Unpredictable abilities emerging from large AI models
quantamagazine.org
Unpredictable abilities emerging from large AI models
1–10 of 326 posts
Re: Unpredictable abilities emerging from large AI models
#2I used ChatGPT to generate erotic stories. Now I want a model which can produce porn videos from prompts.
Re: Unpredictable abilities emerging from large AI models
#3I used ChatGPT to generate erotic stories. Now I want a model which can produce porn videos from prompts.
I am excited to live inside my anime Haram hentai
Re: Unpredictable abilities emerging from large AI models
#4>“That language models can do these sort of things was never discussed in any literature that I’m aware of,"
I had previously the expectation that unpredictable emergent behavior would exist in any sufficiently complex system? Based on layman's readings in chaos and complexity theory.
Re: Unpredictable abilities emerging from large AI models
#5I used ChatGPT to generate erotic stories. Now I want a model which can produce porn videos from prompts.
Give it 2 years, it will happen, and it will make billions.
In 5 years, unlimited interactive NSFW video games that will be personalized and remember you.
Re: Unpredictable abilities emerging from large AI models
#6I used ChatGPT to generate erotic stories. Now I want a model which can produce porn videos from prompts.
I am excited to live inside my anime Haram hentai
We all are. Some want porn of Emma Watson, some want porn of werewolves and some just want to see porn of their ex. And soon we are all gonna get what we want.
That‘s the silver lining to the massive job destruction that is to come.
Re: Unpredictable abilities emerging from large AI models
#7From the review comments...
> Key Weakness: The paper largely focuses on showing how much emergence occurs in a “sudden” manner, bringing reports from previous work. It relies on the “magic” of emergence, rather than providing new insights on why this is happening and when it happens/does not happen.
> Requested change: More fundamental evidence on the claim "further scaling will likely endow even-larger language models with new emergent abilities" with more concrete discussion (with possibly evidence) on how those new emergent abilities would look like and how further scaling will be possibly in a approachable way.
Re: Unpredictable abilities emerging from large AI models
#8Basic statistical relations in observable world (on Human-Level scale) should somehow be reflected in human language and in mathematics. Deeper models tend to learn those relations, and produce good looking connections which fit in the realm of possible "thoughts" perfectly.
Re: Unpredictable abilities emerging from large AI models
#9In the past few days I've been shellshocked by GPT-4. I'm still and believe I will be for the near future.
Re: Unpredictable abilities emerging from large AI models
#10This caught my attention as I found it implausible:
> One DeepMind engineer even reported being able to convince ChatGPT that it was a Linux terminal and getting it to run some simple mathematical code to compute the first 10 prime numbers. Remarkably, it could finish the task faster than the same code running on a real Linux machine.
Following the link, there's a screenshot to a screenshot [0] of a code-golf solution to finding primes which is quite inefficient, and the author notes
> I want to note here that this codegolf python implementation to find prime numbers is very inefficient. It takes 30 seconds to evaluate the command on my machine, but it only takes about 10 seconds to run the same command on ChatGPT. So, for some applications, this virtual machine is already faster than my laptop.
So it's not quite calculating primes; more likely it recognizes the code as being code to do so, and recites the numbers from memory. That's interesting in its own right, but we won't be running Python on an LLM for a performance boost any time soon. In my experience this interpreting is apparent as a limitation of the model when it keeps insisting on broken code being correct, or having its mistake pointed out, then apologizing, saying it's got some new code that fixes the issue, and proceeding to output the exact same code.
[0] https://www.engraved.blog/content/images/2022/12/image-13.pn...
