- cross-posted to:
- nyt_gift_articles@sopuli.xyz
- cross-posted to:
- nyt_gift_articles@sopuli.xyz
A radio station in Poland fired its on-air talent and brought in A.I.-generated presenters. An outcry over a purported chat with a Nobel laureate quickly ended that experiment.
AI is a lie.
It’s not a lie, it’s just not what most people think it is. There’s a lot of ignorance and a lot of lies about AI.
Was it Google who claimed 25% of their production code last year was written by AI? Microsoft? Anyway, I’m going to call bullshit on that right now. Or they require a curious amount of bullshit code to run their business.
To be fair to all those people that misunderstand it, they are marketing it as Artificial Intelligence, which it isn’t. So one could argue it is in fact a lie, as most marketing seems to be these days. It’s difficult for us humans to see the difference between intelligence and an “alright prediction of what might come next”. Such as when we struggle to tell the difference between the truth and a lie someone told us. It can be deceiving.
Since marketers have bastardized the term, and we’ve begun using AGI in place of the old meaning, confusion is only going to get worse until existing LLMs become somewhat boring, and marketing latches onto some other trend.
With that said, I find the utility of this thing we now call AI to be pretty useful for my own needs, but that’s not stopping people from trying to fit this square shaped solution into circle shaped holes.
This exactly. AI is great for certain things. What everyone tries to use it for, like the bullshit in this article, is not great.
“25% of our code was copied off of GitHub”
Well, their products have been going downhill so I suspect that 25% was AI genersted and then their QA is less effective because there are no devs doing unit testing.
In fairness, about 50% of my code by lines is written by AI these days, and I don’t have it linked into my code base. That claim isn’t ridiculous
Now, of that 50% is 88% long repetitive crap that I could easily write but find mentally draining, the other 10% is something simple that I would normally copy paste from elsewhere because I forgot the exact syntax (and don’t exactly remember where I used it last) and me giving it a shot with things I don’t want to do, like restyling a page. The last 2% is me giving it a shot with business logic for shits and giggles, occasionally I’ll try to coach it through the solution but usually I just grab bits and pieces and rewrite it myself
Granted, this is the easiest and most simple and repetitive code, but it’s still a godsend. Now can AI write the other 50%? With a proper setup where it ingests the code base into a vector store it might get up to 75%, if I was willing to coach it through my tasks carefully (taking more time than the task would take me) I could probably get it up to 85% or 90%, but that last 10%? It just can’t, it’s not even close
It’s not taking my job without a paradigm shifting breakthrough or two on the scale of “all you need is attention”. Even then, it only works if you write your prompts like code… If you don’t understand how to use it and understand the code well enough to communicate the goal explicitly and unambiguously, you’re not going to be able to drive it where you want it to go
To put it another way, you can build 90% of the system in 10% of the time it takes to complete the last 10%
I have access to AI integrated with my IDE. It mostly guesses at the line I’m going to write. It probably gets it right 50% of the time.
It also very, very often suggests stuff that works but isn’t very good. Like it offered some convoluted suggested for adding audit fields to Firebase. Ultimately it did suggest the solution I went with, but only after starting down the road of stupid ideas.
Like, if your code base is pretty good and you just need to tweaks stuff that is already good enough that’s one thing. I frequently look at the code base and wonder if it was implemented by someone who really knows Java at all.
I suppose it might be fair to assume a huge technology company would have their shit together, but technically I work for a huge tech company… just not the same core business. Tech enough that we have a whole mess of internal AI tooling to create AIs for specific things.
We can create an AI agent, but we can’t follow simple fucking rest standards.
Anyway it’s hard to quantify, but I get less mileage out of integrated AI tools than I do bouncing ideas off ChatGPT.
I think that’s fair.
I don’t have AI integration in my ide, mostly by choice -if I pushed for it I could make it happen, but I just don’t think that’s a good idea at this point
AI can be a crutch . One that limits you to the level of a baby developer. If you can’t effortlessly understand what it gives you, frankly you shouldn’t be using it.
Bounce ideas of chat gpt. It sounds like you’ve got the right idea - your reaction sounds correct to me, you should never ever trust it… You must only use it, and that’s the tone I get from your post.
It is a tool, you are a programmer. You exploit tools, you do not trust any tool. You are the one who turns ideas into actions, never forget that and you can use this new tool anywhere it makes your life easier
What are you even talking about? The story is about the use of AI to recreate a dead person and create a fake interview.
It’s not AI that’s lying. The AI is just doing what AI does.
So, lie?
You are anthropomorphizing it. It can give truth or falsehood the same as pages of a book or a funhouse mirror.
If you ask me today, “what is the meaning of life?” I might give you an answer. And if you ask me tomorrow, I might give a different one. You have no way of knowing whether I’m correct today, tomorrow, or ever. But if one of those answers, right or wrong, helps you find meaning, it’s still useful. (As a rhetorical point. I’m definitely the last person anyone should look to to find meaning.)
AI is a lot like that. You give it input, it gives you output, and whether you get anything of value depends greatly on what you are looking for.
I’ve gotten some advice on improving some of my writing. And some of the advice I took, some I ignored, and some I modified before using. I think the writing turned out better, and since I largely write for myself I’m pretty happy with that.
I’ve asked it for help programming, and at times it was helpful and other times cost me hours circling around the same old wrong answers, but there’s every chance I would’ve struggled just as much looking online.
The other day my daughter was making a slushie and it was turning out really wet and gross, so I explained to an AI what we’d done and asked if it had any idea why it didn’t turn out. And it turns out, we were using zero sugar soda which doesn’t work—the sugar is necessary. So we added some simple syrup and it turned out perfectly.
And it was much faster and easier than Google. But if the advice had been wrong, nothing of value would’ve been lost.
You could say it’s a L.A.I.
Technically a checkers program is also AI if you check the dictionary definition. Many will say it’s not “real AI” as it doesn’t have general intelligence but it’s still AI. Snake oil salesmen love that.
Selling AI as a solution for everything is absolutely a lie. Modern snake oil in nearly all the public facing uses.
There are a lot of good uses for AI in science and entertainment as a way to enhance work being done by humans. Pattern recognition for lidsr in finding kist civilizations. Finding exoplanets. Finding new compoubds for medicine and manufacturing.
Hell, there is even some usage for some personal things and LLMs are great at addressing certain people’s hesitation when interacting with computers. But like the prior examples, it can’t solve anything on its own. It must be controlled and thigs like search engines and processing job applications are terrible uses for AI when humans are having trouble doing that thing already.
AI only works for technical work that requires precision when it is doing something humans were already good at. It can do a lot of those things faster, but has to be babysat if the output matters.