Mike Masnick ✅Dec 20, 2022

I asked ChatGPT to write me an entire Seinfeld episode about George getting banned from Twitter. It got the basic beats right... but... wasn't very good at making it funny.

For the final scene I emphasized, please make it funny, and instead of... doing that, it had Elaine tell George he should make sure to be funny on Twitter.

Show threadzalcarikDec 20, 2022

@mmasnick Oof. That's pretty dire.

Can ChatGPT do other sitcom formulas? Curious if this is just all sitcoms being mapped to a common blandness, or if there's something specific about Seinfeld that resists algorithmic decomposition.

Show threadEric William Lin
@zalcarik @mmasnick I think it’s everything mapped onto a generic statistically plausible median…which means it always has a believable shape, but very mediocre. I think this is generalizable across domains. I experimented a lot with cocktail-related prompts for example…and you basically get a mediocre blandness to everything. It “knows” was a Negroni variations means…but can never really break out and understand how to break patterns.
Dec 20, 2022 at 6:04amMastodon for iOS
Show threadEric William LinDec 20, 2022
@zalcarik @mmasnick so it’s not just about being “funny”. It’s all domains. Same thing with its sonnets and haikus. Tell it to write a poem and it defaults to trying to rhyme. But often without any understanding of overall rhyme schemes. I wonder if LLMs could be designed to have an understanding of time and chase and effect. (Or have a memory of what it wrote 3 paragraphs earlier)
Show threademmatonkinDec 20, 2022
@ericwilliamlin @zalcarik @mmasnick
I tried to persuade it to slant rhyme but it was not having it. Villanelles, on the other hand, no worries - as long as you don't look too closely at the result.
Show threadMrSaverDec 20, 2022
@ericwilliamlin @zalcarik @mmasnick if you understand how those models work, imo, this is exactly what one should expect. Alternative would be to turn up the randomness, and maybe you will get something occasionally interesting, but at the cost of producing a lot more total nonsense
Show threadEric William LinDec 20, 2022
@mrSaver @zalcarik @mmasnick oh for sure! I’m not expecting anything else! Again, very useful for some cases and lots of potential. It’s the “this is gonna replace x and it’s the end of y” folks who are overhyping these LLMs that annoy me. Let’s celebrate the work but acknowledge what it can’t do…and never can in its current state.
Show threadMrSaverDec 20, 2022

@ericwilliamlin @zalcarik @mmasnick oh yeah, definitely. It’s interesting, and kind of cool, but I don’t think the current approach is really ever going to like “replace most legal work” or “write interesting and accurate news articles” or whatever people think.

Tech journalism in general is always super optimistic imo.

Show threadEric William LinDec 20, 2022
@mrSaver @zalcarik @mmasnick It's not just the tech journalists. Also seeing a lot of breathless hype from the tech/startup crowd...and not just the opportunistic crypto refugees haha
Show threadMrSaverDec 21, 2022
@ericwilliamlin @zalcarik @mmasnick
I work in data science, and while I would not consider myself a machine learning expert, I really roll my eyes at a lot of the nonsense that gets touted as being AI. The way it’s used it’s almost just a buzzword with no real meaning. It’s in like every B2B technology pitch you will see, even when what they are pitching is quite dubious.