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 “Technology doesn’t need liberation, people do.“ It seems so crystal clear to me.
That doesn't seem to be the best idea @pluralistic
AI and LLM output is 90% bullshit, and most people don't have the time nor the patience to work out which 10% might actually be useful.
That's completely ignoring the environmental and human impacts of the AI bubble.
Try buying DDR memory, a GPU or an SSD / HDD at the moment.
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 There must be something in the air; I think this is something that has been on many folks' minds for a while... not criticisms of Cory, but rather the general cognitive difficulty of reconciling survival in a world far less under our control than we thought it was when our character and values accreted and congealed.
It's a nuanced phenomenon especially for a limited-attention-span world. It's good to see people taking up the challenge. Thanks for this.
@tante Dunno where you got the idea that I have a "libertarian" background. I was raised by Trotskyists, am a member of the DSA, am advising and have endorsed Avi Lewis, and joined the UK Greens to back Polanski.
@pluralistic @tante My impression was, Tante meant this specific argument and the way it is structured, and the way it functions. I hold the both of you in high esteem, and I don't have the impression that he'd somehow characterize anything beyond that argument he discusses.
> Cory shows his libertarian leanings here...
> Many people criticizing LLMs come from a somewhat leftist (in contrast to Cory’s libertarian) background.
This falls into the "you are entitled to your own opinions, but not your own facts" territory.
@tante Using LLM's for various tasks may be convenient, but it may also put you at a disadvantage because you lose the ability to do those things by yourself.
@tante@tldr.nettime.org That's not the only thing where the actions and words of Doctorow do not match.
@tante I, for one, enjoyed the write-up.
As you, I do plenty of things that I either cannot morally or ethically defend or simply am ignorant of the harms of.
For other technology, I am either pressured to use it (by government or similar institutions) or I can at least use it 'for good' (facilitating some other good - in current case, student administration) and also make a living.
For LLMs, I fail to see the redeeming qualities that make the compromise 'worth it'. It's all nett negatives.
@tante My point is that LLMs 'try' (and very often fail) to solve the wrong problem.
Writing all the boilerplate -> should be solved by better frameworks.
Spell check -> I think this is already invented.
Summarize long texts -> Executive summaries.
Produce (verbose) text -> Writing in my view is as much thinking as it is writing - skipping the thinking part is counter-productive.
@tofticles @tante @onepict it's a bitter pill to be forced to use a spell checker that until very recently didn't even know how many rs are in "strawberry " (or ls in "blackball").
@perigee @tofticles I think @tante captured the difficulties existing and surviving in this world.
I'm still disappointed with Doctorow, especially since he felt the need to get his defence in first.
His choice of wording there was saddening to me. He outsourced his feelings of guilt and judgement on the rest of us.
Whenever I hear or read someone using the term purity anything I feel a mild disappointment. Because I do expect better of our leaders who take the public stage.
@tante At Tuesday night's #Faradayprize presentation from Mike Woodbridge, more than one of the slides about #LLMs brought #Trump to mind
@tante One thing I'd love to hear more about from you, and @pluralistic, @simon and others is which of these models are doing the best, most interesting things to mitigate harms? Some are being trained with thought about & limits on the inputs, and the public interest in mind in general. To at least some degree. (e.g., Apertus)
Who's reviewing, comparing models with that lens? If someone understands most companies are part of the problem and asks for alternatives, what do you offer them?
@tante I still haven't completely unpacked these arguments. To dwell on the gramnar checker thing, I assume that pre-LLM checkers were to some extent developed by building statistical models from a large corpus of existing text. That's not quite the same thing as the mass plagiarism used to build generative AI models. For myself I've never used such tools, I consider them an annoyance: If there's a mistake in my writing, the human reader will make a better job of correcting it from context.
@tante (Not meant as nitpicking, but it's the very first stentence and I stumble over it. Should be "Life is complicated".)
@tante It's also annoying tone policing: Sure, you can protest against it, but not like *that*.
Install bonfire.cafe
Get the full app experience
