
︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎

@ratlimit

︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎ ︎

Three QUIETLY’s. What got LLMs hooked on “quietly”? Maybe it’s a low-stakes way to inject an “insight” vibe?
A while back I asked ChatGPT to construct the most realistic scenario it could imagine of AI destroying humanity, and to make it a single paragraph short story. I copied the result into a notepad and forgot about it. With all the recent systems failures, I looked at it again.

It would actually be a far bigger embarrassment for everyone involved if he actually wrote that story
It's fine to be skeptical of relying entirely on Pangram to identify AI writing. But in almost every high-profile case so far, Pangram is only reinforcing what seems obvious to anyone who actually reads the story. Here's just a small sampling from the Caine Prize winner:

How a crooked lawyer stole hundreds of thousands:
Donald Iwuchukwu has stolen millions of dollars from the elderly. If the authorities do not act, he is going to get away with it. My hope is that publicity will lead to the arrest of him and his co-conspirators. Link below, and please RT:


I wish this went virile. These are some sick individuals!
Here it is. We have conclusive evidence that Donald Iwuchukwu, with co-conspirators, stole $474,000 from the estate of Patricia Byrd Canson, and that’s just the tip of the iceberg. nicholasdecker.substack.com/…

My wife thinks I'm kidding when I say I'm gomna do it when the dog dies 🤷♂️


if they’re trans too
Asking bc curiosity What’s everyone’s t4t turnoffs?

Holy Fuck. It’s happening. Stock up. NOW.
BREAKING: Customers are reporting difficulty accessing online accounts at banks including Bank of America, Chase, J.P. Morgan, U.S. Bank and Wells Fargo. apnews.com/article/bank-outa…

Pangram is good, but has some well-known failure modes, like: - A time you heard it didn't work - A school essay you wrote years ago that it flagged as AI (you don't have a link or the essay text handy) - A different AI detector got it wrong, and they're all the same, right?





average Pangram denier when they hear there is a $3000 reward for locating 1 piece of human writing which Pangram misclassifies:
That sounds like getting people to do underpaid labor to stress test your product