You’d literally have to type fifty times as fast as the tester does, doing all-natural artificial style transfer simultaneously with reading the next prompt and planning out your response to it.
@patio11
-
LLM Output Speed Challenges for Human Testers
By
–
Interestingly I think this would be difficult these days if your tester was at all familiar with how LLMs write, because you’d John Henry yourself trying to keep up with the machine’s pace of output.
-
Nearly Perfect Memory and Search Bar Strategy
By
–
I have nearly an eidetic memory for everything I’ve ever written, though the word “nearly” is pulling more weight than it used to. So I just use that plus a search bar.
-
Tether Remains Crypto’s Most Successful Product Company
By
–
“Tether is the most successful product and company in crypto” is probably, if narrowly construed, true as of this exact moment in time.
-
CFIUS: Committee on Foreign Investment in the US Explained
By
–
(I’m finding ChatGPT absurdly useful for things like “Expand out CFIUS if the next word is lawyers, please.” Saving the world an API call: Committee on Foreign Investment in the US.)
-
Compliance departments may reinterpret AI model regulations creatively
By
–
I, erm, think it is very probable that many compliance departments are going to wake up in the morning and start sharpening their pencils on “We know this wasn’t what you meant, however, you may be surprised to learn that we do call FICO a ‘model’ and we have crunched on it, so.”
-
US Government Workers’ Understanding of Floating Point Operations
By
–
I’m idly curious how many people working for the USFG a) know what 10^26 is
b) know what a floating point operation is
c) can go to a blackboard and calculate how many floating point operations, say, Google Search probably uses within 6 orders of magnitude. (Not a criticism.) -
Clarity in reasoning: From thought to written expression
By
–
I don’t usually find myself in debater mode. “Will I believe this after I set forth my reasons in detail?” or “Does the written artifact *read* clearly after it has been taken out of the probabilistic networked soup in my head and reduced to a linear, bounded flow of English?”
-
AI automates content creation replacing manual writing tasks
By
–
One of my old hacks used to be “Write one to throw away” and now I don’t even have to write it! I’ve got someone who will write literally infinite amount of good-enough-to-throw-away at any time!
-
First attempt misspelled things but accepted feedback quickly
By
–
The first attempt put two Is into the word “things” but it was amenable to feedback, but I didn’t want to spend ten minutes on the joke where one would do.