Post
What is the incremental environmental damage created by running an existing LLM locally on your own laptop?
As to "90% bullshit" - as I wrote, the false positive rate for punctuation errors and typos from Ollama/Llama2 is about 50%, which is substantially better than, say, Google Docs' grammar checker.
@tante Is refusing buy a Tesla or use X because I disapprove of Musks racism and suport of Trump, "neoliberal purity culture" ? Maybe putting solar on my house is "neoliberal purity culture" too, as is refusing to cross a picket line. Seems like Doctorow uses "purity culture" to mean what I would call "moral responsibility". As for "neoliberal", insofar as it retains any meaning, it's much more applicable to the kind of free market libertarian argument that Doctorow is making than otherwise.
@tante
> Again, it twists the argument in the way that the AI corporations like to do it as well: Search engines scour the web so AI companies should be allowed the same. It’s the same technology!
Just to add to your argument, there is another key difference between scraping for searching and scraping for LLMs.
Scraping for searching does have the concept of consent (if imperfect): it's opt-out, you can always say that you don't consent to your website (wholly or in parts) be scraped (by specific search engines or all of them), in your `robots.txt` controlled by you, and your decision will be respected. And this works also for revoking previous implicitly (or even explicitly) given consent.
As far as I'm aware no LLM honors `robots.txt`, and even if they would, they still don't and cannot offer any mechanism for consent revocation.
@tante can you refer me to the blogpost you are quoting there? I would really like to understand what the context is.
@tante I mean the post by doctorow that includes "neoliberal purity culture" ... I like to do my due dilligence (sourxe work) when forming an opinion and I couldn't find it by ctrl+f "neoliberal" in a couple of recent posts. So where do I find that?
@tante thank you! ... weird how I didn't find it earlier.
@tante I don't think anyone should have to resort to ad hominem when there's so many of his actual words to point to over the past decades.
@tante "neoliberal purity culture" ? Menuda tontería!
@tante that doesn't sound like Cory "AI is a bubble and a scam" Doctrow. Can you provide link to where that's published? I couldn't find any kind of reference online.
@tante
I disagree with Corey. Now I'm going to get on with the rest of my day.
Not that I care about Doctorow but do you use only ethical technology? I think it's fair not wanting to use ChatGPT but calling LLMs or worse "AI" (a whole computer science field) the devil is plainly silly. And it's not like it's going anywhere
@soy nobody can use "only ethically" tech (or products in general). It's impossible these days. But I reduce harm as much as possible
@tante computers make copying essentially free. LLMs are just one instance of that. We should adapt to copying being free, because after LLMs there will be inventions that make copying even easier. Strongly reducing copyright protection (and other forms of IP) is part of that. As is enforcing anti-monopoly law
@tante Gawd, you don’t even need an LLM to do that. There’s Antidote and Grammarly, just to name two…
@tante I've struggled for a while to reconcile people's justified outrage at the AI bubble's incalculably harmful unsustainability with the need to deny that even the most banal utility of its outputs could ever possibly have merit.
Credit to @pluralistic for really articulating why I find the "Not Even Once" crowd to be largely missing the point, even if we all agree that the "AI True Believers" are ghoulish monsters.
Install bonfire.cafe
Get the full app experience
