Is chat GPT a danger to humanity?

[Mr. Dazzle]

Interesting comments from OpenAI, the creators of ChatGPT.

Supposedly the idea behind the recent explosion in AI performance is already dried out and the next explosion will have to come from elsewhere. But I suppose they would say that, maybe they're the ones dried out!

https://www.wired.com/story/openai-ceo- ... eady-over/

[Mr. Dazzle]

But if the models don't get any bigger wouldn't they just be recirculating nonsense in that case?

They've now gone past politicians and reached "half an hour before closing time".

[Count Steer]

But if the models don't get any bigger wouldn't they just be recirculating nonsense in that case?

They've now gone past politicians and reached "half an hour before closing time".

There comes a point that increases in scale just result in faster GIGO. These things just process data and there's no quality control on the data input or on the conclusions drawn from it. I'm intrigued by the suggestion that the quality of the output will need to be judged and fed back in. Who is going to judge it? The public, based on whether they like an answer or not? May your deity of choice help us all.

[ChrisW]

Who is going to judge it? The public, based on whether they like an answer or not?

Oh no, captchas are going to get massively more annoying.

[Horse]

His idea is to use them when the owner doesn't need it. A brilliant idea actually and saves on parking!

Is it actually 'his' idea? I've heard it being talked about for years, amongst many other potential opportunities from them.

The other "hidden" feature being that all of that data capture from each individual robot/car is shared with the millions of others.

Hidden? Tosh and nonsense. It's just one part of the 'connected' that puts the C into CAV. Has been for years and the idea of transferring information has, similarly, been part of CAV development for years.

So hidden that uk government knows about it:
https://www.gov.uk/government/organisat ... s-vehicles

CCAV has been about since at least 2017.

[Cousin Jack]

If every AV knows about every pothole it encounters, and shares that with others, why the fuck can councils pretend they don't know about a pothole unless it has been reported to them with precise location?

[Mr. Dazzle]

My car tells me about pot holes, but only sometimes. Bugs still to be worked out it would seem.

[MrLongbeard]

What's the current state of autonomous vehicles in the UK? I thought they weren't allowed yet.

Ford Mach-E on certain roads has been signed off and approved
https://news.sky.com/story/self-driving ... t-12856889

[Mr. Dazzle]

Mercedes S-Class is too I believe, maybe only in Germany. They took great pride in beating Tesla to having a properly homologated AV.

https://europe.autonews.com/automakers/ ... -class-eqs

[Horse]

Don't know how it works around your way but in London they don't get filled because Sadiq Khan is a cunt.

Simple questions, for anyone:
1. What don't you want councils to spend on to allow budget for potholes?
2. Are you going to whinge and whine when extensive lengths of roads are closed for repairs? Ideally, any utilities work, such as replacng old water mains, would be done at the same time.

One if the best cost effective ways of stopping roads deteriorating is tar and chippings; happy with that?

[Horse]

https://www.economist.com/britain/2023/ ... riving-bus

The ab1 bus service to Edinburgh seems much like any other. It leaves Ferrytoll Park and Ride, on the north bank of the River Forth, and crosses onto the m90 motorway, reaching a top speed of 50mph and encountering a smattering of junctions, roundabouts and traffic lights on its 25-minute journey into the city. None of this would be at all remarkable—but for the fact that it does so without any input from a human driver.

[weeksy]

The other "hidden" feature being that all of that data capture from each individual robot/car is shared with the millions of others. Instantly. When one of them knows where I live, they all do. If one of them sees a pothole, they will all know. Spooky but just goes to show how much data the big boys are playing with.

I hope I am right about Elon and he really is both as smart and as altruistic as he appears (at least to me anyway).

But that's not AI, that's just a database sharing information between computers. If i post a document on a share ALL the users who have access can see it... because it's 'there' that doesn't mean there's any magic, it just means they've share communications between computers. There's nothing deep and meaningful in there. It's just added a GPS coordinate to a database pointing to a pothole.

[Count Steer]

Can't resist repeating myself here. While I am an avid fan of Elon Musk (who basically invented "Open AI" meaning for it to be open source instead of a closed shop) he is also working on his own AI. X.AI I think it's called.

That probably explains why he wants the other developers to freeze activities for 6 months.

[Horse]

The other "hidden" feature being that all of that data capture from each individual robot/car is shared with the millions of others. Instantly. When one of them knows where I live, they all do. If one of them sees a pothole, they will all know. Spooky but just goes to show how much data the big boys are playing with.

But that's not AI, that's just a database sharing information between computers. If

... because it's 'there' that doesn't mean there's any magic, it just means they've share communications between computers.

With implications for data security.

https://www.dotmagazine.online/issues/o ... ected-cars

May 2019
DATA PROTECTION FOR CONNECTED AND AUTONOMOUS VEHICLES

[Cousin Jack]

Don't know how it works around your way but in London they don't get filled because Sadiq Khan is a cunt.

Simple questions, for anyone:
1. What don't you want councils to spend on to allow budget for potholes?
2. Are you going to whinge and whine when extensive lengths of roads are closed for repairs? Ideally, any utilities work, such as replacng old water mains, would be done at the same time.

One if the best cost effective ways of stopping roads deteriorating is tar and chippings; happy with that?

I would slash the budget for adult social care for a start. Housing too, I ibject to dossers being put into hotels. It costs a fortune but is totally ineffective.
I can live with tar and chippings,, roll them un and brush them up.

[weeksy]

The other "hidden" feature being that all of that data capture from each individual robot/car is shared with the millions of others. Instantly. When one of them knows where I live, they all do. If one of them sees a pothole, they will all know. Spooky but just goes to show how much data the big boys are playing with.

I hope I am right about Elon and he really is both as smart and as altruistic as he appears (at least to me anyway).

But that's not AI, that's just a database sharing information between computers. If i post a document on a share ALL the users who have access can see it... because it's 'there' that doesn't mean there's any magic, it just means they've share communications between computers. There's nothing deep and meaningful in there. It's just added a GPS coordinate to a database pointing to a pothole.

But of course, it is an AI so while it shares some features of a dumb database, you have to remember the AI can model behaviours and work out a strategy based on "what works". And it ain't just potholes. Every microphone and every camera can and will record everything within its reach. I read recently that every bit and byte of information (voice, text, video etc.) that has ever been on a wide area network has been captured, recorded and stored for the past 20 years.

Your AI equipped robot/car will of course be trying to work out how to get money out of you either through targeted advertising or more insidious psychological "tricks" for which there is an entire library of research. These are neurological adapter which target particular parts of the brain to put you in the right mood for easy manipulation. If you think that's "tinfoil hat time" I should remind you this has been going on already for years... AI will just be way more targeted and devious. It will be incredibly persuasive because, a bit like J.A.R.V.I.S. your personal AI will also be a "friend" and confidante.

Nah, disagree with all of that.... compltely completely disagree... You're wrong.

[weeksy]

Nah, disagree with all of that.... compltely completely disagree... You're wrong.

So you disagree that “AI can model behaviours and work out a strategy based on "what works".”

How can you disagree with that? That is practically a definition of how AI is commonly used.

I disagree that that's what's happening. It's a computer, it's given information, it uses it, stores it and applies the algorithm associated to it. The car isn't AI, it's a car with a computer chip that does as it's told.

It's like saying Google maps is AI because it will give you different directions based upon traffic issues. No it's not... it's told "plot the fastest route" and it then uses the data associated to set that route based upon calculations from previous traffic, it's not AI, it's not thinking, it's not intelligent, it's just a computer.

[Horse]

But of course, it is an AI so while it shares some features of a dumb database

So a pair of tin cans on a string, which allows data sharing, could - of course - also be an AI?

Every microphone and every camera can and will record everything within its reach

I bet no-one has realised that! Oh, hang on, that 2019 article on AV data that I linked to.

Your AI equipped robot/car will of course be trying to work out how to get money out of you

So just a localised version of what Google, Facebook, et al do now?

[Wossname]

…. Stuff…

I read recently that every bit and byte of information (voice, text, video etc.) that has ever been on a wide area network has been captured, recorded and stored for the past 20 years.

Etc

I don’t believe everything I read, especially the more outlandish stuff. “If it sounds too good/bad to be true, it probably…”

It stops me worrying unnecessarily. You should try it some time.

[dern]

…. Stuff…

I read recently that every bit and byte of information (voice, text, video etc.) that has ever been on a wide area network has been captured, recorded and stored for the past 20 years.

Etc

I don’t believe everything I read, especially the more outlandish stuff. “If it sounds too good/bad to be true, it probably…”

It stops me worrying unnecessarily. You should try it some time.

Too right, I hardly believe anything these days. Instead I form an opinion based on a wide variety of sources over periods of time from which I can form a fuzzy picture of what I think is happening.

Then I post it here to see which aspects of my reasoning are incorrect. So it's disappointing when people wave a hand vaguely at a post and tell me "nah bollocks". Hey ho. I will continue to maintain an opinion until I am given sufficient information to refute whatever factual content may either be missing or incorrect.

As for the data capture, what do you suppose the CIA/FBI/MI6 are doing all day? Why has Amazon become so ally with the military industrial complex? What does this all mean: https://aws.amazon.com/blogs/industries ... rm-on-aws/

Anyhow, a quick Google confirms data is being captured and stored routinely, no surprise to me. What is less surprising is that it is China who are the biggest player. Read 'em and weep: https://theconversation.com/the-worlds- ... red-159964

Some interesting numbers anyway...
But they're just numbers.

But none of your sources are right. If you took a couple of hours out to read up on basics of the theory of machine learning you wouldn't be posting such nonsense. Even just reading the wiki page on machine learning would stand you in good stead. It's just a technique for pattern matching without explicitly coding the patterns that you're looking for. You either user supervised or unsupervised learning to establish whether a thing is the same or similar to something else and how similar... that's it. You can apply this to predictive modelling, face recognition, generative work and loads of other stuff but, and this is the key, with this kind of technique you cannot model and will never ever be able to model sentience. You might be able to fool someone in to thinking a program is sentient but that doesn't mean that it is.

What you can do is that you can generate stuff that looks very convincing, even indistinguishable from the 'real thing'. Therefore, to my point a few days ago, the only danger is to the unwary who can't (or won't) fact check. To those people it's very dangerous indeed *but* the motivation will be provided by the people who have created the ML algorithm and most certainly will *not* be provided by any mechanism produced by AI.

The picture you're painting is a paranoid view of incorrect information. You could not possibly be more wrong, that danger comes from people wanting to manipulate you to believe in stuff (including the stuff you've posted). AI can't think for itself any more than a gun could murder someone.