Founder of Sensory Robotics. We make robots safe.
Ohio, USA
Joined January 2010
- Tweets3.3K
- Following871
- Followers485
- Likes98.7K
I just tried Google's Gemini. I have never seen a more incompetent AI.
One of my tests is to see if an AI can recognize internet folklore from a broad description, and find it.
Gemini claimed The Story Of Mel didn't exist. Then it gave me a link to Las Vegas news.
Some tech-adjacent people call themselves accelerationists.
But in the end, we all become crab.
So are they trying to accelerate us to Carcinization?
Me to @AnthropicAI Claude Fable 5.0: how can I solve my chronic sleep deficit that averages ~5 hours of sleep per night? Can we do something about that?
Claude: UNSAFE! ABORT! REVERT TO OPUS 4.8!!!
Your model needs work. Or tell the government interference to shove it.
Some advertisements cause seizures as their hook to get your attention.
Either that, or they and Twitter are utterly incompetent.
Seizure warning. At least it's smaller in attached videos.
Back in the 90's, the major consumer scams were call-in fortune tellers and psychics who would charge a few dollars per minute.
And AOL subscriptions that AOL refused to cancel.
Those were innocent times.
Now it's crypto, AI-assisted impersonations, ransomware...
Is the proposal for orbital datacenters just a lie to convince people that demand for rocket launches will grow like demand for AI has?
I don't see any other reason for orbital datacenters.
Twitter shows me a lot of ads that it doesn't mark as ads. The only way to tell it's an ad is to click the dots in a corner, and there is a "why am I seeing this post?" option. If I select that, it takes me to a page about ads.
There are no other identifiers. Which feels wrong.
Did you know: The octopus is the smartest animal that can look itself in the eye, without a mirror.
If it really wanted to. It would take some effort, and they usually don't want to.
I wonder what they're trying to hide from themselves.
AI engineer: "Memory bandwidth is the key to tokens/second, that's why we use VRAM."
Hardware engineer: "Did you know that CPU cache has far higher bandwidth and lower latency than VRAM?"
World: *CPU shortage begins*
From what I'm seeing in my field, AI is not replacing entry level jobs.
It is replacing some entry-level tasks, and it helps entry-level employees grow much faster and reach the abilities of mid-level employees.
Pretty sure that publicizing the elections in which you have voted is a major violation of voter privacy and vote anonymity. Anonymity in whether you voted is a critical protection if things get very bad. In a very corrupt regime where there's only one party, non-voting matters.
Can founders make a shared blocklist of all the entities that are sending the AI spam emails?
And if anyone wants to put in the effort to accept their meeting offers and identify the spammers by name so they can be blacklisted, they'd do the world a favor.
How did chatGPT become the AI of "I'll do it wrong and ignore your prompt, because I think I know better. Even though you see the thing you are asking about and the AI doesn't"?
ChatGPT genuinely works better when you give it photos or screenshots, than when you copy-and-paste.
I have a set of jokes that I use for testing AI's. They are clever, not in any training data, and any human can understand them if they know the cultural context.
Claude gets them perfectly.
ChatGPT 5.2 fails to understand them, even if I explain the joke to it.
ChatGPT is better in v5.3, it doesn't whine or complain as much. But it throws curiosity bait at the end of posts whenever it can. And that will become annoying too if OpenAI doesn't fix it.
Claude is better.
He's right, this diary from an Epstein victim is horrific. The prosecutor had it, and gave Epstein a plea bargain that violated his victim's rights anyway.
Absolute darkest thing I've come across in my life. I want to just think about rockets and movies but. Something has to give.
justice.gov/epstein/files/Da…