"Forty-three percent had used one for work" … "The numbers are slightly different for the under-16s. Fifty-four percent said they had used a GenAI tool, with more than half (53 percent) of those saying they had used it for schoolwork."
https://www.ofcom.org.uk/siteassets/resources/documents/rese...
Your wording choice there is excellent as "transforming" can either be extremely positive, or negative, or just different. Given how many students now, judging by various outcries on social media by teaching staff are borderline illiterate in middle school, it does make one wonder what the long term effects of this will be.
More interesting would be the percentage of people that have used GenAI in the past year with a purpose.
I suspect a significant number have just played with it. Likely many didn't make it a habit, let alone ever do anything useful with it. It would be a bit of a stretch to say anyone in this group has 'adopted' AI.
Also how many would continue to use genAI if it weren't artificially cheap/free. Remember that nearly all AI companies are losing money hand over fist, there's bound to be user attrition when the walls close in and they have to start shoving in ads or aggressively upselling paid subscriptions.
If we keep testing young people in AI-friendly tasks, they will keep using AI to a larger rate than the rest of us.
I will confess to enjoying bad AI _image_ generation, mind you. Look at the images on this website! https://theonlineaccountants.uk/file-company-accounts.htm
AI generated text generally just makes me cringe, tho.
As a lifelong computer nerd I've always dreamed of having a computer friend who knows all about my life and who I can bounce ideas off, it's a core sci-fi trope. But I tried these new LLMs everyone was so hype about and... nah, they all feel like I'm wasting my time. Anything I want to actually know I can look up myself faster than negotiating with these things to try guess what they want me to tell them so they will actually do what I wanted. And as far as just being a nice personality to spend my time with goes, they all come across to me as awkward or pretentious or fussy, and not what I want my computer friend to sound like. If I need to spend 20 minutes trying to "train" my computer friend how to sound vaguely like a real person, it's a fail, sorry. It feels like trying to socialize on Reddit or Discord, you waste a bunch of time going back and forth with strangers who are trying to be funny or clever but never actually say anything meaningful. Which is why I don't use those sites either.
But then I have real life friends who have mentioned they talk to these chatbots all the time, almost like a therapist, and they ask them questions about their lives. Obviously they don't take the advice as seriously as they might coming from a professional, but they still consider it a valuable interaction. To them, having this kind of stilted interaction with a hokey-sounding AI is worth something. And I don't get what they are seeing that I am missing. Perhaps my standards are too high? Or perhaps they just have a greater tolerance for wasting time?
Anyway, when surveys ask me if I've used AI, I always say yes. The worst surveys then do not ask a follow-up question of "did you think it was actually useful/compelling/fun", but of course they don't because the whole point of them is to build hype to sell the gimmick to another company who will shoehorn it into whatever.
but anyhow, the part about the lame personality (they are prompted as branding now) were the main focus of the current AI renascence, which everyone likes to lie to themselves started with attention paper, when in reality is started with a new wave of companion bots, which ironically were praise for adapting their personality really well.
From statements to spelling.
"Beyond all recognition" seems like quite an extraordinary claim.
That's honestly pretty low considering the sheer number of products that now have AI integration. I honestly think a lot of the hype around AI taking over jobs is because the people who are spreading it are more likely working in some pretty dull coding environments where copilot can shine, because you just plain don't write a ton of interesting or otherwise off-the-beaten-path code, and that's what copilot excels at. And then those people have spare time at work to evangelize about this great technology.
And before anybody starts with me, yes I have tried it. It's fine. A lot of what I work on isn't standardized or boilerplate enough to where copilot can really help me, and the situations where it can, I found the time I spent describing what I wanted, getting the answer back, copying it into my IDE and then customizing it as required was frankly just better spent writing it myself. Maybe it made me SLIGHTLY faster? But nothing approaching what I would suggest the term "coding AI" implies, and frankly, I enjoy writing the easier code on occasion because it's a nice break from the harder stuff that copilot can't touch.
Like if you're a freelancer who jumps between gigs of refactoring ancient websites into something vaguely web 2.0, you would probably get a lot of mileage from copilot where you're just describing something as written (or hell, giving it the existing code) and asking for it to be rewritten. But if you're doing something novel, something that hasn't been posted on StackOverflow a thousand times, you will run into it's limitations quite quickly and, if you're like me anyway, resign it to the bin because fundamentally asking it to make something, finding out it can't, and then making it myself is FAR more annoying than just assuming it can't and moving on.
I have a friend who works in logistics for GE and they’re getting training on the basics of GenAI and then they have to go out and find ways to integrate it into their workflow. The problem is that management isn’t doing the legwork to understand how to integrate the tooling, they’re just handing that responsibility off to the actual users. Those people wind up complaining that taking time to integrate these tools winds up slowing them down and they struggle to find meaningful applications for the LLMs.
It’s like management is saying “here’s a new hammer, we don’t know how to use it but the guy who sold it to us convinced us you can figure out how to use it. So go out and do it and be better and faster at your job, good luck”.
I'm not anti-AI. I don't mind asking an LLM to write me some boilerplate. But I don't want to change my tooling, nor have an integrated assistant. My output is fine.
I'll hold out as long as I can, but it feels like this may one day just become the reality.
There's several free options, so what's stopping you?
Here are a few things keeping Me from looking into it.
I might feel so entrenched in my process that I can’t figure out where to start.
The current crop of offerings only interest me from a self hosted point of view. I’m interested in an ai that is aligned with closely to my own world view. If there is a bias built in I would rather be generated by me.
When this ai wave first started the story was it was so complex the origin of the answers was considered in unauditable. Which I took as deceptive or the pushers didn’t understand what they were doing.
When I stated my interest it goes back to the ‘80s and I was playing with OpenCYC about 15 years ago. I keep saying I will take a look but other projects keep grabbing my interest.
Good news on a few of those points, but not all of them.
> I might feel so entrenched in my process that I can’t figure out where to start.
This is the easy one. You can just tell them that, in those words, with what you're up to, and they suggest stuff back in English. (And if you're not a native English speaker, use your native language and many will respond in that language instead of in English).
> The current crop of offerings only interest me from a self hosted point of view.
My experience with current self-hosted ones is… they're not worth bothering with. (But also, my machine is limited in the model size it can use locally, so this may not impact you so much).
> I’m interested in an ai that is aligned with closely to my own world view. If there is a bias built in I would rather be generated by me.
Good news here, but for what I consider to be bad reasons. They're often wildly sycophantic, so they'll often display whatever bias they "think" you have.
> When this ai wave first started the story was it was so complex the origin of the answers was considered in unauditable. Which I took as deceptive or the pushers didn’t understand what they were doing.
It's getting slightly better, but unfortunately the problem is mostly the "don't understand", there's too many different teams for me to feel it's deceptive. If you really need auditable results, I'd stay away from them too — they're definitely not in the vein of AI that CYC ever was.
This summer I was identifying every weed in my yard because every photo has that info. I also try to keep spell check turned off.
Your point about sycophantic ai is what I'm looking for. I see it as an opportunity to remove the white noise of the internet. I also understand your point about regarding the dangers.
My response is that with ai being no different than any technology.
Know thy self.
this included teens doing homework, and everyone else filling job applications. same mindset.
ai? sure. google search and adapting any template on the very first page? sure!