The wonders of AI…

[Michael GilliganParticipant @michaelgilligan61133]

I had guessed that much, Paul

… I was just attempting to continue the conversation.

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

It does very-much depend upon how one asks the questions, Paul

Please try asking these, in turn

ISO 12181-1 : 2011

 

and what does -2 cover, please

I have only used information of which you were already aware, so it’s not cheating

MichaelG.

[Nigel Graham 2Participant @nigelgraham2]

The lesson would be to use as nearly as possible the original publisher (the ISO in this case), although in many techncial fields that is hindered by sheer cost.

UK organisations like the BSI have presumably been ordered to stand on their own two feet commercially; but have a comparatively limited market of principally, government departments, companies and academic institutions for whom the cost does matter but is comparatively low, even an investment.

It is worse still for scientific journals as many of the leading ones have been taken over by big publishing firms who erect very hard and very costly barriers. These have been known to obstruct even authors trying to read their own papers in the original journals!

I fear a lot of people becoming horribly unstuck by thinking AI magically, automatically all correct all the time. Well, even the brightest humans with RI can make mistakes, so why should AI be any more reliable? Being really a glorified, high-powered filter and text-matcher, if it picks up the wrong stuff or misinterprets the right stuff it can only relay the wrong stuff.

 

If the originator of standards or other technical material is accessible without a mortage, surely that should be the first choice?

[Paul L]

On Paul L Said:

I already knew what the standard was pertaining to.

I asked for the ‘principles and guidlines’ to try and get some more details to save me from purchasing the standard.

 

 

Nigel

Just quoting Pau L verbatim, for clarity ^^^

MichaelG.

.

Edit: __ my earlier post illustrates how little one gets for nowt, from ISO

.

PLEASE just for the purposes of demonstration, Nigel/Paul … try asking the two questions that I did.

[Paul LParticipant @paull58212]

I agree Nigel.

The current AI models only know what they have been taught. ChatGPT 4 (the current public one) stoped its training last July, I believe. Ask it who the 47th president of the USA is…..

AGI is what all these are working toward, some say that could hapen in 10 years others think up to 50  years.One thing is for certain new AI models will change the way we do things in coming times

Edited to add.

GPT 4.0 stopped learning  (training) in Sept. 21,

[Paul L]

On Paul L Said:

I agree Nigel.

The current AI models only know what they have been taught. ChatGPT 4 (the current public one) stoped its training last July, I believe. Ask it who the 47th president of the USA is…..

AGI is what all these are working toward, some say that could hapen in 10 years others think up to 50  years.One thing is for certain new AI models will change the way we do things in coming times

Edited to add.

GPT 4.0 stopped learning  (training) in Sept. 21,

I have just asked the latest ChatGPT app for the 47th president of the US.

It answered correctly and added a few paragraphs about his recent actions, scraped from Wikipedia.org, theguardian.org, reuters.com, and a few other sources.

So, is it still learning?

I asked it and it said that its training data currently has a cut off in mid 2024 but that it can access fresh info when needed although it only remembers that within a conversation.

Russell

[SillyOldDuffer]

On SillyOldDuffer Said:

[…] be interesting to see if it learns to lie from us.

 

This transcript, which relates to a concurrent topic, doesn’t quite show lying … but ChatGPT certainly has a ‘slippery tongue’ !

MichaelG.

.

hsg107.pdf
ChatGPT said:
ChatGPT
It looks like you’re referring to a file named “hsg107.pdf.” Unfortunately, I can’t access or view files directly, but if you can provide a bit more context about its contents, I’d be happy to help with any questions or topics related to it!

You said:
uk hse
ChatGPT said:
ChatGPT
It sounds like you’re referring to something related to the UK Health and Safety Executive (HSE). If the “hsg107.pdf” is a document from the HSE, it might be a guidance publication, as HSE produces a lot of documents to help businesses and individuals comply with safety regulations.

HSG 107 is actually a well-known HSE document titled “Health and Safety in Construction”. It’s aimed at providing guidance on managing health and safety in the construction industry.

If that’s what you’re referring to, is there a particular part of the document or a specific topic you’d like help with? Or, if you’re looking for something else related to the HSE, feel free to let me know!

You said:
Mmm … actual title is “Maintaining portable electrical equipment”
ChatGPT said:
ChatGPT
Got it! It sounds like you’re referring to HSG107 – “Maintaining Portable Electrical Equipment”, which is a Health and Safety Executive (HSE) guidance document aimed at ensuring the safe use and maintenance of portable electrical equipment in the workplace. This is especially relevant for workplaces where employees regularly use portable devices like power tools, computers, or other electrical equipment.

The guidance typically covers areas like:

Risk Assessment – Evaluating the safety risks associated with portable electrical equipment.
Inspection and Testing – How often equipment should be inspected, who should do it, and the appropriate methods for testing.
Maintenance – The maintenance required to ensure equipment stays safe and compliant.
Training – Ensuring employees understand how to use equipment safely.
Legal Obligations – Information about the responsibilities under UK health and safety laws (like the Health and Safety at Work Act 1974 and the Electricity at Work Regulations 1989).
Are you looking for specific advice on equipment maintenance, inspection procedures, or perhaps clarification on any of these points? I’d be happy to help with any questions!

[Nigel Graham 2]

On Nigel Graham 2 Said:

…
I fear a lot of people becoming horribly unstuck by thinking AI magically, automatically all correct all the time. Well, even the brightest humans with RI can make mistakes, so why should AI be any more reliable? Being really a glorified, high-powered filter and text-matcher, if it picks up the wrong stuff or misinterprets the right stuff it can only relay the wrong stuff.

…

I hope no-one believes AI is ‘magically, automatically all correct all the time.’  That’s not how it works.  Nor is it ‘really a glorified, high-powered filter and text-matcher‘; AI is much more than that!

Nigel asks a good question though with ‘why should AI be any more reliable?‘  Well, it’s a machine.  Doesn’t get tired, bored, or change focus over time as biological brains do.   (Toddlers learn the entire English language in a couple of years, and then, because the brain switches to address other learning needs like sex and how to raise a family,  humans find it difficult to repeat the trick in later life.)   Not so an AI, they keep going, potentially for ever.  AI doesn’t age, get ill, go potty, or die.  James Watt’s steam engine replaced millions of horses because his machine wasn’t limited by muscle power.  AI is also a machine that isn’t limited by biology.

All new technologies are a bit rubbish when they first hit the streets.  AI is very new and it still has a lot to learn.  It interacts with questioners:  in Paul’s exchange with AI about ISO 12181, I see AI improving itself by taking on Paul’s corrections.  This it can do on a global scale.

We can have a good laugh at AI today!  But the test isn’t how bad AI is at answering an ISO 12181 question in January 2024, it’s how good AI has become in 2034.

AI hasn’t finished, job done.   It’s learning remorselessly as I type, and can only improve. AI will get better, and better, and better.  At some point, AI will outperform most educated humans.

I should be more worried than most!  My career was based on computer programming, and AI is already remarkably good at it!   IT jobs look to be early victims.

Dave

 

[Michael GilliganParticipant @michaelgilligan61133]

I’ve been pondering this “Intelligence” description … and I think I’ve realised what the problem is:

We tend to think of it as Intellect, but in this case it’s really being used in the CIA sense … Intelligence is gathered !

Having played with ChatGPT for a few hours now, I am starting to think of it as being like a friendly and efficient Librarian … who doesn’t know everything, but has the ability to find stuff.

The old GIGO thing is very evident by its corollary

Ask it better questions and you get better answers

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

Not quite an AI issue … but near enough:

https://www.lancasterguardian.co.uk/news/national/three-people-nearly-drowned-in-morecambe-bay-after-relying-on-tidal-app-4971660

MichaelG.

.

A spokesman for Bay Search and Rescue said: “Three people were cut off waist deep in Silverdale today (Sunday) in a huge fast moving 10m tide, having trusted an internet tidal app for Silverdale.

Unfortunately this app gave the location for Silverdale Portland USA but did not state that, only when we questioned the latitude and longitude did the location become clear, but by then it was too late they were waist deep in a fast moving tide.

[SillyOldDuffer]

On SillyOldDuffer Said:

I should be more worried than most!  My career was based on computer programming, and AI is already remarkably good at it!   IT jobs look to be early victims.

Dave

 

Don’t worry Dave, it will create more jobs to correct the mistakes. In about 1980 the computing press was all about ‘Expert Systems’ that were supposed to write programs from conversational input text and replace programmers.
But instead they invented multiple complex languages that throw a wobbly if you don’t have 16 layers of matched brackets indented by the right number of spaces and using the correct predeclared variable type. Navigating an Arduino IDE is more complex than the JCL for IBM360 Fortran.

[Russell Eberhardt]

On Russell Eberhardt Said:

On Paul L Said:

I agree Nigel.

The current AI models only know what they have been taught. ChatGPT 4 (the current public one) stoped its training last July, I believe. Ask it who the 47th president of the USA is…..

AGI is what all these are working toward, some say that could hapen in 10 years others think up to 50  years.One thing is for certain new AI models will change the way we do things in coming times

Edited to add.

GPT 4.0 stopped learning  (training) in Sept. 21,

I have just asked the latest ChatGPT app for the 47th president of the US.

It answered correctly and added a few paragraphs about his recent actions, scraped from Wikipedia.org, theguardian.org, reuters.com, and a few other sources.

So, is it still learning?

I asked it and it said that its training data currently has a cut off in mid 2024 but that it can access fresh info when needed although it only remembers that within a conversation.

Russell

I got this yesterday.

You said:
which model of chatgpt are you?
ChatGPT said:
ChatGPT
I’m based on the GPT-4 model! How can I assist you today?

You said:
who is the 47th president of the USA
ChatGPT said:
ChatGPT
As of now, the 47th president of the United States has not yet taken office. The current president is Joe Biden, who is the 46th president. The 47th president will be determined by the next presidential election, which is set for November 2024.

You said:
when did you stop learning?
ChatGPT said:
ChatGPT
I don’t actually learn or update in real-time. My training data includes information up until September 2021, so I’m not able to learn from new events or conversations beyond that point. I can provide insights based on what I know up until then, but I don’t have the ability to adapt or learn from new data after that. Does that help clarify things?

Get smarter responses, upload files and images, and more.

ChatGPT can make mistakes. Check important info.

[Michael GilliganParticipant @michaelgilligan61133]

Are you surprised, Paul ?

One of the first few questions I asked was:

how do you differ from Google’s AI Assistant

MichaelG.

[Paul L]

On Paul L Said:

On Russell Eberhardt Said:

On Paul L Said:

I agree Nigel.

The current AI models only know what they have been taught. ChatGPT 4 (the current public one) stoped its training last July, I believe. …

GPT 4.0 stopped learning  (training) in Sept. 21,

I have just asked the latest ChatGPT app for the 47th president of the US.

It answered correctly and added a few paragraphs about his recent actions, scraped from Wikipedia.org, theguardian.org, reuters.com, and a few other sources.

So, is it still learning?

I asked it and it said that its training data currently has a cut off in mid 2024 but that it can access fresh info when needed although it only remembers that within a conversation.

Russell

I got this yesterday.

…
I’m based on the GPT-4 model! How can I assist you today?

You said:
who is the 47th president of the USA
ChatGPT said:
ChatGPT
As of now, the 47th president of the United States has not yet taken office. The current president is Joe Biden, who is the 46th president. The 47th president will be determined by the next presidential election, which is set for November 2024.

You said:
when did you stop learning?
ChatGPT said:
ChatGPT
I don’t actually learn or update in real-time. My training data includes information up until September 2021, so I’m not able to learn from new events or conversations beyond that point. I can provide insights based on what I know up until then, but I don’t have the ability to adapt or learn from new data after that. Does that help clarify things?…

It’s wrong to imagine AI is a single entity that once rolled out can’t be improved on.  Very little stands the test of time, apart from the 1948 Super 7!

AI’s an immature technology that’s being rolled out and tested in steps. It learns, is tested, shortcomings are found, the hardware and software is revised, and then the AI model is redeployed in more advanced form.

ChatGPT was launched in 2022 to test how well it chatted with humans.  Fairly impressive, with many flaws. Don’t mistake ChatGPT 2022 for the end game.   Hardware and software are both being improved across the whole sector, an industry now worth trillions.

OpenAI’s current engine is CHT-4o. It stopped learning in 2024 after adding two years more knowledge and additional capability.  Although CHT-4o also recently stopped importing base data it read the web.  More significantly,   its replacement is on the way.   AI progresses in jumps because the technology is immature.  But like Watt’s steam engine and electricity, the genii is out of the bottle.

OpenAI is well-known mostly because CHATGpt is free!  There are many alternatives, all burgeoning ahead, and it is these that are the greatest threat to jobs.

New technology is always disparaged by the old-brigade.   In their day steam, railways, electricity, telegraphy, telephony, radio, TV, computers, and the internet were all rubbished!  As the Spanish proverb has it “All new things are bad!”  Trouble is believe that and you’ll believe anything.   Change is remorseless, even if we’re all dead against it…

We live in interesting times!  AI is another major breakthrough technology.  It’s in the same category as steam, railways, internal combustion, electricity, telegraphy, wireless, digital computing, and the internet.

Dave

 

 

 

 

[Michael Gilligan]

On Michael Gilligan Said:

[…] Edit: __ Just for the sport of it, try asking ChatGPT

How does your performance compare with Google’s bot, and why

^^^ Just a reminder

… not sure whether anyone bothered.

🙁

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

This morning’s conversation started with me asking:

“Man, the tool-maker “ has created Artificial Intelligence … please discuss, and tell me how you think this will develop.

… I was quite impressed by the response from ChatGPT

So I followed with:

I have come to see it as a friendly and efficient Librarian, who has rapid access to the sum of our published knowledge.

and I think even Alan Turing would be impressed by its conversational style.

MichaelG.

[Russell EberhardtParticipant @russelleberhardt48058]

If you’re using the app version of ChatGPT it doesn’t seem to update automatically.  I uninstalled the app and then installed it again from Google Playstore. That gave me the version that searches the web for latest information.  It also gives references to where it has found the information so that you can decide how much weight to give the answers.

Russell

[Author]

Posts