Forum menu
My dentist recently showed me my x-ray that had been analysed by machine learning. ****ing robot overlord decided I needed 1.5 hours of torture that I had to pay £300 for the privilege of enduring. Nice little earner.
In Iain M Banks' Culture books, a galactic society is run by AI and humans have no need to work. It only works because human greed and lust for power has somehow been eliminated.
Go back 30 years and look at how beneficial the internet could have been, and then how it's been corrupted.
My current view of LLMs… it’s like the biscuit factory in Bagpuss…
In Bagpuss the mice pretend to work a biscuit making machine… but all that happens is they put a biscuit in, and then take a biscuit out. Obviously you can’t make money this way, unless… you steal the biscuits that you put in the machine… and then use the output to your benefit
I think the big issue is how you make money, what were they valued at $1.5bn 🙂
The Bloomberg website literally does an ai summary on the the article for fear that people can’t cope with reading more than one paragraph 🙂
My dentist recently showed me my x-ray that had been analysed by machine learning. ****ing robot overlord decided I needed 1.5 hours of torture that I had to pay £300 for the privilege of enduring. Nice little earner.
I think your comment is meant to be tongue in cheek, but by parallel to the cancer diagnoses, it's identified earlier that you need that work and prevented you needing 4 hours of work over 2 visits and a £3000 crown?
Good question about what constitutes AI, is just automating processes (even if it's doing it better/faster) really AI or does there need to be a learning element to it, so the more it does the better it gets.
The LLM stuff is unsustainable BS. A bad bubble that a business-oriented ‘tech’ industry has seized on as it is low/bereft of actual innovation. They’ve been burbling on about this stuff for a few years now and there’s not much there despite the claims of a ‘revolution’. A less innovative, more beta metaverse imo.
Wasn't it chocolate biscuits made from butterbeans?
“Breadcrumbs and butterbeans”… but that was all part of the sleight of hand… it was really just biscuit in, biscuit out. Great if you have a free supply of biscuits stolen from the world’s biscuit makers.
I hate it as in a majority of cases it’s going to be used to flood social media,films, music and literature and news with low quality material.
This is where I’m at. But rubbish produced without the need for loads of skilled workers is, in so many cases, going to win in the marketplace over excellent content that comes with an expensive wage bill. Especially if the “marketplace” is carved up between the big tech cos that have the most to gain from this. Unions in the USA have fought back hard, and in many case are holding the line… but the tech companies they are up against understand the “danger” of unions, and are out to break them in the long term, while also continuing their takeover of spaces where we are used to traditional media and culture companies and organisations thriving.
The story can be anything that the author wants though.
If you don’t need character illustrators, animators, story board artists, soundtrack composers, colourists, editors etc… you definitely don’t need “authors”. See the writers strikes in the USA. They were fighting for their very existence, and they knew it.
In the medical side with the diagnostic stuff though it’s great.
Agreed. And already achieving things people can’t do. It’s not just replacing human diagnosis, it’s spotting patterns in medical data for people with known outcome that human’s can’t, and then finding those patterns in the undiagnosed that would otherwise be missed. Healthcare is going to be transformed for the better. This stuff is hard though, unlike the cheap wins the big tech cos are after.
I think your comment is meant to be tongue in cheek, but by parallel to the cancer diagnoses, it's identified earlier that you need that work and prevented you needing 4 hours of work over 2 visits and a £3000 crown?
Well yes, I see this as one of the beneficial applications, computers helping in areas where they excel. But it isn't AI it is pattern recognition. The dentist said he had already identified the areas of decay and the machine learning analysis agreed.
The growth isn't linear, predicted AGI is 5-10 years away, ASI is 10-15yrs, predictions yes but at 45yrs old id expect to see the changes roll out well with my lifetime all being well, my lads (5yrs old) will see the most part, and the bit that worries me more so.
The fact we may be in a situation where a mind far advanced of our own may have the ability to destroy us as a species in the next 20yrs? The 'God father' of AI saying we have a 10-20% chance of that scenario happening - this is why I started the thread in the main part.
Akin to living in the cold war but with a bomb that can think for itself
"AI" is a very general term that covers a huge range of potentially valuable applications.
Chatbots generating random text, however, are literally destroying knowledge at a rapid rate. I fear for the future. And if anything can be learnt from the Turing Test, it is that humans are gullible and easily tricked.
The growth isn't linear, predicted AGI is 5-10 years away, ASI is 10-15yrs,
30 seconds of google showed answers ranging from "some time this year" to "100 years"... About the same time Fusion power plants arrive.
LLMs are here to stay. They're pattern matchers and probability and zero true understanding of source material. A tool to mash up what you feed in and provide a "most likely" answer. Tremendously useful but just a tool.
Im finding all this very interesting.
At the moment i dont think we have a problem. We have a useful tool and we can build on it.
If and IF we get to generative AI then yes i think there is a real issue.
I think the main problem with AI is the fake stuff and massive potential for misuse, which we're already seeing and is only going to get worse.
Social (and traditional) media can be very useful but is already a hot mess of misinformation, manipulation and outright lies. Add into that the AI-generated deepfake stuff and it's getting impossible to know what's true without considerable research. Even more worrying is that the models themselves can be trained to manipulate - just look at Elon's Grok and the recent news of him tweaking it to output his propaganda.
The long and short of it is that AI output cannot be trusted. And not only because of hallucinations. None of them are simple unfiltered LLMs, even the common "trusted" ones such as ChatGPT, Gemini, CoPilot etc are trained to always be polite, always positive etc etc... that is manipulation of the raw results. Who knows what other hidden prompts and adjustments are made to the results? I think it's a dangerous slippery slope.
On the other hand, I don't think it'll get that far re. the 'AI takeover'. Yes there are dangers to AI but the prophesied meltdown of all creative content won't happen - yes it'll get bad but people will pull back, AI usage will shift (I have no idea to what) and IMO the internet as we know it won't be destroyed as some seem to think.
I am a software developer and have recently turned CoPilot on in Visual Studio. As I type it suggests code snippets and I have to say it's quite amazing. It's an odd mix of very helpful and a bit thick, but for many tasks it's a big timesaver. Write a few comments, hit tab and boom, saved me a few minutes of typing. Granted sometimes it's providing rubbish but overall (learning how to prompt it) it's very useful. I feel dirty. (But have zero concerns over it taking my job anytime soon.)
Your faith in human nature is greater than mine. I'm already seeing apparently intelligent people wholly outsourcing their analysis and thought processes to chatbot drivel. They do it because it's easier than thinking and they just don't care much about the validity of the output so long as it looks truthy.
If it gets rid of tedious jobs then good. Although we had fun this week using AI summary on a report our client had overpadded using Ai. Hopefully it will reduce reports to a series of bullet points in the first place.
That of course is the utopian dream - we get AI to do the chores like washing up, ironing, writing reports so we can enjoy things like drawing, taking photos, making music etc.
Instead we've got "AI" making 'music', creating images and doing awful drawing so that we can spend more time washing up, ironing and writing reports.
Maybe we’ll get AI-assisted route planning. Louise, plan me a 25 mile circular route from my current location with 3000 ft of total elevation that is mostly off-road and avoids footpaths that have no history of cheeky use, or that are frequented by surrons, bitey dogs, grumpy farmers, doggers, gammons, militant ramblers or other miscreants. Busy A roads should be avoided, as should fields likely to have cows with calves. At least one cake shop near the route is essential, preferably two.
Your faith in human nature is greater than mine. I'm already seeing apparently intelligent people wholly outsourcing their analysis and thought processes to chatbot drivel. They do it because it's easier than thinking and they just don't care much about the validity of the output so long as it looks truthy.
Thankfully the world is not run by such people though.
...What am I saying? We're doomed.
i found this interesting
Maybe we’ll get AI-assisted route planning. Louise
Wait, AI identifies as "Louise" ?
We're already dead
First they came for the models
I am somewhat astonished that the A.I. vanguard isn't being lead by the porn industry. Stuff like this usually is.
Im finding all this very interesting.
At the moment i dont think we have a problem. We have a useful tool and we can build on it.
If and IF we get to generative AI then yes i think there is a real issue.
How are you defining "generative AI" though? ChatGPT, Grok etc can all make new content.
Im finding all this very interesting.
At the moment i dont think we have a problem. We have a useful tool and we can build on it.
If and IF we get to generative AI then yes i think there is a real issue.
How are you defining "generative AI" though? ChatGPT, Grok etc can all make new content.
They can make new content based on rules. Once they can develop themselves then all bets are off as the speed of development will be exponential to our demise...
Don't have time to find a better link, AI was told it was being replaced for updated version, AI then jumped to newer server, when asked denied knowledge of jumping servers...
How are you defining "generative AI" though? ChatGPT, Grok etc can all make new content.
They make content by mashing up what bits of old content they deem most likely to be "useful" when put together, depending on the prompt. It's not strictly new, more a collage of old. If you only ever fed it Charles Dickens, you wouldn't get Irvine Welsh.
I am somewhat astonished that the A.I. vanguard isn't being lead by the porn industry. Stuff like this usually is.
You’d have to be studying porn in depth thou to answer this question 🙂
There’s apps and sites like nudify will take a normal picture and turn it into a naked picture.
So a lot of jolly japes as these are passed around 🙁
They can also swap faces on video with deep fake tools.
But we will probably see more wholly computer generated porn as the floodgates open.
They make content by mashing up what bits of old content they deem most likely to be "useful" when put together, depending on the prompt. It's not strictly new, more a collage of old. If you only ever fed it Charles Dickens, you wouldn't get Irvine Welsh.
I think this is the gotcha it’s easy to assume that ai is actually ‘intelligent’ and there’s a bit of thought going on rather than clever text processing.
it’s easy to fall into the trap as it does appear that there is intelligence , it sort of gives answers that look good but sometimes they look good but are totally wrong.
it’s the ultimate bullshitter 🙂
Grok is frightening, imagine a generation believing everything grok tells them 🙁
Still if you’ve a news website and want 50 stories based on the same event then it’s happy cheap content generation days for you 🙂
Write once and ai generate the other 49.
for many tasks it's a big timesaver
You’re one of the few for whom LLMs have some use then. Seems sometimes that time saving is an illusion.
https://www.theregister.com/2025/07/11/ai_code_tools_slow_down/
for many tasks it's a big timesaver
This morning i’ve been using the frankly brilliant Adobe AI in photoshop. The generative fill feature did in a couple of minutes what would have taken me a couple of hours of retouching, allowing me to get on with the actual design work.
I use it on a daily basis to do the work that isn’t remotely creative, but just time-consuming and laborious
clever text processing.
the ultimate bullshitter
Make up your mind which it is, because there is no balancing of the scales between them.
You’re one of the few for whom LLMs have some use then. Seems sometimes that time saving is an illusion.
Interesting article, though not entirely surprising. I tend to use it for simple stuff, basic functions or something which would otherwise be just repetitive typing or copy/pasting. For more complex functions the results tend to be either "looks ok on the surface but is complete twaddle" or "pretty good, needs tweaking though" which I can see how that easily leads to actually losing time when you think it's being helpful.
I guess my usage is fairly similar to binners':
This morning i’ve been using the frankly brilliant Adobe AI in photoshop. The generative fill feature did in a couple of minutes what would have taken me a couple of hours of retouching, allowing me to get on with the actual design work.
I use it on a daily basis to do the work that isn’t remotely creative, but just time-consuming and laborious
How are you defining "generative AI" though? ChatGPT, Grok etc can all make new content.
They make content by mashing up what bits of old content they deem most likely to be "useful" when put together, depending on the prompt. It's not strictly new, more a collage of old. If you only ever fed it Charles Dickens, you wouldn't get Irvine Welsh.
That's exactly what generative AI is - it means that the AI can make new content based on patterns learned from the training data.
just watched that video Somafunk posted on the first page. The part about the chatbot that starts about 22 minutes in is pretty jaw dropping. 🤔 Forget about worrying about global warming or world war three, I am off to look up Sarah Conner in the phone book!
Prediction: AI will be blamed for problems actually caused by HS (Human stupidity).
Older STWers will remember the prevalent excuse 'computer error' which was usually a cop out for human failure.
Are these still labelled as ID-ten-T errors?
It's still a case of garbage in, garbage out. If the AI is trained on an LLM full of drivel, then it will spout drivel.
It's still a case of garbage in, garbage out. If the AI is trained on an LLM full of drivel, then it will spout drivel.
The biggest problem is going to be the feedback loop. Half the pages on the web were written by AI and these are being fed into the next generation of AI as training data.
It's so hard to tell what was written by an AI and what wasn't that errors are going to pervade the training data and compromise all of the output.
This site was served up to me quite prominently in a search yesterday. I am certain it is entirely AI generated, from the text to the images to the social icons that don't actually link anywhere. So it cannot be trusted. It's American in tone but based out of Czechia and presumably just trying to gain ad revenue. But it's ostensibly a health site. This shit is going to overwhelm the internet, really quite soon IMO
https://www.theonlineallergist.com/
I thought I would check if I could fake my own artwork in AI to save me time thinking and being creative.
I wrote an AI app that let you upload an image of a painting and also a photograph of something different. It analysed the painting and applied the same style too the photograph to create a new picture.
https://photos.app.goo.gl/wXdTpbhLR3AFzbAw9
If you are wondering why my painting shows his sword pointing up and the statue has the sword pointing down the come to the exhibition in Winchester September 19th - 26th at The Nutshell where you will read
"
"
Think your AI App might be the problem. Done with ChatGPT
We are in the edge of enshittification of chatbots. The are now consuming ai generated content which is inherently bollocks.
The only models that are useful are taking llm base models and performing refinement training in a locked environment on a companies own data. That way you have a level of control and monitoring of output.
Specialised models should be good though
Its the rapidness of the advance I think is the most of my issue, there seems to be a lot of high level people asking for legislation, laws, controls being put in place but its coming at such a slow pace compared to the advances of the technology
I'm surprised nobody else seemingly picked up on the drive for legislation, but my gut feeling is that it's being backed by the existing big players for the protectionism that could bring to their marketplace. If there's rules that say your new whizz-bang AI tool needs to pass incredibly complicated and expensive certification, that plays into the hands of those big players and makes it difficult for small startups to break into the market, no matter how good their idea is. Small startups will just have to sell their soul to a major player to even cover the cost of basic certification, and the big guns get bigger off the back of it. And round and round it goes...
As for my own actual use of AI tools; I'm starting to have more faith in them for software development where the answers I seek are binary. It either works or it doesn't, and generally that's then independently verifiable. I certainly wouldn't trust the current regurgitation AI tools to come up with truly "new" ideas, designs or concepts, nor would I ask it to write its own tests to verify its own output!
We are in the edge of enshittification of chatbots. The are now consuming ai generated content which is inherently bollocks.
I’ve noticed a helluva amount of ai bots posting in the comments of YouTube videos, all with a similar response such as “your content is an inspiration to me, I admire your professionalism” usually with a picture of a woman’s arse or tits etc.
What do they get out of such a thing?
Noticed similar on twitter comments and more so in political comments/replies along with a huge amount in the comments under Gaza tweets, if you check out the posters behind such tweets they all appear to have accounts that are created years previously but they are very obviously bots as they all post variations on the same comment.
The Enshitification is well under way.