"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
Source (Bluesky)
I work at a company that uses AI to detect repirstory ilnesses in xrays and MRI scans weeks or mobths before a human doctor could.
This work has already saved thousands of peoples lives.
But good to know you anti-AI people have your 1 dimensional, 0 nuance take on the subject and are now doing moral purity tests on it and dick measuring to see who has the loudest, most extreme hatred for AI.
Nobody has a problem with this, it's generative AI that's demonic
Generative AI uses the same technology. It learns when trained on a large data set.
Generative AI is a meaningless buzzword for the same underlying technology, as I kinda ranted on below.
Corporate enshittification is what's demonic. When you say fuck AI, you should really mean "fuck Sam Altman"
I mean, not really? Maybe they're both deep learning neural architectures, but one has been trained on an entire internetful of stolen creative content and the other has been trained on ethically sourced medical data. That's a pretty significant difference.
No, really. Deep learning and transformers etc. was discoveries that allowed for all of the above, just because corporate vc shitheads drag their musty balls in the latest boom abusing the piss out of it and making it uncool, does not mean the technology is a useless scam
This.
I recently attended a congress about technology applied on healthcare.
There were works that improved diagnosis and interventions with AI, generative mainly used for synthetic data for training.
However there were also other works that left a bad aftertaste in my mouth, like replacing human interaction between the patient and a specialist with a chatbot in charge of explaining the procedure and answering questions to the patient. Some saw privacy laws as a hindrance and wanted to use any kind of private data.
Both GenAI, one that improves lives and other that improves profits.
I think DLSS/FSR/XeSS is a good example of something that is clearly ethical and also clearly generative AI. Can't really think of many others lol
All this is being stoked by OpenAI, Anthropic and such.
They want the issue to be polarized and remove any nuance, so it’s simple: use their corporate APIs, or not. Anything else is ”dangerous.”
For what they’re really scared of is awareness of locally runnable, ethical, and independent task specific tools like yours. That doesn’t make them any money. Stirring up “fuck AI” does, because that’s a battle they know they can win.
First of all, intellectual property rights do not protect the author. I'm the author of a few papers and a book and I do not have intellectual property rights on any of these - like most of the authors I had to give them to the publishing house.
Secondly, your personal carbon footprint is bullshit.
Thirdly, everyone in the picture is an asshole.
I'm just sick of all this because we gave to "AI" too much meaning.
I don't like Generative AI tools like LLMs, image generators, voice, video etc because i see no interests in that, I think they give bad habits, and they are not understood well by their users.
Yesterday again i had to correct my mother because she told me some fun fact she had learnt by chatGPT, (that was wrong), and she refused to listen to me because "ChatGPT do plenty of researches on the net so it should know better than you".
About the thing that "it will replace artists and destroy art industry", I don't believe in that, (even if i made the choice to never use it), because it will forever be a tool. It's practical if you want a cartoony monkey image for your article (you meanie stupid journalist) but you can't say "make me a piece of art" and then put it on a museum.
Making art myself, i hate Gen AI slop from the deep of my heart but i'm obligated to admit that. (Let's not forget how it trains on copirighted media, use shitton of energy, and give no credits)
AI in others fields, like medecine, automatic subtitles, engineering, is fine for me. It won't give bad habits, it is well understood by its users, and it is truly benefical, as in being more efficient to save lifes than humans, or simply being helpful to disabled people.
TL,DR AI in general is a tool. Gen AI is bad as a powerful tool for everyone's use like it is bad to give to everyone an helicopter (even if it improves mobility). AI is nonetheless a very nice tool that can save lifes and help disabled peoples IF used and understood correctly and fairly.
AI in others fields, like medecine, automatic subtitles, engineering, is fine for me. It won't give bad habits, it is well understood by its users, and it is truly benefical, as in being more efficient to save lifes than humans, or simply being helpful to disabled people.
I think the generative AI tech bros have deliberately contributed to a lot of confusion by calling all machine learning algorithms "AI".
I mean, you have some software which both works and is socially beneficial, like translation and speech recognition software.
You have some software that works, and is incredibly dangerous because it works, like facial recognition and all the horrible ways authoritarian governments can exploit it.
And then you have some software that "works" to produce socially detrimental bullshit, like generative AI.
All three of these categories use machine learning algorithms, trained on data sets to recognize and produce patterns. But they aren't the same in any other meaningful sense. Calling them all "AI" does nothing but confuse the issue.
I sure am glad that we learned our lesson from the marketing campaigns in the 90's that pushed consumers to recycle their plastic single-use products to deflect attention away from the harm caused by their ubiquitous use in manufacturing.
Fuck those AI users for screwing over small creators and burning down the planet though. I see no problem with this framing.
AI is a marketing term. Big Tech stole ALL data. All of it. The brazen piracy is a sign they feel untouchable. We should touch them.
I would not to get close to bike repaired by someone who is using ai to do it. Like what the fuck xd I am not surprised he is unable to make code work then xddd
They only real exception I can think of would be to train an AI ENTIRELY on your own personally created material. No sources from other people AT ALL. Used purely for personal use, not used or available for use by the public.
So I'll be honest. I use GPT to write Python scripts for my research. I'm not a coder and I don't want to be one, but I do need to model data sometimes and I find it incredibly useful that I can tell it something in English and it can write modeling scripts in Python. It's also a great way to learn some coding basics. So please tell me why this is bad and what I should do instead.
Didn't you read the post? You're bad and should feel bad.
I'd say the main ethical concern at this time, regardless of harmless use cases, is the abysmal environmental impact necessary to power centralized, commercial AI models. Refer to situations like the one in Texas. A person's use of models like ChatGPT, however small, contributes to the demand for this architecture that requires incomprehensible amounts of water, while much of the world does not have enough. In classic fashion, the U.S. government is years behind on accepting what's wrong, allowing these companies to ruin communities behind a veil of hyped-up marketing about "innovation" and beating China at another dick-measuring contest.
The other concern is that ChatGPT's ability to write your Python code for data modeling is built on the hard work of programmers who will not see a cent for their contribution to the model's training. As the adage goes, "AI allows wealth to access talent, while preventing talent from accessing wealth." But since a ridiculous amount of data goes into these models, it's an amorphous ethical issue that's understandably difficult for us to contend with, because our brains struggle to comprehend so many levels of abstraction. How harmed is each individual programmer or artist? That approach ends up being meaningless, so you have to regard it more as a class-action lawsuit, where tens of thousands have been deprived as a whole.
By my measure, this AI bubble will collapse like a dying star in the next year, because the companies have no path to profitability. I hope that shifts AI development away from these environmentally-destructive practices, and eventually we'll see legislation requiring model training to be ethically sourced (Adobe is already getting ahead of the curve on this).
As for what you can do instead, people have been running local Deepseek R1 models since earlier this year, so you could follow a guide to set one up.
Do people who self-host count? Like ollama? It's not like my PC is going to drain a lake.
Ethics and morality aside.
Yes, they count, the process of making and continuing to update the underlying LLM is also what drains the lakes, they are all made on pirated info (all the big ones for sure, I've not heard of a widely available, usable model trained 100% on legally obtained data, but I suppose it could exist).
To that person, yeah self hosting still counts.