@nprofile1q... Bouncing on “it’s a tool” paradigm: I guess varying use also have very different costs. I don’t imagine sorting saved blog posts (which btw were immensely important for me for personal reasons, and also very difficult to parse) is anywhere in the same cost league as “generate some random slop video”.

@nprofile1q... (If anything, my post was intended as a slight slap to AI enthusiasts reacting to Linus’s post, more than a reaction to said post and for this part I could have been more precise)

@nprofile1q... my choice of word “perfectly reasonable” for the Linus message wasn’t particularly careful about details in the message, more about general stance “the tool can be useful” and if anything I have been frustrated to see AI-bros reacting to that Linus message feeling vindicated because they are extrapolating much more than what was said.

@nprofile1q... The unhinged use of it I reckon is problematic (vibe coding being problematic on so many fronts: close to IP infringement, produce bad quality software which tends to be throwaway making the resource waste much worse, alter developers skills etc). But I feel there are good possible uses that will come out of this, perhaps from more specialized models.

@nprofile1q... wait, there are several topics bundled. I was answering to your message suggesting the robot was not intelligent enough, to this specific argument I am saying it doesn’t matter.
The cost matters. But I am not sure performing a well scoped task with no alternative solution
is a meaningful portion of cost vs big use of AI to brute force a way through things.

@nprofile1q... It also feel disingenuous to bring the "statistical word prediction" when it does have use. Sure its "intelligence" is weird, but I had Claude parse hundreds of messy manually saved out-of-order html blog posts and generate a neat indexed version. I would never have bothered writing a tool to do this otherwise.

@nprofile1q... I honestly don't care that it's not genuinely intelligent, if it can provide some value to me. A few weeks ago I pointed it at tricky problems I have and it helped me bounce idea, theories, and find edge/useful cases.

@nprofile1q... I'm not a person who is particularly excited by AI, nor do I have the slightest amount of admiration/trust for big tech. But it intuitively feels off to me to be resisting it using the argument that they stole FOSS ode.

@nprofile1q... Putting aside the other (real) issues, I sense a difference between using AI to make a full project mostly inspired by existing ones vs using AI to summarize, optimize or evaluate existing code. Learning from existing FOSS is "only" doing what humans would do at a much larger scale.