yes, comforting, they will keep one or two people on to form prompts and validate ai outputs
at least intitially, it'll soon become unnecessary
I don't know what blend of narcissism and misunderstanding of the history of free market (and totalitarian) capitalism that makes people think they won't or can't be replaced. AI already is smarter and massively more capable than you, and is advancing frightening rapidly as it starts to be employed to support it's own evolution. People seem blinded to the massive and obvious downside to their own interests, like home weavers cheering on Arkwright.
To put it another way, imagine it wasn't a computer, but instead we had the capability of building organic beings that were much smarter and more capable than humans that would multiply and spread around the globe. Why the fuck we would want to do that? Wouldn't it just show us up for the flawed modest people that we actually are, and push most of us into obsolescence and subservience by comparison. It would be madness, but everyone seems enthusiastic about AI. It's depressing
(
cumquat may will not be commenting further on the allegations, Thu 30 Jan 2025, 6:32,
Reply)
"AI already is smarter and massively more capable than you"
Actually no it's not. I regularly ask it for help with stuff I'm not sure about.
80% of the time it invents APIs that don't exist. If it doesn't have an answer it makes shit up.
It has got it completely right maybe 5 times total. If it is better than you it's cos you are doing trivial stuff
It's shit and you can't rely on it and it's probably about as good as it will ever get until something new comes along rather than LLM
(
simbosan is too tired, Thu 30 Jan 2025, 8:44,
Reply)
seems like a narrow measure
but yeah, AI's skills are uneven across the whole gamut of endeavor
but to take the field of chemistry, for example, in 18 months ChatGPT has gone from writing sub-GSCE level papers to university papers that are impossible to detect without assistance from AI in recent iterations.
The OpenAI 'o1' ChatGPT released the preview for last month performed 7 times better in programming than the previous version. They ran o1 in the Codeforces' competitive programming contests, I don't know what the fuck the format is, but we can assuming these are top programmers, it scored in the 89th percentile fo contestants
So if it's writing shit APIs now, wait until the next release
Anyway, I know I'm just "man shakes fist at sky", but it's such a monumental folly it's sad that's there's not more democratic opposition to it, just a bunch of people reassuring themselves it can't happen to their job without really grasping the pace and scope of AI advancement
(
cumquat may will not be commenting further on the allegations, Thu 30 Jan 2025, 11:52,
Reply)
Worth reading Ed Zitron's stuff on OpenAI
www.wheresyoured.at/oai-business/
There are some major stumbling blocks for it, not least the hallucinations that seem unsolvable, and presumably will get worse considering that it's polluted the internet and is now feeding on its own shit.
Still going to cause a lot of shit and job loss. Just worth being wary that the LLM stories are being fed by hype-bros wanting more investment money to a media that aren't questioning their claims enough.
I feel sorry for people wanting to use AI for things other than content creation, as it does seem to have potential in things like medicine, but the industry is going to be associated with generative slop and cons because of these "entrepreneurs".
(
monkeon schmonkeon, Thu 30 Jan 2025, 12:35,
Reply)
A good article about the finances and early losses, but Trump announced they're raising 100 billion in investment
despite the shaky finances and hype merchants and bullshitters that new tech always attracts, and whatever deficiencies it has in specific areas, the money won't stop and the advancements won't stop, because the carrot is so massive and valuable, and that carrot is not to be some helper service like clippy, that's just the transition, it's to replace waged employment with ai, and businesses will absolutely pay for that.
(
cumquat may will not be commenting further on the allegations, Thu 30 Jan 2025, 23:01,
Reply)
I'm sure they'd also have liked
a $25,000 self-driving car by 2023, or to hold all their meetings with a legged avatar in the Metaverse.
Why do you think the advancement won't stop? Where are they going to get the new data from? How will they fix the hallucinations?
Anyone going all-in on it for anything business critical is risking a Post Office-esque scandal. On a smaller scale, you're already seeing it with cases like this.
(
monkeon schmonkeon, Fri 31 Jan 2025, 10:05,
Reply)
It's a prediction engine that is good to use in some contexts
Pretty good at summarising notes (release notes etc) or scaffolding some template code for example, but people and tech-bros are treating it like it's some omnipresent god. My favourite was when it told someone to kill themselves.
(
Fadgebadger Sweaty sumo wanks for everyone!, Fri 31 Jan 2025, 10:41,
Reply)
it doesn't have to be a god to do most of human work
we're kind of prediction engines ourselves, and have flaws and error rates.
(
cumquat may will not be commenting further on the allegations, Fri 31 Jan 2025, 11:31,
Reply)
even if you're right I see it as more like the dot com boom, when there was stupid money being thrown at this shiny new thing
But 25 years later amazon is the biggest retailer in the world, and it's all online. It never went away becuase the underlying utility of it was solid, it just needed refinement. And now it's fucked countless retail business over the years. AI will do this for jobs. It might not be linear, there might be frontrunners that fall at the hurdles. But it will happen and it will fuck everyone but the few.
(
cumquat may will not be commenting further on the allegations, Fri 31 Jan 2025, 11:39,
Reply)
Amazon always worked, though.
It used to be the go-to example of good usability before they fucked it with dark patterns and shoddy marketplace items. It always made business sense. They were just biding their time for bandwidth costs to come down and internet users to go up for it to become profitable. Same with Google and Facebook. Then they can enshittify.
Not all technical hurdles are solvable and unless they are, OpenAI provides at-best buggy business software and an alternative to Google that needs you to fact check if you don't want to get stung. Lots of people think technical stuff is easy if you can easily describe the frontend, but it's the detail that's usually the hardest part. ChatGPT still feels like a prototype.
Sam Altman keeps claiming to be approaching AGI without any evidence that it's even possible. No one seems to ever challenge him or ask for a live demonstration. He comes across as a shyster promising to turn lead to gold if you give him enough money for chips.
(
monkeon schmonkeon, Fri 31 Jan 2025, 12:49,
Reply)
It won't get noticeably better because the underlying idea is very limited
Just saw this and yeah, i believe this is true. AI writes shit code. And coding exercises/tests are all very formulaic, perfect for a context free network
bsky.app/profile/mattround.com/post/3lgxbbagw5k2r
(
simbosan is too tired, Thu 30 Jan 2025, 12:45,
Reply)
I reckon in 10 years time you'll wonder how you could have ever thought this
but I guess we'll both have to wait and see how it plays out.
I've been listening to interviews with recently employed devs who worked for chatgpt and other ai mobs talk about their experiences with it. They say they're losing oversight over how to comprehend how the models are advancing, as processing capacity and data is added it's developing capabilities in a whole raft of areas they don't have the metrics yet to measure. And a surprising number of people with direct involvement with the project have come to the view it should be stopped.
(
cumquat may will not be commenting further on the allegations, Fri 31 Jan 2025, 0:04,
Reply)