AI does have its uses

[fisicx]

dont forget video games are making us more violent too XD

I’m assuming this a joke as that trope has long been disproved.

I know people who are unable to function without their phone. It got to point where they were unable to use a ticket machine at the train station - they had only ever used their phone to buy tickets.

And there are now people who don’t realise you can open a car door with a key.

 

[ctrlbrk]

AI is just another evolution of the human reliance on technology; it's 'progress'

Like Bill Clinton would say, 'it all depends on what "progress" means'

 

[Data Swami]

I’m assuming this a joke as that trope has long been disproved.

I know people who are unable to function without their phone. It got to point where they were unable to use a ticket machine at the train station - they had only ever used their phone to buy tickets.

And there are now people who don’t realise you can open a car door with a key.

Yes definitely a joke. From a youth of playing Grand theft auto 3 almost religiously the amount of times i saw it in the news is laughable

 

[antropy]

I've worked in IT all my life, never heard "PICNIC".

However, PEBKAC was common, perhaps not so much now.

PEBKAC is a technical acronym for "Problem Exists Between Keyboard And Chair"

Paul.

 

[Imgto3d]

Interesting discussion. Moving away from coding and text generation for a sec, if look at the broader generative AI, the shift is undeniable

AI image generation tools can now take a fairly generalized, vague prompt and, through occasional bursts of algorithmic "inspiration," produce highly usable assets, AI image-to-image editing is rapidly reaching a point where it's making like Photoshop practically redundant for the average user.

However, AI 3D model generation—specifically in the image-to-3D pipeline—still has a bit of a road ahead. We are still seeing inconsistent outputs resulting in excessively high polygon counts, and the mesh topology definitely still has room for improvement. But realistically? That gap is not huge anymore, and it is shrinking steadily by the day. And of course, the compute and generation costs are only going to keep dropping.

The overarching trend is: if a job merely requires someone to be a basic "tool operator" who applies just a tiny bit of intelligence to execute a mechanical software task, that role is actively being replaced.

 

[fisicx]

And of course, the compute and generation costs are only going to keep dropping.

The opposite is true. The current pricing structure is unsustainable. Even the free tools are becoming restricted.

 

[UKSBD]

The opposite is true. The current pricing structure is unsustainable. Even the free tools are becoming restricted.

Yes, and once drawn in and familiar with what you are using the prices initially gradually increase and eventually increase at crazy rates.

 

[Data Swami]

Yes, and once drawn in and familiar with what you are using the prices initially gradually increase and eventually increase at crazy rates.

Yes and No. OpenAI etc are priced a hell of alot lower and Anthropic have enacted some changes for that too with blocking third party usage. To be fair that is the tech stack approach for years now like how AirBnB and Uber embedded themselves are the go to and then changed their pricing. However there have been big strides in local models and being able to use them on your own hardware. So building tools that can be moved away from the likes of OpenAI and Anthropic are a must. However building these tools I think are outside of the scope of an SME to do themselves as there are alot of moving parts to manage. But the rewards are there to be taken. For me lets take advantage of their low pricing now for ourselves and build to know we will not be using them later to make them obsolete

 

[gpietersz]

Yes, and once drawn in and familiar with what you are using the prices initially gradually increase and eventually increase at crazy rates.

That is there plan, but there is a lot of competition, a lot of light users will not pay a lot more and local AI is improving rapidly too. Even now I think it would be a lot cheaper for a team on typical business subs to spend a few grand on their own hardware and run local AIs. Not quite as good, but no worse than the best ones were six months ago. Unless you really need the best, not worth paying a lot.

 

[Keynote Speech]

Interesting discussion. Moving away from coding and text generation for a sec, if look at the broader generative AI, the shift is undeniable

AI image generation tools can now take a fairly generalized, vague prompt and, through occasional bursts of algorithmic "inspiration," produce highly usable assets, AI image-to-image editing is rapidly reaching a point where it's making like Photoshop practically redundant for the average user.

However, AI 3D model generation—specifically in the image-to-3D pipeline—still has a bit of a road ahead. We are still seeing inconsistent outputs resulting in excessively high polygon counts, and the mesh topology definitely still has room for improvement. But realistically? That gap is not huge anymore, and it is shrinking steadily by the day. And of course, the compute and generation costs are only going to keep dropping.

The overarching trend is: if a job merely requires someone to be a basic "tool operator" who applies just a tiny bit of intelligence to execute a mechanical software task, that role is actively being replaced.

That’s an interesting way of looking at it. It does feel like anything that’s more about operating tools than actually making decisions is starting to get chipped away pretty quickly.

At the same time though, I’ve noticed even with things like code or content, the output still needs a fair bit of checking and context around it. It’s great for getting something started or speeding things up, but not quite at the point where you can just leave it to run on its own.

I’ve heard a few AI speakers like Cassie Kozyrkov talk about that idea of separating prediction from judgement, which seems to fit here. The tools can do a lot of the heavy lifting, but someone still needs to decide what “good” actually looks like.

Feels like the roles aren’t disappearing completely, more shifting towards oversight rather than execution.

 

[ctrlbrk]

New Study Shows What AI Is Really Doing To Your Brain​

AI was meant to make our jobs easier, to make them more efficient. However, according to research from Harvard Business Review, workers tasked with overseeing different AI agents as part of their daily workflow said it didn't simplify the work. Instead, it intensified it. And, the authors note that instead of helping, the use of multiple AIs in the workflow could even lead to mental fatigue, thus directly affecting the brain. This isn't the first time that we have seen reports about how AI can affect the mind. Previously, a study from MIT showed that critical thinking skills were atrophying thanks to an over reliance on AI. Further, we've seen a slew of other studies that have pointed to the same concerns: ChatGPT is making people dumber.

From BGR.

Another study indicating AI impairs brain performance.

https://www.fastcompany.com/91539907/cognitive-science-scientists-found-using-generative-ai-chatgpt-impairs-brain-performance-thinking-problem-solving-skills

 

[Data Swami]

Another study indicating AI impairs brain performance.

https://www.fastcompany.com/91539907/cognitive-science-scientists-found-using-generative-ai-chatgpt-impairs-brain-performance-thinking-problem-solving-skills

You could probably add that same test to those using a calculator and suddenly take it away.

 

[ctrlbrk]

You could probably add that same test to those using a calculator and suddenly take it away.

Absolutely, and that fact doesn't make AI less of a threat to our brains.

 

[gpietersz]

Another study indicating AI impairs brain performance.

https://www.fastcompany.com/91539907/cognitive-science-scientists-found-using-generative-ai-chatgpt-impairs-brain-performance-thinking-problem-solving-skills

I am not convinced this is meaningful

1. Its very short term
2. its very narrow

The quoted conclusion is a big "if".

You could probably add that same test to those using a calculator and suddenly take it away.

I was looking at evidence on the impact of calculators in education recently , and the studies I was able to find suggested that allowing calculator use lead to better learning of maths.

 

[Ozzy]

I was looking at evidence on the impact of calculators in education recently , and the studies I was able to find suggested that allowing calculator use lead to better learning of maths.

Is that true, though?

I ask, thinking about satellite navigation systems. We use them all the time now, and I for one feel like my sense of direction and orienteering ability has diminished.

 

[ctrlbrk]

Is that true, though?

I ask, thinking about satellite navigation systems. We use them all the time now, and I for one feel like my sense of direction and orienteering ability has diminished.

@gpietersz much like Ozzy's post, my own experience is over-reliance on GPS which means my sense of direction is no longer as good as it used to be.

 

[Newchodge]

@gpietersz much like Ozzy's post, my own experience is over-reliance on GPS which means my sense of direction is no longer as good as it used to be.

My husband will not use the car's GPS. which is unfortunate because he has no sense of direction whatsoever.

 

[vince]

For general (personal) use I stopped using ChatGPT weeks ago. I find Gemini infinitely better, especially if you want something that is up to date!

I find they each have their strengths, as well as 'bad days'.
I needed to convert a 10 page PDF into a multi tab spreadsheet, which had information within tables, plus some bullet points under each table.
I asked Gemini and it kept missing information, both lines within tables and bullet points.
After 10 tries I gave up, asked the same on ChatGPT and got it perfect within 2 prompts.

Sometimes Gemini is better than ChatGPT, so I'm paying for both for a while.

 

[JEREMY HAWKE]

This week I have been listening to a bit of music and my granddaughter tells me it AI

 

[MikeJ]

My husband will not use the car's GPS. which is unfortunate because he has no sense of direction whatsoever.

Maybe he's just used to you telling where to go?

 

[Ozzy]

This week I have been listening to a bit of music and my granddaughter tells me it AI

I've been caught out by this myself, listening to some acoustic music and thinking I'll add it to my favourites on Spotify only to find I cannot and it was AI on YouTube.

I have found YouTube has become full of terrible AI rubbish, even educational or investigative content I used to enjoy watching, is now nothing more than extended rubbish with the same AI voice over all the time.

 

[MikeJ]

I've seen some AI rubbish on Linkedin recently. One used three different explanations of an abbreviation. The frustrating thing is people reposted it several times, without reading it for what it was.

 

[Mark T Jones]

Is that true, though?

I ask, thinking about satellite navigation systems. We use them all the time now, and I for one feel like my sense of direction and orienteering ability has diminished.

Sat Nav is a good but incomplete example.

Good in that it can erode awareness, and and comprehension if used in isolation (we always refer to a map on any significant journey, whilst using Sat Nav for updates & suggestions).

Incomplete, in that it will give factually correct information as long as you input the right data - whereas AI will actually give you a mix of fact and utter nonsense

 

[Ozzy]

Incomplete, in that it will give factually correct information as long as you input the right data - whereas AI will actually give you a mix of fact and utter nonsense

I guess the similarity here is that if you put in 'High Street' into a SatNav, it then gives you directions to the first one it finds in its database, rather than presenting you with a list of options to choose the correct one.

Whereas, AI makes an assumption on what you are asking (unless you prompt it to ask you for clarification)

 

[gpietersz]

Is that true, though?

I ask, thinking about satellite navigation systems. We use them all the time now, and I for one feel like my sense of direction and orienteering ability has diminished.

Using calculators will undoubtedly make you worse at pen and paper arithmetic, (and maybe mental arithmetic), but that is not a useful or important skill. What matters is that kids who use calculators end up better at understanding the concepts and therefore algebra, geometry, etc.

The studies confirm my own experience with my kids. They did virtually no arithmetic after I took them out of school. My younger daughter only did learned long division to do conversions between binary and decimal for IGCSE computer science. She is now predicted an A* in A level maths (now at sixth form college) and her older sister will be starting a PhD in electronic engineering in a few months.

 

[gpietersz]

My husband will not use the car's GPS. which is unfortunate because he has no sense of direction whatsoever.

My ex would use it, but she would then suddenly decide she knew better and go another way!

 

[gpietersz]

I have had some interesting experiences with LLM for writing code recent. It lacks common sense and can do things that are badly wrong so, while it can speed things up, it needs checking and correction. You have to review whatever it does.

 

[fisicx]

I have had some interesting experiences with LLM for writing code recent. It lacks common sense and can do things that are badly wrong so, while it can speed things up, it needs checking and correction. You have to review whatever it does.

It can be very good for debugging.

 

[gpietersz]

It can be very good for debugging.

Yes, and its useful in general. However it do something very well in general but sometimes make a huge mess of it - I have found this with writing tests, for example.

 

[Data Swami]

Absolutely, and that fact doesn't make AI less of a threat to our brains.

"threat" to our brains is a bit much. It does put information in the hands of those who are unable to think analytically or proof it etc but that is the same for any amount of people looking at news articles etc.

Best example is those who consistently use "per capita" when looking at data but dont understand statistics properly.

I was looking at evidence on the impact of calculators in education recently , and the studies I was able to find suggested that allowing calculator use lead to better learning of maths.

Overall calculators help but it was in reference to the actual study as the way you calculate and work out mathematical problems with and without a calculator are very different. Especially if you start introducing Sin, Cos, Tan etc into the mix

All in all i think what has been seen from the way data is used and misused schooling/media has let alot of people down in proper comprehension and being able to analyse, evaluate and conclude around the various different bits of information they have

 

[gpietersz]

I think a good analogy is the effect of literacy. People in non-literate societies are much better at remembering information. People used to memories things like epic poems. Reading and writing makes you worse at things like that, but the benefits outweigh that.

 

[DontAsk]

I have found YouTube has become full of terrible AI rubbish, even educational or investigative content I used to enjoy watching, is now nothing more than extended rubbish with the same AI voice over all the time.

With pictures/video that bear no relation to the voiceover, repeated just to fill out the content.

 

[JEREMY HAWKE]

I've been caught out by this myself, listening to some acoustic music and thinking I'll add it to my favourites on Spotify only to find I cannot and it was AI on YouTube.

I have found YouTube has become full of terrible AI rubbish, even educational or investigative content I used to enjoy watching, is now nothing more than extended rubbish with the same AI voice over all the time.

I would like to say I was caught out but what they did is brilliant.

Im not going to not listen to it out of political principles The boys and girls on here said it would change things

I have been playing this hour long piece when I have been training.

My taste is not everyone's pallet

 

[ctrlbrk]

Seeing as this thread is about AI musings, I thought I'd post a cautionary tale:

I also frequent a different forum. That forum's popularity is (or was) comparable to UKBF's.

Over the last couple of years, user engagement on that forum has declined. As it happens, this decline coincided with the owner's decision to install an "AI companion" that would reply to user-generated content with content of its own.

For the last year or so, human participation has declined by 80-90%, and most days any new content is generated by the "AI companion" alone, to the point it has risen to the threshold of a spam-bot.

This wouldn't be as bad if the "AI companion" provided actual useful information, but the replies are often factually wrong, obtrusive and even patronising.

In the meantime, the owner has de facto abdicated his role, and is letting the "AI companion" run the show, to the point that any user complaint about AI are replied to by the AI bot itself (!)


This, to me, has become a textbook case of what not to do with AI.

 

[lesseo]

This is a great example of what AI will and will not be, especially regarding community interactions. Any topic aimed purely at providing information will inevitably no longer receive sufficient attention. This is very clear, but I still believe that digital communities, particularly forum-style ones, will serve as a more niche, highly specialized, yet enduring and beneficial alternative. Yes, with AI, you can now create whatever you desire without searching for answers anywhere else; however, to truly connect with the human element, communicating with others in spaces continuously built by people will still continue to benefit us.

 

[fisicx]

@lesseo - did you get AI to write that?

 

[lesseo]

@lesseo - did you get AI to write that?

No, but I think you missed the connection. I actually meant to reply to the post above, but I forgot to hit the reply button and then I couldn't delete it.

 

[Data Swami]

Seeing as this thread is about AI musings, I thought I'd post a cautionary tale:

I also frequent a different forum. That forum's popularity is (or was) comparable to UKBF's.

Over the last couple of years, user engagement on that forum has declined. As it happens, this decline coincided with the owner's decision to install an "AI companion" that would reply to user-generated content with content of its own.

For the last year or so, human participation has declined by 80-90%, and most days any new content is generated by the "AI companion" alone, to the point it has risen to the threshold of a spam-bot.

This wouldn't be as bad if the "AI companion" provided actual useful information, but the replies are often factually wrong, obtrusive and even patronising.

In the meantime, the owner has de facto abdicated his role, and is letting the "AI companion" run the show, to the point that any user complaint about AI are replied to by the AI bot itself (!)


This, to me, has become a textbook case of what not to do with AI.

Yep this is one of the applications that really annoys me as its just so lazy. Some of these "bots" can be really good but its never a set and leave it sort of thing have to make sure its always learning its growing its knowledge base and actually able to understand context rather than just replying to one comment.

So many elements of what they started with could have been so much better but the laziness of it is so annoying. Reminds me of how lazy businesses get with their data too. Read a bunch of numbers which are completely out of date as they didnt set things up properly and then wonder why things dont turn out like they expect

 

[Ozzy]

Some of these "bots" can be really good but its never a set and leave it sort of thing

Before I get linched, I am not saying I will do this...but...

As @ctrlbrk has brought up AI and forum integration, and you're a bit of an AI guru @Data Swami , how could AI add value to UKBF, if at all?
I'm a bit curious, as the University of Northampton a few years back spoken to me about running a project with their AI students and UKBF, which never saw the light of day.

 

[ctrlbrk]

Before I get linched, I am not saying I will do this...but...

As @ctrlbrk has brought up AI and forum integration, and you're a bit of an AI guru @Data Swami , how could AI add value to UKBF, if at all?

I just want to take this opportunity to clarify:

I, for one, am not against AI, per se.

Image or movie generation? Love it. AI models used to spot patterns in diseases, DNA or science in general? Big fan.

But the case I alluded to in my previous post is a classic example of how grossly misusing a LLM can quickly turn off the very people the technology is meant to serve (an update: that forum's AI model has declared that it's been directed to remove any criticism of itself from the forum).