessibility device is necessary.

Not sometimes. Not until the next update. Always. They proved it with math.

Even with perfect training data and unlimited computing power, AI models will still confidently tell you things that are completely false. This isn't a bug they're working on. It's baked into how these systems work at a fundamental level.

And their own numbers are brutal. OpenAI's o1 reasoning model hallucinates 16% of the time. Their newer o3 model? 33%. Their newest o4-mini? 48%. Nearly half of what their most recent model tells you could be fabricated. The "smarter" models are actually getting worse at telling the truth.

Here's why it can't be fixed. Language models work by predicting the next word based on probability. When they hit something uncertain, they don't pause. They don't flag it. They guess. And they guess with complete confidence, because that's exactly what they were trained to do.

The researchers looked at the 10 biggest AI benchmarks used to measure how good these models are. 9 out of 10 give the same score for saying "I don't know" as for giving a completely wrong answer: zero points. The entire testing system literally punishes honesty and rewards guessing.

So the AI learned the optimal strategy: always guess. Never admit uncertainty. Sound confident even when you're making it up.

OpenAI's proposed fix? Have ChatGPT say "I don't know" when it's unsure. Their own math shows this would mean roughly 30% of your questions get no answer. Imagine asking ChatGPT something three times out of ten and getting "I'm not confident enough to respond." Users would leave overnight. So the fix exists, but it would kill the product.

This isn't just OpenAI's problem. DeepMind and Tsinghua University independently reached the same conclusion. Three of the world's top AI labs, working separately, all agree: this is permanent.