Comment on Linus Torvalds used AI to fix a Linux bug, and now LLM critics on Linux face an uncomfortable choice
Ooops@feddit.org 2 days ago
To rephrase that title in another context: “I hammered in a nail with my new hammer, and now I face the uncomforatble choice if I should use it on all my screws, too”.
No, ffs… this story -and the one hundred copies in the last week- is bullshit. LLMs are a tool. There are things they can do, like helping to automate some debugging tasks with precise instructions. There are things they can’t do, like producing maintainable vibe -coded code.
There is no uncomfortable discussion. Use a tool where it’s useful, don’t where it isn’t.
The whole “discussion” is fake, started by that insane faction treating AI like their new savior and trying to push the idea that everyone not using AI for whatever task is doing so out of ethical, moral or ideological reasons (and it’s pushed by the creators of said AI because only a total and absolutely insane movement to do everything with AI now will ever justify their investments - and even then it’s still questionable…). No, that’s just how those idiots make decisions.
Using AI for one task (it can do) does not mean someone just converted to your branch of insanity and will use it for everything. Using an LLM to assist in debugging your code and banning all vibe-coded shit from your code base is no contradiction, no matter how hard you try to push that fairy tale.
Zephorah@discuss.online 2 days ago
Here’s an example of how not to use it. Medicine. The Internet soup includes medical advice from new agers, mommy blogs, anti-vax, chiropracty, chakras, well meaning but mistaken posters, Reddit, aggregate scrape sites, legit sites, and more.
If you take the time to correct the AI it will apologize and then move to other attempts until it nails it, maybe. Alternatively, you tell it the right answer and it confirmation biases you. (But are you actually right? The AI will never tell.)
The medical professional community is getting hit hard lately with aggressive waving of phones, yelling, and unwillingness to listen, because the Google bot said so and the medical staff doesn’t agree with the Google bot. Or, the chatGPT subscription said so. The ChatGPT guys go hard on defending the AI.
There used to be a bit of embarrassment and willingness to listen to credentialed individuals giving medical advice in contradiction to WebMD or random posts. Not always, but often.
Lately, people are defending the AI answers like it’s their own life’s work.
It’s starting to resonate, at times, like people caught in an addiction but not ready to get clean: defensive, they know what they’re about, you don’t know, and they’ve got this all under control. That’s not quite right, but it’s the closest equivalent in my experience.
Computer things though, I use it sometimes, depending.
whoisearth@lemmy.ca 18 hours ago
Disagree. I’ve heard many doctors on the news talk about the specific agentic AI they’re using and how, as an example, it’s better at finding cancer sooner, or diagnosing diseases in general.
Example -
cbc.ca/…/artificial-intelligence-emergency-rooms-…
Zephorah@discuss.online 7 hours ago
They are testing quite a few things atm. Radiology image recognition is definitely there. But is this what the phone wavers are accessing? No. The Google search bot and the bots being tested are two different things.
I wouldn’t get too excited about the future of AI in medicine, not given the level of venture capital and such in medicine already.
LLMs are programs that can be tuned by people who have not sworn an oath or spent 12 dedicated years and $250k debt to pursue medicine. What will motivate how these bots are tuned? Publicity and PR can create excitement, but what have the last 30years of medicine taught us? Even the radiology bot can be “turned down” in ways a human rarely would allow.
Profit speaks. Profit drives. Profit decides patient care levels and outcomes. What keeps it afloat are the people clawing back at administrators. Don’t underestimate the power of angry nurses or pushy physicians in just how much they’ve kept healthcare afloat. For you. While admin battles them from the shadows in the name of profit.
What they don’t mention among the many bots being tested is the EOLI, or, End Of Life Index. The most used electronic medical record had a bot introduced this year, unannounced, that is a probability predictor of a patient dying in the next 6 month. The mouse over admonishes caregivers not to discuss this with patients or families as it may upset them. Does/will the EOLI factor into insurance decision making?