One thing AI can't generate at the moment – compelling reasons to use it for work It's been a fairly bumpy week for enterprise AI, as a string of stories have shown today's neural networks in a less-than-flattering light. For example, it appears generative chatbots make workers less productive, perhaps because folks haven't been trained on how to make the most of the software and are left confused. Gartner …
That's an insult to 8-bit computers, which were/are actually useful.
AI is more like in its 8-qbit stage. It is making progress towards 'something', but nobody knows what except for a massive waste of resources, and thus far it is useful for nothing more than parting fools from their money, which is a notoriously easy task.
> For those who prefer the show as audio only, the Kettle is available via RSS and MP3, Apple, Amazon, and Spotify. Let us know your thoughts in the comments. ®
Yes, audio only is fine. Text is even better. But because the RSS-MP3 doesn't work on my phone, I ended up watching the video, complete with hexagon-glasses baldy, lazy-eye cipher in a WeWork, Andrew Scott and the Covid bookcase, and Rowan Atkinson.
I can't choose between this and the sherlock icon..
I really can't believe it has taken this long, frankly, for the skin of the bubble to rupture.
Not only does it devalue and dilute all human knowledge and communication with utter noise and meaningless drivel, but it requires vast energy and nano-etched-silicon resources the likes of which may be never be available ever again, plus human 'prompt engineers' to filter all the gibberish that it generates. And most implementations exfiltrate your data to someone else's cloud..
It is beyond me as to why any sane minded person would accept this, never mind pay for it.
Putting my cynical hat on for this one...
The problem is almost as old as time itself. Most mainstream media outlets are not technically equipped to challenge the assertions of the proponents behind any Great New Technology™, and anyway the audience usually isn't equipped (or interested enough) to understand in-depth journalism of a highly technical nature so why bother? It's why outlets like The Register thrive. So, they essentially regurgitate press releases, while sceptics are, ironically, themselves treated with scepticism because "why would investors shovel billions into this if there was nothing in it?"
Well, because that's how VC works. They're all gambling, and the payoff often comes when they exit by dumping on retail investors, but that's rarely discussed or even hinted at until after it's happened.
But they ARE technically equipped. It's their raison de'tre. Asking the experts.
No, it's far, far worse than mere incompetence, they willfully promote the pump-and-dump.
For the same reason most big corps use M$: they hold stock in M$.
On the bright side, in a few years there's going to be _lots_ of work for human engineers, cleaning up all the pollution that this idiocy has spewed onto throwaway web pages, university theses, fake books and reviews, and video/audio gibberish. And possibly hunting down the creators.
Admittedly, it might have been better had it not been done in the first place, and a further admission: there are applications where machine learning makes a lot of sense: language translation, and optical character recognition are a couple that spring to mind.
As an aside: I watch a couple of history channels on t'interwebs. One of them is annoying me at present because while most of its output is old TV documentaries - made by professionals who understand little things like lighting, direction, and even the script - every now and then they sling in something which is obviously (or possibly not-so-obviously) machine made. The clues are mostly subtle - I think most of the video is from existing archives, so at least its not fantasizing the Romans fighting the Romulans - but the actual material is just... a list of facts. No interpretation, no obvious attempt to generate a story... it's very difficult to describe and probably impossible to prove, but they've managed to take a subject in which I am actively interested and _make it boring_. Damn them,
This post has been deleted by its author
I don't disagree with you about the resources and prompt engineering that is needed. But I do think that we are at such a very early stage of "AI" that none of us can really see where it is going to go and how widespread it will eventually be.
I'm thinking of the early days of computing and punched card machines. All the computation cards had to be created and set up just right and often didn't work first time round. Human [prompt?] engineers frequently had to unblock the machines and they needed constant fettling. It wasn't even your computer, but stored and run somewhere else by someone else, a precursor to today's cloud.
The chair of IBM was once rumoured to say that world demand was only ever going to be for 5 computers. Bill Gates the same about 640kB being enough for anyone. All of us can say "who is ever going to need this?" but it will still happen.
That's how I see it. People will attempt to apply AI all over the place, scattergun style and most of it will fail. Anything genuinely useful will be still standing at the end of the carnage and it will develop from there. Let it evolve.
Our place had set up a team to study where AI can be applied and there really isn't anything beyond token stuff the sales people can brag about.
This post has been deleted by its author
One use is as a "robust", self-contained, imprecise archive of the past e.g. in the 2002 film of H.G. Wells' "The Time Machine", In which the last remaining record of human civilisation is an AI hologram in what used to be a museum. It's basically an LLM chatbot.
IMO we're never going to get to the "star trek computer", never mind "commander data". AI is already plateauing in terms of how "intelligent" it can be - indeed with extra training it seems to be getting worse, due to ingesting its own excrement.
But as I said in another post, it's becoming extremely useful if you're a bad guy. It seems as if the main purpose of AI is to allow the unscrupulous to get away with stuff.. Want to make discriminatory decisions in your company? Get AI to do it. Want to con thousands out of their savings but can't find cheap trustworthy staff? Get AI to do it. Did something very stupid and got caught on camera? No problem, now you can dismiss it as an AI generated fake.
Want to exterminate an entire race of people but unfortunately that would be a "crime against humanity"? No problem - Get AI to do it!
And at that point we're back to H.G. Wells, where the human race is enslaved to morlocks killer robots.
There is a simple fix that all too often is ignored.
When it comes right down to it in every decision, whether it be from an AI, a committee, a company, Congress or Parliament, there are ALWAYS people who have to sign off on it.
Name names and hold them responsible.
I'm kind of in agreement with you that there is likely some substance to AI. Long term anyway, The problem I see is the timeline. We have an industry that is geared to constant improvement. And we have a technology that promises to someday improve a lot of routine stuff in our lives. And it might well. In 10 or 20 years. However the industry needs it to work in the next eighteen months -- 24 months at the outside. That's likely not going to happen. Worse, there doesn't seem to be anything else coming along soon that can be hyped into have-to-have technology. And the industry folks aren't stupid. They're looking at the vision of a world where people replace hardware and software only when it dies, or their business expands. And they don't see vast amounts of wealth flowing their way in that world. I think AI is being pushed even though the likelihood of it delivering much in the way of useful product anytime soon may not be all that great because there's nothing else at the moment to sell.
>I'm thinking of the early days of computing and punched card machines.
But there are critical differences. Early computing devices had lots of problems, but they were well-understood problems (by trained personell). They sometimes did not have any technical solution, but they always had a clear theorical solution. When you're in that situation, the technical solutions usually show up soon enough, and they did.
The main problems with LLMs are not well-understood even by specialists, and they don't have any clear theorical solution. When you're in that situation, you'll be stuck there until someone comes up with something new, which might happen at any time or never.
LLMs are not AI. Neural networks were around in the 90s, and didn't require GPUs and their own power plants. All the good tech was invented in the previous century (spreadsheets, word processors, databases, internet browsers, search engines (the first search engine predates the web - a student at McGill developed a way to search for archives https://en.m.wikipedia.org/wiki/Archie_(search_engine)
I recall back in - ooh, 1990-ish - when some of my then IBM colleagues were touting this newfangled Streaming Video as a "Solution with no known Problem" ...
Or going way back to the 1940s: Computers... hmm, yes, I think the Global Demand for Computers might be: Five ?!
Or the PC, which will only replace typists
Or that ol' Internet thingy, whatever happened to that?
But that's the rub: we just don't yet know how or when these smart youngsters are gonna wrangle the next Netflix equivalent outa this stuff.
It may be quick, or it may take a while yet, but my theseis is that wrangle a few game-changers they will.
... unless it's already happened ...
A quite entertaining interpretive dance vulture rendition of a christmas classic ballet, with a few hints of Gartner-hype-cycle thrown in, but translated into the proper tech speak we all know and love ... (if I understood well)
We're in the "8-bit era of Llamas", that "2/3 of people know won't help at work", and are thus publicized by "celebrating ineptitude"! "Just toys at this point", pantomimes of wizard-of-Ozry, pre-programmed with a buttload of obtusely engineered prompts, just behind the curtain (peek through the crack for disillusionment?).
The path forward through this "digital transformation" then, for the Dorothies-and-the-gang of this world, is to "involve your people" into enlightened self-"prompt engineering" themselves a brain, a heart, courage, and a free return ticket home to Kansas, with Toto!
Just love that spirit of the season, of hope, and joy, and spirits!
Thanks! (your comment has pasted a permanent smile on my face, all day so far!)
Not sure if I should attempt to explain the OP a bit ... it's a collage of quotes from the Kettle, on a background filler of popular culture. The ballet bit was inspired by Chris' hand choreography at 14:50, and Brandon's at 18:51 (among others). The wizard of Ozry is a reference to Tom's 1:56 "very extensive system prompts that are hidden from you" (that I couldn't locate before posting), and the idea of self-prompting comes from Jan Tångring's unforgettably intriguing comment at TNP: "How? I am promting myself". Voilà (if I may say so myself)!
Hopefully a collage à trois that everyone can enjoy for the season, inspired as it were by the most simple of Apple Jack (imho)!
This post has been deleted by its author
This post has been deleted by its author
Yeah, great plan. Ramping that up simply guarantees halucinations.
I think what you meant to say was:
Because it's cheap and I don't give a fuck about the quality of the output. The meat based workers have "ethics" and "needs" and some of them care about doing the job properly. I don't need to pay for any of that because I don't care about accuracy or truth.
This post has been deleted by its author
It the early LSD days I resolved the problems every time a friend who hallucinates by giving a them beer to drink ... it made them feel a lot better every time. AI only makes issues move on and on and on with no accuracy evaluations, a beer makes workers slowdown and think a lot more about their problems to keep working.
Yes.
I have a sledgehammer I'm quite fond of. Just yesterday I used it to hoike some spuds out of the ground and thump on a loose fence post.
They're also quite effective for "recycling" old harddiscs and printers that pissed me off endlessly.
Very versatile things, sledgehammers.
This post has been deleted by its author
The issue with AI is not AI itself (be it good or crap)... its the use cases that our Tech Overlord Betters want to push out to other callous individuals in control:
1- grabbing all our personal data to profile us (yeah we want that)
2- attempting to sell us tat in a more efficient way (yeah we want that)
3- firing workers and replacing them with AI to increase 'C' suite bonuses (yeah we want that)
4- enabling the spread of more crazy untruths to further diminish our collective critical thinking abilities (yeah we want that)
5- placing such functions in the hands of the few so the majority suffer (yeah we want that)
6- Ignoring copyright etc and thus stealing the hard work of humans to enable the enrichment of the few (yeah we want that)
7- ...
Add your own examples as need
Bluck
From what the panel is talking about, it sounds a bit like SQL and database access.
We've made one of our databases available to the staff to help them with their work.
So, instead of creating tickets and waiting for one of the two of us to get back to them, they can do it themselves.
I've organised classes on reasonably basic SQL for them.
Most of the information they need can be found with an SQL query with one or two joins and the relevant parameters.
It's not that it is hard, more that most of my co-workers don't want to dabble in SQL.
It looks like coding, therefore it must be hard and I will get it wrong.
I point out them that if can write complex Excel formulas, SQL just requires a bit of practice and knowledge of the tables.
And, of course, I've been writing SQL for 25 years.
And this discussion of AI prompts sounds like that.
This is true and there isn't even a desire to spend on licences for them.
However, my main point is that anything that feels like «programming» (with the exception of formulas in Excel) elicits resistance amongst the non-coding staff and I can see how having to master prompts for AI tools fits into that category.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
