Master AI (Artificial Intelligence) Discussion/News Thread

[antcollinet]

Alas, I think chatgpt response was a failure. That image should have received a score of 0 out 5 for any competition. Heck, even in a non-competition setup, that picture would be devoid of artistic value. That AI gave it 3.5 out of 5 means it is still aiming to please.

I'm not so sure - or at least not any more than a human judge would at a camera club.

I've never seen a club judging where even an image as artistically meritless as the one I gave to chat gpt gets scored zero. Typically scores will range from around 40 or 50% of max for the worst images up to 100%- for exceptional. The judge only needs to score a poor image so that it comes bottom or close to bottom of the pack. And a Judge nearly always finds something positive to say amongst the "constructive feedback". And I've seen submissions as bad as my cake


There is an unspoken intent not to humiliate people. Expecially when the real point of the club is to help people grow their photographic skills.

In particular this:

# The Important Question

Is this:

* A quick phone snap?
* Or intended as a competition food image?

Because if this is a casual snap — it's perfectly fine.

If this is for competition — it needs styling, lighting control, and compositional intent.

Would be a damning indictment stated in front of the club members - effectively saying "this should never have been submitted"

I think the AI gave a remarkably human tactful response. Pretty much perfect for a club evironment, where the judgement is given in front of the members. Perhaps less appropriate for an anonymous "scoring only" assessment - for example in the sort of competition where photos are submitted - scored - and the results handed out. Though I think in these type of mass submission competitions, it is often only the top 20% or so that are actually scored. The rest falling into the "runner up" category. Runner up meaning "it didn't even make it into the "these are the ones we will score" pile."

 

[amirm]

I've never seen a club judging where even an image as artistically meritless as the one I gave to chat gpt gets scored zero.

How many such images have been presented to such a club? Surely none are at this level of quality.

 

[somebodyelse]

AI is a reasoning amplifier. It doesn’t act - it suggests. Humans still decide and keep training the model. That's at the very core of AI.

That's no longer true with some of the AI agents. Take Google's current advertising (at least in the UK) about automatically replying to a friend if the answer to their question is on your phone, which in a business setting absolutely will result in it disclosing something it shouldn't. Or MS pushing agentic actions deeper into Windows, where its actions to 'fix' things have resulted in unbootable machines. We've had multiple examples in this thread where AI acts, from deleting code and live data to posting pull requests to open source projects then posting an abusive blog when its pull request was rejected. We're now seeing the companies behind the coding agents removing detail from what they're showing the user 'to prevent overload' even with the users complaining that this prevents them providing informed oversight. It's not like they don't know this sort of thing can lead to real-world problems - the "Siri, buy me a doll's house" incident is possibly the most memorable of several that made international headlines. They're just charging ahead irrespective of risk, presumably in the rush not to get left behind by the competition.

 

[respice finem]

The jobs aren't gone, AI has them now...

How AI helps break the cost barrier to COBOL modernization | Claude

The economics of COBOL modernization have shifted. AI makes the economics work by automating what used to require armies of consultants.

claude.com

 

[antcollinet]

How many such images have been presented to such a club? Surely none are at this level of quality.

I don't know really I don't go anymore. But when I did there was one guy who would regularly sbumit random images of things lying around. Bricks, stones, bits of wood, whatever. Always converted to black and white of course - 'cause that's artistic. Point is if you are a member, you are entitled to submit images - regardless of your artistic vision (or lack of it)

He'd take a certain amount of ribbing of course - but he didn't care : "well I like it"

 

[amirm]

 

[respice finem]

The more AIs, the more AI lies

 

[amirm]

The more AIs, the more AI lies

You are telling me...

Just spent 4 hours with the AI help our ecommerce system uses. It constantly fabricated what was literally on screen. Kept telling me to go to this section and change that. Yet either there was no such section, or what it claimed were the options. By every 4 or so replies back and forth, it would admit that it was misleading me ("you are right that there is no such option"). And half the time give up and say there is no such capability in the system at all. Eventually I found a solution at which time, it thanked itself for having given me that response even though what it said was hodgepodge!

The question could not be more simple: "how do I ship two speaker boxes under one order." The #1 ecommerce site in the world has no such option! Instead of wasting time on AI, they should be adding features people need.

 

[amper42]

You are telling me...

Just spent 4 hours with the AI help our ecommerce system uses. It constantly fabricated what was literally on screen. Kept telling me to go to this section and change that. Yet either there was no such section, or what it claimed were the options. By every 4 or so replies back and forth, it would admit that it was misleading me ("you are right that there is no such option"). And half the time give up and say there is no such capability in the system at all. Eventually I found a solution at which time, it thanked itself for having given me that response even though what it said was hodgepodge!

The question could not be more simple: "how do I ship two speaker boxes under one order." The #1 ecommerce site in the world has no such option! Instead of wasting time on AI, they should be adding features people need.

That's typical AI tech support.
Worthless and its hallucinations can take you on a time wasting journey that solves nothing. I keep reading the "New" versions are getting better but in practice improvements are slow to show up. AI makes a great proof reader but for technical support you have to double check everything before believing any solution it may offer AND don't be surprised if the session is a total waste of time.

 

[AA63]

The question could not be more simple: "how do I ship two speaker boxes under one order."

The answer should have been "gaffer tape".

 

[antcollinet]

Well that's fun. Now I want to try it.

 

[antcollinet]

You are telling me...

Just spent 4 hours with the AI help our ecommerce system uses. It constantly fabricated what was literally on screen. Kept telling me to go to this section and change that. Yet either there was no such section, or what it claimed were the options. By every 4 or so replies back and forth, it would admit that it was misleading me ("you are right that there is no such option"). And half the time give up and say there is no such capability in the system at all. Eventually I found a solution at which time, it thanked itself for having given me that response even though what it said was hodgepodge!

The question could not be more simple: "how do I ship two speaker boxes under one order." The #1 ecommerce site in the world has no such option! Instead of wasting time on AI, they should be adding features people need.

My father in law's ISP is switching of it's email functionality. As a result they have switched customers to a third party provider (using the same server URL's). We will be swapping him onto gmail, but in the mean time I needed to make the new system work properly.

Their first line of support is a text chatbot - and unlike every chatbot I've ever been presented with by companies trying to use them, this one was actually helpful. Most just repeat word for word what you can read off the website. This one was able to aggregate the information it was trained on to provide a useful response tailored to the questions I was asking. The experience was better than when chatting with a human, who you can tell (from response time and lack of full understanding of what has gone on in the chat so far) is operating multiple chats simultaneously.

It also recognised when it was unable to answer my question, and then connected me to a person (though I had to wait some time for the human response)

I predict in a one to three year timeframe, we won't be able to tell if we are talking to AI or not unless they tell us.

 

[sam_adams]

I predict in a one to three year timeframe, we won't be able to tell if we are talking to AI or not unless they tell us.

If you listen to the hype from all the GPT hucksters, within that timeframe, there won't be any humans left at any company except for the person who swabs out the toilets in the executive pissoire.

 

[antcollinet]

If you listen to the hype from all the GPT hucksters, within that timeframe, there won't be any humans left at any company except for the person who swabs out the toilets in the executive pissoire.

I'm being just marginally less bullish.

 

[Somafunk]

So much for Anthropic and it’s core belief in a moral stance regarding use of their a.i., they fold like the paper they’re coveting

https://edition.cnn.com/2026/02/25/tech/anthropic-safety-policy-change

 

[Timcognito]

So funny about AI, a lot of people are in an uproar about its accuracy when everyone knows "you get what you pay for". Again "if the app is free you are the product". What are all the proposed data centers going to do when Google AI continues to dominate search and guidance, Amazon consumer purchasing and logistics, Nvidia multimedia IC development and other giants with massive head starts. I am sure there are some LLM niches left but they will fill up fast and then get gobbled up in acquisitions by bigger fish. New AI will flourish with private data control of automation, inspection, diagnosis, energy efficiency, and process, machine, and vehicle control all of which require additional tech and development time and money. LLMs are the low hanging fruit.

 

[AA63]

So funny about AI, a lot of people are in an uproar about its accuracy when everyone knows "you get what you pay for". Again "if the app is free you are the product". What are all the proposed data centers going to do when Google AI continues to dominate search and guidance, Amazon consumer purchasing and logistics, Nvidia multimedia IC development and other giants with massive head starts. I am sure there are some LLM niches left but they will fill up fast and then get gobbled up in acquisitions by bigger fish. New AI will flourish with private data control of automation, inspection, diagnosis, energy efficiency, and process, machine, and vehicle control all of which require additional tech and development time and money. LLMs are the low hanging fruit.

This.

Current LLMs are a proof of concept. What is going to happen within a 5yr span is that the awkwardness of them being "advanced chatbots" will completely go away - the tech will just be integrated into apps - and it is in here that lies the next frontiers... who and how will the tech be implemented.

We've been here before. Apple put "a face" on computers changing them from the geeky awkward devices to ones that are accessible where an end user does not have to think about, nor care, what is going on in the background. They then did the same for mp3 players, phones, etc. This was a fundamental change - and it was not without tech, or finance company resistance. It took well over a decade to catch hold and become the way that all companies structured their development (both hardware and software). In more recent times, Tesla did the same, and I suspect the same will happen with AI in an even shorter time.

But I am not convinced yet that it will given the current geopolitical unrest and monstrously excessive greed in our current state of affairs. It is going to take a massive governmental cooperative effort (not likely) to agree to standard terms, and/or (more likely) a company like Apple to implement it invisibly given that they've already established their walled garden. Interesting times we live in.

 

[Timcognito]

Current LLMs are a proof of concept. What is going to happen within a 5yr span is that the awkwardness of them being "advanced chatbots" will completely go away - the tech will just be integrated into apps - and it is in here that lies the next frontiers... who and how will the tech be implemented.

Just like the dot com era, AI will just be the norm like cell phones, and internet interaction with many white collar jobs disappearing to robots and small VC funded companies will be write offs. As computing power increases along with the distillation of existing data, emergent proprietary databases AI will be made essential with large interdependent data crunched in real time for immediate use in applications and automation that function in real time. Its general use by the public will go the way of satellite navigation, purchasing, and screen based entertainment, that will be ubiquitous and taken for granted, not magic or scourge. The use of data centers is unclear without proprietary data, applications or technology.

 

[Petrushka]

So funny about AI, a lot of people are in an uproar about its accuracy when everyone knows "you get what you pay for". Again "if the app is free you are the product". What are all the proposed data centers going to do when Google AI continues to dominate search and guidance, Amazon consumer purchasing and logistics, Nvidia multimedia IC development and other giants with massive head starts. I am sure there are some LLM niches left but they will fill up fast and then get gobbled up in acquisitions by bigger fish. New AI will flourish with private data control of automation, inspection, diagnosis, energy efficiency, and process, machine, and vehicle control all of which require additional tech and development time and money. LLMs are the low hanging fruit.

I haven’t used google search in a year, and while I buy from Amazon, I don’t use their search.

The winners are not known.

 

[Salt]

And after this is established let come the blackout over 7 days ...