thing is there's actually good uses for it! Just one example: there are NGOs using AI to tabulate satellite data on elephant herds and this helps them ID and prevent poaching. There's another that is using similar sat data tabulation to track coral reef recession. The llm bros are cocking it all up
I have also grown convinced that our modern conception of AI is evil because, as I look back in history, I notice that a great many of the horrific events and practices of the modern world came about because of evil attempts to reduce humans to machines.
feels really weird because in a world where we hadn't let the dumbest people with wallets in the world co-opt the entire conversation I'd probably be arguing for a bunch of the actuall practical applications that do exist. But noo, we had to hand the mic to the most dead end clownwaffles out there
ChatGPT and Midjourney are shit. It that’s not what LLMs are designed for! My university is using them to do analysis on massive datasets. These are huge research projects that would take years to sort through, being analyzed in weeks. But they’re secure because they aren’t from the open web.
I’ve already seen a “this shows promise, but the data cleanup and error checking needed makes it useless now” paper in regards to using it in drug discovery. A lot of people are getting big grants and kicking the wheels of the tech right now.
Hallucination is going to be AI’s “last mile” problem.
my literal first book was about how machines if they achieve true sentience should be given equal rights and they actually turned me into their enemy, incredible achievement
I work with and research civilian drone tech for a living and I think it's really cool that we have semi-autonomous flying robots and I still absolutely hate this shit!
We had the same thing with the dot com bubble. Didn't stop the utility of the internet. LLMs are great tools with a lot of applications. They're still rough around the edges for some broader uses with extensive always on integration though.
Yes, I've heard it's great for programmers. I don't know why they gave the AI idea to the same folks that gave us NFTs and cryptocurrency, who all seem to be furious that they can't fucking draw.
It's terrible for programming in my limited experience so far. I had to turn off the annoying second-guessing "assistant" in my IDE. I was wasting time discarding its suggestions.
I do think it's very specific, and moreover unnecessary. Like it helps but doesn't revolutionize the process. I truly think C Suite bozos got bamboozled en masse and are just trying to not seem stupid.
Yeah these tools are not helping anyone to write really good code. The thing is writing good code is already so devalued a skill I'm not sure how perceptible the difference would be in a lot of fields. Which is really part of the problem.
Yeah, “by tokenizing 100,000,000 examples of code written by humans, I can offer this code as a statistically likely solution to your query” is not a sound approach to coding.
It sounds bad enough already but it can only get more ridiculous once a substantial amount of the code isn't even written by humans any more. Sucks all the joy out of the work too but I guess the people pushing for this either don't enjoy programming or have no skills to take pride in...
Yeah the impact on open-source projects allowing anyone to contribute must have been disastrous. AI code generation just emboldens people to contribute above their level of ability. More skilled programmers would be out there making better, more judicious use of the tools.
The good uses of AI were being used before crypto and NFTs. The latest LLM craze is the weird one nobody likes.
Though they can be helpful for programming, probably their best use right now.
Machine learning has definitely proven really useful for IDing critters in photos for conservation and a bunch of related purposes. Different technology than LLMs or diffusion networks that have lately been lumped in with ML under the heading of AI.
And it's even more annoying because every other use normal people see are "we put a chatbot on the website and fired 1000 people" and "tech company spies on you"
It's just a tool. Tools are good when in service of men and diabolical when men are in service of them.
And stupid when used how they weren't meant by people who don't understand them
As someone who was using clunky ol' AI techniques for image classification 20 years ago, seeing that kind of thing conflated with LLMs and "makin' plausible stuff up" engines is weird.
Even the generative use of LLMs is weird. I was using natural language processing roughly 15 years ago for sentiment analysis. LLMs are basically the next step of natural language processing, and probably would have been great for speeding up my work, but I never called what I did 'AI'.
Yeah as a long-time AI enjoyer it's rough out there these days, thanks to these grifters I feel like I'm screaming, "NOT ALL AI!!!" on a constant loop. Mind you I felt for years that machine learning was already filled with these types overstating the 'intelligence' of what they were achieving.
Pattern recognition software (which is not what "AI" is) may help with these kinds of things. But if anyone is using AI to tabulate data on herds or tracking coral reef recession, their software is probably giving them BS answers and they just don't know it yet.
Setting LLM garbage aside, I expect that some nice medical applications will be developed, say around radiology, and instead of being used to make care more reliable by augmenting human workers, it will be used to reduce professionals to machine operators with lots of human tragedy.
It can scrape Stack Overflow and make it even easier to throw together a bunch of code you don’t understand to produce an insecure application that won’t scale and get paid for it. Like it was with php but with less effort.