Anonymous 09/28/2026 (Mon) 12:54 Id: c841d2 No.194326 del
>>194294, >>194295, >>194296, >>194297, >>194298, >>194299, >>194300, >>194301, >>194302, >>194303, >>194304, >>194305, >>194306, >>194307, >>194308, >>194309, >>194310, >>194311, >>194312, >>194313, >>194314, >>194315, >>194316, >>194317, >>194318, >>194319, >>194320, >>194321, >>194322, >>194323, >>194324, >>194325
DataRepublican (small r) @DataRepublican - Yeah AI is surprisingly bad at “identify this brand/estimate this” problems because they’re biased towards what’s common in their training data
Parent Post:
James Tucker @JamesATuckerDC
Im not saying AI is stupid because I come off looking stupid in this post. But I bought a uses car and I thought it had some unconventional tires and rims. I wanted to see if they were stock or if they had another name. It immediately misidentified the tire size (claiming it read the size on the side wall) and thus the entire conversation became incomprehensible. It took me circling the tire size in the picture before it admitted it was wrong. It finally gave me an answer and “oh by the way you can find the correct size of tires on the sticker on the inside door jamb” (which is why I look stupid). But interactions like this is why when someone claims AI is going to rule the world I just shake my head.
https://x.com/DataRepublican/status/2104279454034051134

DataRepublican (small r) @DataRepublican - I don’t even know what he’s trying to argue from a technical perspective. If you’re trying to test AI will misbehave, why not look at the attempted commands from an audit log? Why do you need deliberate security holes? “Oh because AI is so smart it can detect you’re trying to audit it” … everything with these guys lead back to manifesting their silicon deity
Parent Post:
Hensen Juang @basedjensen
>Except some of us have been researching and working in the field for over 10 years and have a far better understanding of its workings than whatever you think you have.
No you have not worked on alignment research for 10 years. You all think reading less wrong and yud fan fiction is "research" it is not go show us the papers that go 10years back if you have been doing this
>Your knowledge on how these systems work is strictly just wrong lol. Go read a book
Mf I am systems architect that has built and run largest clusters in the world for over 15 years. I KNOW exactly how these systems work far better than you safetyist who think you have basic understanding of computing and networks
https://x.com/timducktim/status/2104214534252609668
https://x.com/DataRepublican/status/2104278699889877312

Message too long. Click here to view full text.