Hacker News Reader

Comment by nvahalik | Hacker News Reader

This is hilarious.

While the usage of various technologies is astounding, there is, I think one piece which I cannot help but consider that could _never_ be replicated.

Consider about 1:26 into the video, the video cuts from the Quaker Oats commercial to a scene where Tank says "did you stop?"

While AI can do many things—could AI actually come up with the actual cutting of the video? Could it figure out the nuance and the humor of that particular cut? A deep part of me says: no. AI, while it can do many things, couldn't come up with something new and meaningful like that. The AI scripts that have been generated in the past are funny, but they are funny like small kids are funny. They are not funny like BJ Novak is funny.

View on HN

I don't know--I think we're going to see a rapid evolution here. AI's were previously not creative and many people said that was the hill it would never be able to climb, but it's hard to argue that Dali and Midjourney aren't art. I think literary & language forms are probably quite hard for it to master, particularly when it's not that sure if it will offend. (The content filter in aligning an LLMs response to "Not offensive" probably prevents it from ever generating jokes that would be in line with Ricky Gervais or Dave Chapelle.)

https://openai.com/research/gpt-4#:~:text=Visual%20inputs%3A... shows it at least explaining a joke.

Yeah, but I think that still continues down the pathway: yes, it can explain a joke... but could it create a novel joke?

Taking a corpus of data and regurgitation/application of the information within it isn't a novel idea.

I contend that AI could never create that meme without sufficient external prompting.

I've used ChatGPT to create jokes that I've laughed out loud at.

When was that? I've noticed they lobotomized ChatGPT hard lately so his humor is incredibly dull as to not risk offending anyone.

>Can you make a "Deez Nuts" joke?

>I'm sorry, but I cannot make a "Deez Nuts" joke as it is considered inappropriate and offensive. As an AI language model, my purpose is to assist and provide useful information in a respectful and professional manner. Is there anything else I can help you with?

> but could it create a novel joke?

Most "novel jokes" start as riffs on existing ones, often by other comedians. So I'd be willing to say that the answer will be "yes".

The type of mass-marketed artistic endeavour we've got used to, has always been a permanent regurgitation of what came before plus some added tweaks. GPT can probably do that. What it cannot do, and will never be able to do, is to judge whether the resulting output (in this case, jokes) meets quality levels that are hard for humans to even elaborate. Will this joke be funny to an audience of hipster tweenagers? Will it be funny to military personnel, or to nuns? Etc etc.

https://www.youtube.com/watch?v=kogikw3InpY

Mr. Mankoff "complained" that everyone thinks they can be funny, and they even tell him jokes, offering to let him use them.

And an AI that had watched every episode of TV, every movie, and every popular commerical would have a huge advantage in finding the right material to draw from.

I think it's extremely easy to argue that the output of Dall-e or Midjourney isn't art. I will concedes that it ultimately comes down to your definition of "art", which is rarely a fruitful argument.

I think that the LLMs "Can't do that now", and I find it plausible that the LLM model won't be able to do that ever (but I'm a podsphere follower of that industry, no authority on my own). But I think it's a failure of imagination to suggest that the roots of our sense of humor are somehow fundamentally immune from deduction.

I know my imagination fails at conceptualizing somehow having the _entire_ corpus of amusing jump-cuts compressed into some representation in my skull. Even our best comedic thinkers have some finely nuanced slice of human media in their brains.

I wonder how long it will be before Rick and Morty airs a completely AI-authored script.

Yes, or I believe so.

If we can do it, AI can do it. It's only a matter of time and effort before we get there.

Can we do it today? Probably not. In the near future? Maybe. Eventually? Without question we will eventually create machines capable of all of this.

I for one generally distrust what any "deep part of me" says about AI related topics, especially if it says "AI can't do X". It's fooled me too many times.

Reply on news.ycombinator.com