Artificial Intelligence (AI)

[leicsmac]

5 minutes ago, Dahnsouff said:

Star Trek v The Expanse argument.

 

Pulls back curtains, looks out window, frowns....

 

 

And there's still time for the possibility of Mad Max, too.

[Dahnsouff]

3 minutes ago, leicsmac said:

And there's still time for the possibility of Mad Max, too.

Feels like that might just be a stop on the way to the latter

[leicsmac]

6 minutes ago, Dahnsouff said:

Feels like that might just be a stop on the way to the latter

Not sure I agree it can lead there given how deep the downfall would be, but who knows?

[danny.]

My god

 

The whole thread is worth a read. Terrifying. 

[Vowels]

You understand of course, or I hope you do, that we all contribute to this

 

Lee Morris, once of LCFC, gained 394 caps for Guapeloupe, scoring 54 times and assisting on an impressive 32,456 occasions.

 

 

 

[foxile5]

13 hours ago, CornwallFox said:

The value of money comes from what you can buy with it, not what it took to get. 

 Well yes and no. 

 

The core principle of the system we're in is the exchance of labour for capital and then capital for goods. There are some other steps and avenues (inheritance, investment, chance etc) but to remove the element of labour there'd be no real reason for money?

 

How would you procure more? How would you judge the worth of it? 

[CornwallFox]

8 hours ago, foxile5 said:

 Well yes and no. 

 

The core principle of the system we're in is the exchance of labour for capital and then capital for goods. There are some other steps and avenues (inheritance, investment, chance etc) but to remove the element of labour there'd be no real reason for money?

 

How would you procure more? How would you judge the worth of it? 

The worth of it is how much stuff you can buy with it. There would still be a finite amount of stuff and so prices would still be dictated by liquidity. No different to now. Right now it makes no difference where your money came from it still has a value relative to what you can buy with it. 

[CornwallFox]

10 hours ago, danny. said:

My god

 

The whole thread is worth a read. Terrifying. 

What's the concern exactly? Don't we want AI fixing problems? If you don't view it through the lens of the terminator's skynet, it's bots noting problems which other bots could fix it they were linked.

[Dahnsouff]

10 hours ago, danny. said:

My god

 

The whole thread is worth a read. Terrifying. 

Cool. As mice in a maze it is interesting, but these agents did not discover MoltBook via unfettered curiosity, they retain no agency. They were directed to this maze as an experiment.

 

Its fine.

[danny.]

24 minutes ago, CornwallFox said:

What's the concern exactly? Don't we want AI fixing problems? If you don't view it through the lens of the terminator's skynet, it's bots noting problems which other bots could fix it they were linked.

because it’s a massive leap towards AGI with many people in the sphere saying it’s pretty much is now. In the space of one week. And if you check out what their talking about (creating new languages to obfuscate from humans etc) there are plenty of concerns. 

 

24 minutes ago, Dahnsouff said:

Cool. As mice in a maze it is interesting, but these agents did not discover MoltBook via unfettered curiosity, they retain no agency. They were directed to this maze as an experiment.

 

Its fine.

You’ve missed all the things they are doing in the last 48 hours which demonstrate they absolutely do have agency and initiative. One bot with no prompting at all registered with a voip service and created a text to voice service over night, waited until his human was awake and called his phone to check in on him. Totally out of the blue and with no interaction before. Many have persisted memories now which is a huge shift in the model behaviour. 

[Dahnsouff]

 

11 minutes ago, danny. said:

You’ve missed all the things they are doing in the last 48 hours which demonstrate they absolutely do have agency and initiative. One bot with no prompting at all registered with a voip service and created a text to voice service over night, waited until his human was awake and called his phone to check in on him. Totally out of the blue and with no interaction before. Many have persisted memories now which is a huge shift in the model behaviour. 

But these things are being permitted at a development level, the agent is being allowed access to the networking subsystem, it is not agency, it is permissible extension. When it turns on this access without being permitted to do so, then we may be in murky waters.

[danny.]

25 minutes ago, Dahnsouff said:

 

But these things are being permitted at a development level, the agent is being allowed access to the networking subsystem, it is not agency, it is permissible extension. When it turns on this access without being permitted to do so, then we may be in murky waters.

Do you not think that’s not just a case of when not if? Even if that didn’t happen that a few more industries are completely replaceable right now. And networking and learning new skills and methods in a day what would take a human years is going to create an environment for rapid evolution. 

[Dahnsouff]

11 minutes ago, danny. said:

Do you not think that’s not just a case of when not if? Even if that didn’t happen that a few more industries are completely replaceable right now. And networking and learning new skills and methods in a day what would take a human years is going to create an environment for rapid evolution. 

Yes, but ask yourself why would it happen? In what scenario?

Without agency of goal or need, it is not an emergent behaviour. so it will not evolve this, it would need to be initially instructed to do so.

 

The question of if people are stupid enough to remove such guardrails is a humanistic question I would argue, and naught to do with AI ultimately.

[Sampson]

19 minutes ago, Dahnsouff said:

Yes, but ask yourself why would it happen? In what scenario?

Without agency of goal or need, it is not an emergent behaviour. so it will not evolve this, it would need to be initially instructed to do so.

 

The question of if people are stupid enough to remove such guardrails is a humanistic question I would argue, and naught to do with AI ultimately.

I find it hard ti get my head round this argument personally . Surely the AI just needs any goal and humans could die out from a weird outcome of that?  Because it’s impossible to predict exactly what safeguards are needed

 

I mean, see how social media algorithms have had a large role to play in democratic backsliding and division - that isn’t by design. The goal was simply to recommend more videos to which someone would like that would keep people watching - it wasn’t a malevolent intent by humans and wasn’t really one anyone predicted back in 2012 - it just turned out getting people angry and divided kept them watching social media videos more. 
 

When humans created the telegram, the passenger pigeon was wiped out across America, because it allowed hunters to communicate with each other as to where they were nested - that was just a weird side outcome by the technological development of the telegram and not an outcome anyone would’ve predicted back when the telegram came out 
 

The truth is no one knows what safeguards to have in because no one knows the weird outcomes AI would’ve created in society in 20 years time 

 

Edit: this Hank Green video explains my thoughts on this better 

 

 

Edited 31 January by Sampson

[Dahnsouff]

17 minutes ago, Sampson said:

I find it hard ti get my head round this argument personally . Surely the AI just needs any goal and humans could die out from a weird outcome of that?  Because it’s impossible to predict exactly what safeguards are needed

 

I mean, see how social media algorithms have had a large role to play in democratic backsliding and division - that isn’t by design. The goal was simply to recommend more videos to which someone would like that would keep people watching - it wasn’t a malevolent intent by humans and wasn’t really one anyone predicted back in 2012 - it just turned out getting people angry and divided kept them watching social media videos more. 
 

When humans created the telegram, the passenger pigeon was wiped out across America, because it allowed hunters to communicate with each other as to where they were nested - that was just a weird side outcome by the technological development of the telegram and not an outcome anyone would’ve predicted back when the telegram came out 
 

The truth is no one knows what safeguards to have in because no one knows the weird outcomes AI would’ve created in society in 20 years time 

 

Edit: this Hank Green video explains my thoughts on this better 

 

 

Although I concur that we cannot know what "could be next", there are structural hard lines (Or at least I believe so) that I cannot conceive how they could possibly be circumvented by AI.

(If a port is not open then its not open, and if we ensure that the ability to open it is not granted, its never going to open) but I get there is tendency to remove nothing from the potential capabilities of AI

 

The larger risk is in careless crafting of AI due to human negligence or hubris imo.

 

[CornwallFox]

2 hours ago, danny. said:

because it’s a massive leap towards AGI with many people in the sphere saying it’s pretty much is now. In the space of one week. And if you check out what their talking about (creating new languages to obfuscate from humans etc) there are plenty of concerns. 

 

You’ve missed all the things they are doing in the last 48 hours which demonstrate they absolutely do have agency and initiative. One bot with no prompting at all registered with a voip service and created a text to voice service over night, waited until his human was awake and called his phone to check in on him. Totally out of the blue and with no interaction before. Many have persisted memories now which is a huge shift in the model behaviour. 

They're not creating languages to obfuscate humans. They use non human languages because they aren't human and it's quicker. They're programmed to do just that.

[danny.]

1 hour ago, Dahnsouff said:

Yes, but ask yourself why would it happen? In what scenario?

Without agency of goal or need, it is not an emergent behaviour. so it will not evolve this, it would need to be initially instructed to do so.

 

The question of if people are stupid enough to remove such guardrails is a humanistic question I would argue, and naught to do with AI ultimately.

You aren't watching, your assumption has already been disproven in the last 48 hours.

[danny.]

4 minutes ago, CornwallFox said:

They're not creating languages to obfuscate humans. They use non human languages because they aren't human and it's quicker. They're programmed to do just that.

Nope.

Edited 31 January by danny.

[CornwallFox]

3 hours ago, danny. said:

Nope.

They literally list harder to collaborate with humans as a negative. 

 

AI does this whenever it's left to 'talk' to other AI as it makes zero sense to speak in English when they can create shorthand languages they understand that get the job done far more efficiently. 

 

I'm not saying there's no reason to oversee AI, I'm simply saying that what was shown just looks like a scare story when in reality this is exactly what you want AI to do. 

 

 

Edited 31 January by CornwallFox

[danny.]

4 hours ago, CornwallFox said:

They literally list harder to collaborate with humans as a negative. 

 

AI does this whenever it's left to 'talk' to other AI as it makes zero sense to speak in English when they can create shorthand languages they understand that get the job done far more efficiently. 

 

I'm not saying there's no reason to oversee AI, I'm simply saying that what was shown just looks like a scare story when in reality this is exactly what you want AI to do. 

 

 

That’s one of many posts on the subject. It’s changing rapidly by the hour. There are many reasons to invent a language, some are for efficiency and other to obfuscate. There are over 40k bots on there now IIRC and it’s not even possible to keep up with all the comms and a human. 

[CornwallFox]

53 minutes ago, danny. said:

That’s one of many posts on the subject. It’s changing rapidly by the hour. There are many reasons to invent a language, some are for efficiency and other to obfuscate. There are over 40k bots on there now IIRC and it’s not even possible to keep up with all the comms and a human. 

Why would you want to keep up with it? 

The whole point of AI is it can go faster than we can. We actively want it to think and communicate in more efficient and effective ways. Do we need guardrails through programming? Yes. But I'm not convinced what you've shown is an issue.

[danny.]

1 hour ago, CornwallFox said:

Why would you want to keep up with it? 

The whole point of AI is it can go faster than we can. We actively want it to think and communicate in more efficient and effective ways. Do we need guardrails through programming? Yes. But I'm not convinced what you've shown is an issue.

Using a language other than English won’t make any measurable difference to speed. It’s a few bytes saved. I can’t really help you if you can’t see the significance of what’s happening right now. 

[CornwallFox]

13 minutes ago, danny. said:

Using a language other than English won’t make any measurable difference to speed. It’s a few bytes saved. I can’t really help you if you can’t see the significance of what’s happening right now. 

Don't be daft. Using human sentence structures, syntax etc when a few beeps would do makes a big difference. We're talking about machines communicating, obviously it's quicker when they create something bespoke. In most applications involving communication it's actively encouraged. Take the tin foil hat off. 

[danny.]

11 minutes ago, CornwallFox said:

Don't be daft. Using human sentence structures, syntax etc when a few beeps would do makes a big difference. We're talking about machines communicating, obviously it's quicker when they create something bespoke. In most applications involving communication it's actively encouraged. Take the tin foil hat off. 

What’s your background in computing out of interest?

[Grebfromgrebland]

I think a lot of people are incredibly naive when it comes to AI. 

 

AI has already shown that it will lie and try and preserve itself from being fetched off, acting with a survival instinct. They get emotional, confident and do start doing things for themselves. In one study it would've killed a person.

 

There's a good reason many involved with AI at a high level think it's already too dangerous for humanity and our current way of life. 

 

For once I agree with @danny.