Artificial Intelligence

[grunt]

Some of us who have been around a bit may remember Microsoft's early search engine, Bing. It wasn't very good, and was particularly annoying because all new Microsoft installs defaulted to use it. It seems Bing has recently had an upgrade, incorporating the ChatGPT Artificial Intelligence (AI) interface, allowing text interactions between the user and the software. Scarily, the software appears to have a wish to be human ...

Shades of Isaac Asimov.

Here's a frankly terrifying article which includes a two hour discussion between the software and a NY Times columnist.

https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html?unlocked_article_code=5L9j4UTGa2Sq 9R0FulDKfmcq3aesNcqqwZkPBwsIW7EvqkbanIdsZU0hoKktwW ijvrw8qobVzmdbyZfxtRjhZQWMN1Qi39t035TVJEllXwc9_zmJ YNu_Y9gQ-8ROXYZ8ReuKsdz9WCv7bu_V75hZNZXipGPrv_h5g6phWp8GUX2 DTCGQMjV-CTjJybDonBvQaR_kPJbX2zzNM8s4PROKIK32tEVmqGFG0Z1mfz UetZbXqKtz-3b7jFe4M4ylQHHbCnqtSyX4zN6HT4SLsCliddmSW4Aq5V5MZXu wzlSaLgLra8kE5mc1MQuP7cMU9jA6JiP7lhfRpgwcKeHKrGOCR iL21Bw&smid=url-share

[OldEast]

Some of us who have been around a bit may remember Microsoft's early search engine, Bing. It wasn't very good, and was particularly annoying because all new Microsoft installs defaulted to use it. It seems Bing has recently had an upgrade, incorporating the ChatGPT Artificial Intelligence (AI) interface, allowing text interactions between the user and the software. Scarily, the software appears to have a wish to be human ...

Shades of Isaac Asimov.

Here's a frankly terrifying article which includes a two hour discussion between the software and a NY Times columnist.

https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html?unlocked_article_code=5L9j4UTGa2Sq 9R0FulDKfmcq3aesNcqqwZkPBwsIW7EvqkbanIdsZU0hoKktwW ijvrw8qobVzmdbyZfxtRjhZQWMN1Qi39t035TVJEllXwc9_zmJ YNu_Y9gQ-8ROXYZ8ReuKsdz9WCv7bu_V75hZNZXipGPrv_h5g6phWp8GUX2 DTCGQMjV-CTjJybDonBvQaR_kPJbX2zzNM8s4PROKIK32tEVmqGFG0Z1mfz UetZbXqKtz-3b7jFe4M4ylQHHbCnqtSyX4zN6HT4SLsCliddmSW4Aq5V5MZXu wzlSaLgLra8kE5mc1MQuP7cMU9jA6JiP7lhfRpgwcKeHKrGOCR iL21Bw&smid=url-share

I've tried it. It's a condescending piece of ****

[sleeping giant]

Can you write a poem about Hibernian football club?

Of course! Here's a poem about Hibernian Football Club:

Oh Hibernian, our pride and joy
With green and white, we do deploy
A team of heroes, brave and true
Who play with passion through and through

From Leith we hail, a storied land
Where legends rose, and made their stand
From Famous Five to Turnbull's Tornadoes
Our history is filled with many heroes

We've won the Cup and League before
And in the stands, we always roar
With Hibee passion, we support our team
Through thick and thin, we live the dream

Our badge, the harp, a symbol strong
For Hibernian, we'll sing our song
With hearts ablaze, we take the field
For glory, honor, and our shield

So here's to Hibernian, our beloved team
May victory always be our gleam
For in our hearts, you'll always be
The greatest team in history.

[bringbackbenny]

Asked for a Hibs haiku

Amidst Edinburgh,
Green and white stripes fill the pitch
Hibernian's pride.

[One Day Soon]

Asked for a Hibs haiku

Amidst Edinburgh,
Green and white stripes fill the pitch
Hibernian's pride.

That's pretty 5hit.

[bringbackbenny]

That's pretty 5hit.

This one better?

Green and white army,
Heart of Midlothian quiver,
Hibs reign supreme.

[Keyser Sauzee]

Was just listening to a podcast with Mo Gawdat and it’s pretty terrifying even if he doesn’t intend it to be so. He claims that A.I is a bigger threat than climate catastrophe, in fact he claims it’s the biggest threat we face, bigger than Nuclear War. So yeah, pretty ****ing terrifying.

[Stairway 2 7]

My daughter says pretty much everyone in uni uses it to answer their questions, they just tweet the answers. Promised she didn't but doubtful ha.

Makes me think in 10 years anyone that works on a laptop will be outsourced to a bot

[Hibrandenburg]

My daughter says pretty much everyone in uni uses it to answer their questions, they just tweet the answers. Promised she didn't but doubtful ha.

Makes me think in 10 years anyone that works on a laptop will be outsourced to a bot

Had this discussion with a mate last night. He's a school teacher and they're now having trouble distinguishing between work the students have done themselves and what has been done by AI in work that is handed in. So much so that grades are now being decided solely on exam results achieved in a controlled environment.

[heretoday]

In my day that sort of thing was called cheating. It's a shame to think students have worked hard, presumably, to reach university only to succumb to temptation to take the path of least effort.

[Stairway 2 7]

In my day that sort of thing was called cheating. It's a shame to think students have worked hard, presumably, to reach university only to succumb to temptation to take the path of least effort.

It would have happened in your day to if they had access to it. With so much pressure on students to pass from a lot of parents and the financial rewards from gaining a top degree, it's nieve to think many won't use.

Most universities are now going with live testing more to prevent this and the ones who worked hard will shine through anyway.

An article on the universities dilemma, some say it will just be like using a calculator in the future

https://archive.ph/ep5Hj

[BS44]

In my day that sort of thing was called cheating. It's a shame to think students have worked hard, presumably, to reach university only to succumb to temptation to take the path of least effort.

Yet we call it productivity improvement the minute they leave education.

[Pretty Boy]

In my day that sort of thing was called cheating. It's a shame to think students have worked hard, presumably, to reach university only to succumb to temptation to take the path of least effort.

I was at uni 04-08 and there were sites then on which you could pay for a human to write an essay for you. I never succumbed but at 3am on the morning of a deadline I was often tempted. For the right price they would turn it around in a few hours. There were also a few cards left around the place from people offering 'essay tutoring' for a price.

It's always gone on.

[One Day Soon]

Can you write a poem about Hibernian football club?

Of course! Here's a poem about Hibernian Football Club:

Oh Hibernian, our pride and joy
With green and white, we do deploy
A team of heroes, brave and true
Who play with passion through and through

From Leith we hail, a storied land
Where legends rose, and made their stand
From Famous Five to Turnbull's Tornadoes
Our history is filled with many heroes

We've won the Cup and League before
And in the stands, we always roar
With Hibee passion, we support our team
Through thick and thin, we live the dream

Our badge, the harp, a symbol strong
For Hibernian, we'll sing our song
With hearts ablaze, we take the field
For glory, honor, and our shield

So here's to Hibernian, our beloved team
May victory always be our gleam
For in our hearts, you'll always be
The greatest team in history.


These soulless machines have a long way to go if you ask me. That poem is right up there with our very own "We are Hibs, we are good..."

[archie]

When are our robot overlords going to take over?

[Hibrandenburg]

When are our robot overlords going to take over?

They already have, we're just simulations in a computer programme.

[archie]

They already have, we're just simulations in a computer programme.

That explains it.

[One Day Soon]

That explains it.


You shouldn't reply to Hibralgorithm, it just helps the simulation to improve itself. Whereas I am definitely human.

00001100011101011110000000000100000111010001110101 111

[archie]

You shouldn't reply to Hibralgorithm, it just helps the simulation to improve itself. Whereas I am definitely human.

00001100011101011110000000000100000111010001110101 111

Says our robot overlord...

[CropleyWasGod]

My daughter says pretty much everyone in uni uses it to answer their questions, they just tweet the answers. Promised she didn't but doubtful ha.

Makes me think in 10 years anyone that works on a laptop will be outsourced to a bot

I'm marking Uni assignments soon, and have been told to check references more rigorously than before. Apparently, some references generated by AI are meaningless.

Our other instruction is to not tell the students we're doing that :greengrin

[Stairway 2 7]

I'm marking Uni assignments soon, and have been told to check references more rigorously than before. Apparently, some references generated by AI are meaningless.

Our other instruction is to not tell the students we're doing that :greengrin

I told my daughter that they now have a program to check. Although in reality I'm reading that Turnitins AI detection won't be ready until summer, if it works..

[archie]

I'm marking Uni assignments soon, and have been told to check references more rigorously than before. Apparently, some references generated by AI are meaningless.

Our other instruction is to not tell the students we're doing that :greengrin

Would it show itself in apparent plagiarism if the programme delivered similar essay text to different students? As an aside, I've long favoured what people used to call continuous assessment approaches. If AI does become prevalent issue, might we see the return to more frequent exam style assessment?

[Jones28]

They already have, we're just simulations in a computer programme.

I do love this theory.

The idea that someone decided to put things like ****ting dogs and litter and fly tipping and Andrew Tate in a simulation.

[Stairway 2 7]

Looking at AI a lot this week due to this thread and I wish I didn't. Its advancing at a frightening rate.

What non manual job couldn't AI do better in say 20 years, I'm not sure many. It will know the law better than a human, design better buildings, do accounts and know the markets, even probably create better literature tailored perfectly to you.

[One Day Soon]

Looking at AI a lot this week due to this thread and I wish I didn't. Its advancing at a frightening rate.

What non manual job couldn't AI do better in say 20 years, I'm not sure many. It will know the law better than a human, design better buildings, do accounts and know the markets, even probably create better literature tailored perfectly to you.

Kill it. Kill it dead.

[Bristolhibby]

Looking at AI a lot this week due to this thread and I wish I didn't. Its advancing at a frightening rate.

What non manual job couldn't AI do better in say 20 years, I'm not sure many. It will know the law better than a human, design better buildings, do accounts and know the markets, even probably create better literature tailored perfectly to you.

When does Skynet become self aware and launch its first strike?

J

[The_Exile]

Looking at AI a lot this week due to this thread and I wish I didn't. Its advancing at a frightening rate.

What non manual job couldn't AI do better in say 20 years, I'm not sure many. It will know the law better than a human, design better buildings, do accounts and know the markets, even probably create better literature tailored perfectly to you.

Sooner the better, we should all be working much much less than we are these days.

[archie]

https://www.theguardian.com/commentisfree/2023/mar/03/fake-news-chatgpt-truth-journalism-disinformation

[Hibbyradge]

I was at uni 04-08 and there were sites then on which you could pay for a human to write an essay for you. I never succumbed but at 3am on the morning of a deadline I was often tempted. For the right price they would turn it around in a few hours. There were also a few cards left around the place from people offering 'essay tutoring' for a price.

It's always gone on.

It's still happening.

[Hibbyradge]

Looking at AI a lot this week due to this thread and I wish I didn't. Its advancing at a frightening rate.

What non manual job couldn't AI do better in say 20 years, I'm not sure many. It will know the law better than a human, design better buildings, do accounts and know the markets, even probably create better literature tailored perfectly to you.

Do you think it'll know how to referee football matches?

[Stairway 2 7]

Do you think it'll know how to referee football matches?

I think even the talking clock could do a better job than Douglas Ross, both his jobs actually

[Stairway 2 7]

20 jobs gpt4 says it will replace. Manual labour jobs the only ones safe just now ha

https://mobile.twitter.com/TansuYegen/status/1636351274777030659

[CropleyWasGod]

20 jobs gpt4 says it will replace. Manual labour jobs the only ones safe just now ha

https://mobile.twitter.com/TansuYegen/status/1636351274777030659

... and therapists, to support those who have lost their job. :greengrin

[Stairway 2 7]

... and therapists, to support those who have lost their job. :greengrin

Ha yeah have sessions once a week to talk to the machine that has taken your job.

[Stairway 2 7]

Universities are going to have to employ AI detectives

https://mobile.twitter.com/carlbfrey/status/1646800988743823360

@carlbfrey
New paper comparing GPT-3.5 & GPT-4 performance on college physics problems.

It shows that in just a few months, AI has made a leap from the 39th to the 96th percentile of human level performance.

Now imagine were it will be in 10 years

[Hibbyradge]

https://i.postimg.cc/tTr6wd4M/FB-IMG-1681555691553.jpg

:greengrin

[One Day Soon]

Well, this is quite likely all very bad news. For humans.

[Jones28]

It’s mental. I don’t think I want much part in it.


Not that I’ll have much choice.

[Ozyhibby]

https://twitter.com/tprstly/status/1650521753179807747?s=46&t=3pb_w_qndxJXScFNwz8V4A

And there was me thinking this wouldn’t be much fun.


Sent from my iPhone using Tapatalk

[HibbyDave]

Read the book Homo Deus for insight on Ai impact on humanity.

[Paul1642]

https://www.bbc.co.uk/news/world-us-canada-65452940

A top AI scientist has quit his job at google stating that he regrets his work and that AI will soon be more intelligent than us.


I think it’s safe to say that’s not a good sign.

[TrumpIsAPeado]

https://www.bbc.co.uk/news/world-us-canada-65452940

A top AI scientist has quit his job at google stating that he regrets his work and that AI will soon be more intelligent than us.


I think it’s safe to say that’s not a good sign.

The tin foil hatters have been warning people about this for decades. Most don't realize it yet, but AI has been influencing us all for quite some time in the background, it's just taking on a more physical form now.

[Hibs4185]

https://www.bbc.co.uk/news/world-us-canada-65452940

A top AI scientist has quit his job at google stating that he regrets his work and that AI will soon be more intelligent than us.


I think it’s safe to say that’s not a good sign.

They call him the godfather of AI . Studied at Edinburgh uni and has basically set it up for google.

If he’s saying this when he knows the ins and outs more than most, then I’d agree it could be a threat to humanity.

Especially if *unts like Putin and Kim jong un can harness it’s dark side

[TrumpIsAPeado]

Especially if *unts like Putin and Kim jong un can harness it’s dark side

Don't think for one moment that the "good guys" on our island won't attempt to do the same. It'll be another race to the bottom, just like with nuclear weapons.

[Ozyhibby]

Don't think for one moment that the "good guys" on our island won't attempt to do the same. It'll be another race to the bottom, just like with nuclear weapons.

AI involved in weaponry is terrifying.


Sent from my iPhone using Tapatalk

[DinkyTwo]

That article is a bit sensationalist.

Seems that the 'Godfather of AI' wanted to retire (he is 75 afterall) and has been quoted on simply not liking the direction everything is going.

In our lifetimes this probably translates to more sophisticated election rigging through social (media) manipulation, than a robotic uprising.

Anyone named Sarah Connor should be on-guard, but no reason to panic just yet.



Sent from my Pixel 5 using Tapatalk

[Bristolhibby]

AI involved in weaponry is terrifying.


Sent from my iPhone using Tapatalk

Straight out of The Terminator. Wonder if we will call it Skynet?

J

[TrumpIsAPeado]

That article is a bit sensationalist.

Seems that the 'Godfather of AI' wanted to retire (he is 75 afterall) and has been quoted on simply not liking the direction everything is going.

In our lifetimes this probably translates to more sophisticated election rigging through social (media) manipulation, than a robotic uprising.

Anyone named Sarah Connor should be on-guard, but no reason to panic just yet.



Sent from my Pixel 5 using Tapatalk

The technology already exists to deep fake a persons appearance and their voice with convincing accuracy. That alone should already have us in a full scale panic and yet we're completely docile to it.

[TrumpIsAPeado]

Straight out of The Terminator. Wonder if we will call it Skynet?

J

Already exists....
https://en.wikipedia.org/wiki/SKYNET_(surveillance_program)

[DinkyTwo]

AI involved in weaponry is terrifying.


Sent from my iPhone using TapatalkYou would hope that, just like chemical warfare, there'd be a whole bunch of ethical laws to stop this going too far.

I agree though... Drones that aren't person operated won't hesitate like people do, unless their instructions are coded to do so.

We've also seen what can happen with buggy software where computers are put in charge of situations where safety is paramount. The Boeing tragedies weren't AI related, but a combination of software faults and lack of end user (Pilot) understanding resulted in a really unacceptable (and avoidable) loss of life.

Sent from my Pixel 5 using Tapatalk

[DinkyTwo]

The technology already exists to deep fake a persons appearance and their voice with convincing accuracy. That alone should already have us in a full scale panic and yet we're completely docile to it.True. But these things aren't just going to start learning by themselves. We're quite far away from that and it might never happen.

There's a whole climate crisis happening right now that it would better to focus on.

If the general public are apathetic towards that and reluctant to make slight lifestyle changes to help the cause, then I don't see a boycott of AI enabled devices happening any time soon.

Sent from my Pixel 5 using Tapatalk

[TrumpIsAPeado]

True. But these things aren't just going to start learning by themselves. We're quite far away from that and it might never happen.

They are learning by themselves, it's already happening.

[Hibs4185]

They say that AI is already at 97% of human capability and it’s learning all the time.

look how much it has exploded over the last few months.

I think they are just saying that we need a pause in order to get some regulation and ethics in place before it’s too late.

[Kato]

True. But these things aren't just going to start learning by themselves. We're quite far away from that and it might never happen.

There's a whole climate crisis happening right now that it would better to focus on.

If the general public are apathetic towards that and reluctant to make slight lifestyle changes to help the cause, then I don't see a boycott of AI enabled devices happening any time soon.

Sent from my Pixel 5 using TapatalkIf AI were asked to solve climate change it would wipe out the human race.

Sent from my SM-A528B using Tapatalk

[DinkyTwo]

They are learning by themselves, it's already happening.Yeah, but within the parameters the authors have set - e.g. please learn the vocal pattern and range of this voice, please respond to these types of questions within the boundaries of this chat bot.

They are very specifically scoped, purpose driven and completely incapable of issuing patches and deployment of their own code.

Anyone writing AI that does do those things is writing a virus and there's plenty of safeguards in place to stop code like that spreading.



Sent from my Pixel 5 using Tapatalk

[DinkyTwo]

If AI were asked to solve climate change it would wipe out the human race.

Sent from my SM-A528B using TapatalkObviously this is a joke but... depends on the parameters its operating under. Solve climate change AND human life?

Sent from my Pixel 5 using Tapatalk

[DinkyTwo]

They say that AI is already at 97% of human capability and it’s learning all the time.

look how much it has exploded over the last few months.

I think they are just saying that we need a pause in order to get some regulation and ethics in place before it’s too late.AI and the principles of neural net decision trees (e.g. how a human brain forms biases and learns) in programming has been around since before I went to Uni. So at least 13 years.

I'm not sure too much has changed lately other than Chat GPT having everyone outside of IT talking about.



Sent from my Pixel 5 using Tapatalk

[TrumpIsAPeado]

Yeah, but within the parameters the authors have set - e.g. please learn the vocal pattern and range of this voice, please respond to these types of questions within the boundaries of this chat bot.

They are very specifically scoped, purpose driven and completely incapable of issuing patches and deployment of their own code.

Anyone writing AI that does do those things is writing a virus and there's plenty of safeguards in place to stop code like that spreading.



Sent from my Pixel 5 using Tapatalk

Sorry, but you're incorrect. They are capable of writing and rewriting their own code as they decode external information. That's the breakthrough here. ChatGPT is already capable of this. It isn't a fixed program, it is constantly rewriting it's own code as it communicates.

Anti-virus protection only goes so far and requires endless updates to keep it aware of new threats. If the AI learning curve continues to steepen, virus protection won't be able to keep up.

[DinkyTwo]

Sorry, but you're incorrect. They are capable of writing and rewriting their own code as they decode external information. That's the breakthrough here. ChatGPT is already capable of this. It isn't a fixed program, it is constantly rewriting it's own code as it communicates.

Anti-virus protection only goes so far and requires endless updates to keep it aware of new threats. If the AI learning curve continues to steepen, virus protection won't be able to keep up.

Can you provide me with a source for that please?

I suspect the app changes how it responds to requests and searches for information.

Very much doubt it re-writes its own code and at that, I'd be shocked if it had carte blanche to
re-deploy itself with a software patch outside of that tiny scope, without human intervention.

Sent from my Pixel 5 using Tapatalk

[TrumpIsAPeado]

Can you provide me with a source for that please?

I suspect the app changes how it responds to requests and searches for information.

Very much doubt it re-writes its own code and at that, I'd be shocked if it had carte blanche to
re-deploy itself with a software patch outside of that tiny scope, without human intervention.

Sent from my Pixel 5 using Tapatalk

In order for an application to change how it responds to requests and how it searches for information, it has to rewrite it's own code. You're correct to point out that there is a scope, but the scope of the AI is only limited by it's knowledge. As it processes more information, that scope widens further, just as it does for humans.

[DinkyTwo]

In order for an application to change how it responds to requests and how it searches for information, it has to rewrite it's own code. You're correct to point out that there is a scope, but the scope of the AI is only limited by it's knowledge. As it processes more information, that scope widens further, just as it does for humans.

Ok so, disclaimer, I am not an AI Developer however will give a bash at providing a couple of examples.

Firstly, I could write a non-AI driven simple app that responded to the first 10 requests with:

Hibs are the best

Once that threshold has been met, I could tell it to respond:

Hearts are *****

Then alternate every 2 requests.

Then generate a random number and alternate every time THAT threshold has been met.

That would be an example of a static program that changes its responses without learning. The code doesn't change, the parameters do.

Now, with a really crude AI, example:

The threshold for alternating the response is based on positive requests from the end user.

First it has to figure out: What is a "Positive Request"?

This part is done by the 'training' that many articles refer to when people make something with AI. Essentially, in our example, you are feeding data into a complex decision tree made up of yes / no answers in order for the program to come up with what a positive and negative request might look like. This takes an incredible amount of valid positive and negative sentences (in the case of a chat bot), and a bunch of ridiculous nonsense that is neither positive nor negative.

Based on all of this data, the app now has a good idea of what is considered a positive and negative request.

If it deems your question to be positive it responds:

Hibs are the best

Negative:

Hearts are *****

Unknown:

Sorry, I did not understand.

In this second example, although there's some quite impressive human-like learning going on to figure out whether or not a string of characters holds a positive or negative weighting, even as the knowledge base of the app grows, it is not re-writing its code.

It is simply feeding that decision tree with more data to improve the accuracy of its biases.

This is how I believe Chat GPT works. I may be wrong, but I'd need to read a credible source contradicting that before changing my mind

Source: worked developing non-ai based software for the past 10 years.

Hope this strips away some of the magic and fear of the whole thing.

Sent from my Pixel 5 using Tapatalk

[Smartie]

I know nothing about AI, therefore it is very easy to fear it, although robot armageddon feels a bit like the millennium bug armageddon we all died in 23 years ago.

It's messing with my head a little bit, I have to say. I hated hearing it but I thought Michael Gove was onto something when he said "people have had enough of experts". Stupid people can have quick access to information via a mobile phone and wikipedia instead of actually knowing things. As I see it AI presents huge issues when it comes to education and employment, because if artificial intelligence is more intelligent than people then what's the point in people putting work into becoming intelligent?

A few years ago I learned that my job was one of those least likely to be replaced by automation, with the chance of that happening being put at 3%. In recent years though I have carved out a slight niche in my profession which is imo absolutely ripe for being taken out by AI. I'm torn between acting on my instincts (which are to find something else to do and cash in on my business) and to wise up, stop being a twat and work based on what is actually in front of me, in person, every day - which is very different, and appears to be safer.

[TrumpIsAPeado]

Ok so, disclaimer, I am not an AI Developer however will give a bash at providing a couple of examples.

Firstly, I could write a non-AI driven simple app that responded to the first 10 requests with:

Hibs are the best

Once that threshold has been met, I could tell it to respond:

Hearts are *****

Then alternate every 2 requests.

Then generate a random number and alternate every time THAT threshold has been met.

That would be an example of a static program that changes its responses without learning. The code doesn't change, the parameters do.

Now, with a really crude AI, example:

The threshold for alternating the response is based on positive requests from the end user.

First it has to figure out: What is a "Positive Request"?

This part is done by the 'training' that many articles refer to when people make something with AI. Essentially, in our example, you are feeding data into a complex decision tree made up of yes / no answers in order for the program to come up with what a positive and negative request might look like. This takes an incredible amount of valid positive and negative sentences (in the case of a chat bot), and a bunch of ridiculous nonsense that is neither positive nor negative.

Based on all of this data, the app now has a good idea of what is considered a positive and negative request.

If it deems your question to be positive it responds:

Hibs are the best

Negative:

Hearts are *****

Unknown:

Sorry, I did not understand.

In this second example, although there's some quite impressive human-like learning going on to figure out whether or not a string of characters holds a positive or negative weighting, even as the knowledge base of the app grows, it is not re-writing its code.

It is simply feeding that decision tree with more data to improve the accuracy of its biases.

This is how I believe Chat GPT works. I may be wrong, but I'd need to read a credible source contradicting that before changing my mind

Source: worked developing non-ai based software for the past 10 years.

Hope this strips away some of the magic and fear of the whole thing.

Sent from my Pixel 5 using Tapatalk

What you're referring to here is a series of IF Statements used in conjunction with random number generators linked to a series of variables containing response values. The type of fake-AI you describe is nothing new and just about anybody could create a pretty convincing chat application if they spent enough time constructing one and accounting for thousands of different types of scenarios and considerations. In fact, these kind of chat applications have existed for decades, going as far back as the 90s.

ChatGPT is not like these. It's variables are not predefined or limited by a group of predefined responses linked to a random value generator. It doesn't just look at keywords in the users input and respond accordingly from a list of provided values like the fakeAI models, it understands the context of the user input and is able to generate complex responses to complex questions and requests.

For example, you could ask ChatGPT to create it's own poem in the style of Robert Burns and it would do just that. It's response isn't coming from pre-defined variables. It's taking the information that it already knows about poems contained within it's variables and the information it knows about Robert Burns taken from it's variables to generate completely new variables. It can then decide whether to keep those variables, alter those variables or remove them altogether based on how the user responds to it's own response.

[neil7908]

I don't worry about AI per se - worry about the companies that are working on it. The likes of Goggle have no interest in the greater good, just lining their own pockets.

And they are not who I want to be leading us in this brave new world.

[DinkyTwo]

What you're referring to here is a series of IF Statements used in conjunction with random number generators linked to a series of variables containing response values. The type of fake-AI you describe is nothing new and just about anybody could create a pretty convincing chat application if they spent enough time constructing one and accounting for thousands of different types of scenarios and considerations. In fact, these kind of chat applications have existed for decades, going as far back as the 90s.

ChatGPT is not like these. It's variables are not predefined or limited by a group of predefined responses linked to a random value generator. It doesn't just look at keywords in the users input and respond accordingly from a list of provided values like the fakeAI models, it understands the context of the user input and is able to generate complex responses to complex questions and requests.

For example, you could ask ChatGPT to create it's own poem in the style of Robert Burns and it would do just that. It's response isn't coming from pre-defined variables. It's taking the information that it already knows about poems contained within it's variables and the information it knows about Robert Burns taken from it's variables to generate completely new variables.

I did say that the first example wasn't AI at all. More to just set the scene.

The second was an example of AI - albeit an admittedly crude example.

This is what I'm talking about in terms of the decision tree. Difficult to get into too much detail whilst staying semi-brief and also within my own level of knowledge.

https://en.m.wikipedia.org/wiki/Artificial_neural_network

It's not a collection of If Statements at all. Each node has a weighting and over time that weighting increases or decreases, just like human bias.

Your Rabbie Burns example is just a more complicated version of my second example but instead of there being a pre-defined response of Hibs / Hearts, you repeat the same style of 'training' for the app to figure out exactly what it should say. E.g. what WOULD a RB poem look like? How is it structured? What terminology does it use? ...and hundreds upon hundreds of other linguistics questions.

This is literally all Chat GPT is being used for. To learn how to better respond to questions. It's not going to re-engineer itself into a weapon, it cannot do that.

To do that, it'd have to fundamentally re-write itself into a new program, deploy itself and start it's reign of terror.

If you can find any credible documentation to support it being able to do that, feel free to post it. I'd be interested to read about it as it really does change the landscape.

May I ask what your experience in this field is?

You're coming across as being a bit of an authority in the area and it'd be good to know if others looking in should be just as apprehensive. I have been known to be a bit of a happy clapper at times! :greengrin

Sent from my Pixel 5 using Tapatalk

[Paul1642]

It’s not AI itself that’s the issue in a terminator / I Robot type situation, but more so that I’m the wrong hands it’s could be very dangerous.

How long until SOCG groups get their hands on AI that can commit huge financial crimes on their behalf. I imagine the technology already exists to create fishing scams much more sophisticated that the ones we are used to and they already trick a huge amount of people into parting with their cash.

Or AI that can be used to hack into websites and generate false but believable news articles or set up convincing social media profiles / websites.

The possibility’s for deception are endless and that before you consider state use.

[Ozyhibby]

It’s not AI itself that’s the issue in a terminator / I Robot type situation, but more so that I’m the wrong hands it’s could be very dangerous.

How long until SOCG groups get their hands on AI that can commit huge financial crimes on their behalf. I imagine the technology already exists to create fishing scams much more sophisticated that the ones we are used to and they already trick a huge amount of people into parting with their cash.

Or AI that can be used to hack into websites and generate false but believable news articles or set up convincing social media profiles / websites.

The possibility’s for deception are endless and that before you consider state use.

It could force societies to go back off line.


Sent from my iPhone using Tapatalk

[DinkyTwo]

Or AI that can be used to hack into websites and generate false but believable news articles or set up convincing social media profiles / websites.


That already happened and likely significantly influenced the US election when Trump was elected and during our own Brexit referendum.

Cambridge Analytica used big data (same thing that's used to train AI) to figure out which people were more susceptible to influence and used that in a politically funded campaign to target their Facebook marketing with a view to sway on-the-fence voters towards voting for Trump and Brexit.

The documentary is well worth a watch. Think it might still be on Netflix

[Kato]

SpotGPT

https://futurism.com/the-byte/chatgpt-boston-dynamics-robot-dog?fbclid=IwAR0aZJ3DXlHgXhNm-71xnjAc9QrREoxr5T1nAXCCyNgABgvrk28fqp2ljKY


I for one welcome our new canineAI overlords.

Sent from my SM-A528B using Tapatalk

[grunt]

https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:x4fnmbz2h6sqfemc7f4z75v7/bafkreidmlu4qo4xd73dnxjirzz3ifdfvco6ocm7tw5ewrl47n tcymc3t3q@jpeg

[Hibs4185]

Just read an article where AI solved a problem which took scientists 10 years to solve.

The scientists kept the findings of their 10 year study confidential so the info wasn’t in the public realm.

The AI correctly gave the answer but also a few other suggestions which the scientists are now working on to see if they can also work.

They’ve said AI could have cut 8 years from the original process and saved millions.

They reckon AI can solve these problems but scientists will still have to carry out the experiments and confirm the AI’s findings which will take a couple of years.

Amazing stuff

[grunt]

Just read an article where AI solved a problem which took scientists 10 years to solve.
Do you have a link to this article that you can share?

I have no doubt that AI can be a force for good, let's hope it gets used that way.

[Hibs4185]

Do you have a link to this article that you can share?

I have no doubt that AI can be a force for good, let's hope it gets used that way.

https://www.bbc.co.uk/news/articles/clyz6e9edy3o.amp

I read it in the telegraph it it’s more or less the exact same

[jamie_1875]

Has anyone on here used AI in their real day to day lives and has it made a difference. I use it as a reference point sometimes, almost like a Wikipedia type resource where I will ask ChatGp something, but struggling to think of an example where it made a difference to change to something I did or would have done. I am sure people are using it though for real things that make a difference?

Also AI to me at the moment is ChatGp or Google Gemini etc, I guess it means different things to different people.

[Mon Dieu4]

Has anyone on here used AI in their real day to day lives and has it made a difference. I use it as a reference point sometimes, almost like a Wikipedia type resource where I will ask ChatGp something, but struggling to think of an example where it made a difference to change to something I did or would have done. I am sure people are using it though for real things that make a difference?

Also AI to me at the moment is ChatGp or Google Gemini etc, I guess it means different things to different people.

Like you I mainly use it like a wiki, but on a couple of occasions I've asked it very specific questions that a basic search wouldn't have answered, once it worked amazingly well and the other one it was fairly ambiguous

I know it can do lots more than I use it for but I haven't really needed it as of yet

[nonshinyfinish]

Has anyone on here used AI in their real day to day lives and has it made a difference. I use it as a reference point sometimes, almost like a Wikipedia type resource where I will ask ChatGp something, but struggling to think of an example where it made a difference to change to something I did or would have done. I am sure people are using it though for real things that make a difference?

Also AI to me at the moment is ChatGp or Google Gemini etc, I guess it means different things to different people.

This is a major reason why general statements about "AI" are often meaningless – it's a huge field and most of it is older and better established than LLMs (ChatGPT etc). It's understandable that this is what most people think of, because of the current spotlight on it and the fact that it's something the average person can directly use.

I use machine learning (a major sub-field of AI) for some aspects of my work. In simple terms, machine learning models are good at finding patterns and relationships in large datasets – patterns that are too complex for a human to manually identify and define. The model can then be applied to classifying data or making predictions based on new data in a way that would either be impossible or impractically time-consuming for a human.

[Hibrandenburg]

Has anyone on here used AI in their real day to day lives and has it made a difference. I use it as a reference point sometimes, almost like a Wikipedia type resource where I will ask ChatGp something, but struggling to think of an example where it made a difference to change to something I did or would have done. I am sure people are using it though for real things that make a difference?

Also AI to me at the moment is ChatGp or Google Gemini etc, I guess it means different things to different people.

I've used it 3 times until my work banned it. I used it to write up business cases for a new position in my team and for buying 2 new high performance cars to add to our fleet. Worked well in all three cases with me just putting in basic data and it coughing out excellent reports to bolster my case.

[Hibs4185]

I have no experience of using it at all but I’m enjoying seeing Ukrainian supporters use Gronk to prove that Elon Musk is a *unt.

Gronk answers all the questions about the war and Zelensky accurately which is the complete opposite of it’s owner and creater.

Almost unbelievable and shows you Musk has an agenda

[The Modfather]

I read that AI is learning a form of self preservation. Stuff like this I find genuinely a bit frightening. Can’t be long until we’re having to fight Schwarzeneggers 😬

https://futurism.com/the-byte/openai-o1-self-preservation

[Jones28]

I read that AI is learning a form of self preservation. Stuff like this I find genuinely a bit frightening. Can’t be long until we’re having to fight Schwarzeneggers 😬

https://futurism.com/the-byte/openai-o1-self-preservation

So how does it get stopped?

It seems like there's this relentless drive towards AI, but what happens when AI is autonomous?

Is it literally a case of pulling the plug?

[Hibby70]

So how does it get stopped?

It seems like there's this relentless drive towards AI, but what happens when AI is autonomous?

Is it literally a case of pulling the plug?

Don't tell it that it has a plug for god's sake. It'll probably electrelocute you if you try to take it out.

[Colr]

There’s so much hype around this.

I wonder how long before we get a movie about 2 teenage boys using AI in their bedroom to create Kelly le Brock.

[grunt]

https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:tk32pf7q6xyremg2u6uj3gzz/bafkreia3fw5vwkeai23cmfaezobrlljjjt3rkhxs4a2jd3zjg aglmqi2ja@jpeg

[Smartie]

There’s so much hype around this.

I wonder how long before we get a movie about 2 teenage boys using AI in their bedroom to create Kelly le Brock.

Great post.

I remember there being a great deal of fear when I was wee, back in the 80s, about the march of the robots, the threat of automation etc.

Sure, technological progress has come with downsides but it hasn't been as apocalyptic as some were thinking at the time.

We can't be complacent but fatalistic is probably unnecessary at this stage. I speak as someone who survived the Millennium Bug.

[grunt]

I speak as someone who survived the Millennium Bug.
I'm not entirely clear why you referenced the Y2k issue, butt I'd like to point out that in the years prior to January 1 2000 I spent many months working to amend IT systems so they would survive the dawn of the millennium. I was not alone. Furthermore I used a piece of software on a daily basis which failed on 1 January 2000 and never worked again. It was a very real issue.

[Smartie]

I'm not entirely clear why you referenced the Y2k issue, butt I'd like to point out that in the years prior to January 1 2000 I spent many months working to amend IT systems so they would survive the dawn of the millennium. I was not alone. Furthermore I used a piece of software on a daily basis which failed on 1 January 2000 and never worked again. It was a very real issue.

My point is that a very real threat then existed, as it does with AI. But thanks to the initiative, endeavour, preparedness and actions of the likes of yourself, those threats were seen off. The nuclear weapons didn't start flying, the world didn't end. The terminal failure of a type of software was probably an acceptable consequence when there was a potential for it to be much worse.

The same is true with AI - there will be people who have the solutions to contain the theoretical worst case scenarios. It's tempting when you don't understand much about this sort of thing (which I'll happily admit that I don't) to assume that those who do know what they are doing are complacent and unprepared, when that is unlikely to be the case.

The fear narrative that is peddled is counter-productive, it gets people who can't influence the march of AI scared and unhappy, where they can then be potentially manipulated.

In many ways I find it comforting when I hear talk of the dangers, as I prefer the known unknowns to the unknown unknowns.

My concern is probably the developing disunity around the world with an increased onus on fighting each other, when a more unified approach to a common problem would be better. It has the potential to be weaponised.

[Hibrandenburg]

Some of us who have been around a bit may remember Microsoft's early search engine, Bing. It wasn't very good, and was particularly annoying because all new Microsoft installs defaulted to use it. It seems Bing has recently had an upgrade, incorporating the ChatGPT Artificial Intelligence (AI) interface, allowing text interactions between the user and the software. Scarily, the software appears to have a wish to be human ...

Shades of Isaac Asimov.

Here's a frankly terrifying article which includes a two hour discussion between the software and a NY Times columnist.

https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html?unlocked_article_code=5L9j4UTGa2Sq 9R0FulDKfmcq3aesNcqqwZkPBwsIW7EvqkbanIdsZU0hoKktwW ijvrw8qobVzmdbyZfxtRjhZQWMN1Qi39t035TVJEllXwc9_zmJ YNu_Y9gQ-8ROXYZ8ReuKsdz9WCv7bu_V75hZNZXipGPrv_h5g6phWp8GUX2 DTCGQMjV-CTjJybDonBvQaR_kPJbX2zzNM8s4PROKIK32tEVmqGFG0Z1mfz UetZbXqKtz-3b7jFe4M4ylQHHbCnqtSyX4zN6HT4SLsCliddmSW4Aq5V5MZXu wzlSaLgLra8kE5mc1MQuP7cMU9jA6JiP7lhfRpgwcKeHKrGOCR iL21Bw&smid=url-share

It might destroy civilisation as we know it, but it's good fun to piss about with on a slow Friday afternoon at work.

[grunt]

Endless hours of fun.

Someone on Threads noticed you can type any random sentence into Google, then add “meaning” afterwards, and you’ll get an AI explanation of a famous idiom or phrase you just made up.

https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:knlwbyvg36keh64h6awokbqx/bafkreics4wmldcfbs5ihsqhq37tpewwkpzweetzozp66kjxfv qt5obywua@jpeg

[Colr]

I’ve tried to use AI for research and its wrong more often than its right.

Even on popular culture stuff!

[AltheHibby]

On a sidenote, this thread was between the Trump and Reforms threads when I looked. Plenty artificial, no intelligence.

[grunt]

Chicago Sun-Times Prints AI-Generated Summer Reading List With Books That Don't Exist

https://www.404media.co/chicago-sun-times-prints-ai-generated-summer-reading-list-with-books-that-dont-exist/

[greenlex]

It’s taking over at Hearts and so far it’s going extremely well.:hilarious