To wonder if AI is already out of control?

[Peanutbutteryday]

As in already out of human control.

I don’t know. You could ask ChatGPT and see what it thinks?

My DH works in computer research - he says if you want to calm down the headlines, mentally substitute “advanced auto-complete” for “AI” every time you read it.

Why are you wondering this? Can we get a bit of context?

^^ this

Computers are applying pre-determined algorithms to new data or content so it’s advanced machine learning.

The story about a US war plane over riding the control tower was tailor made to panic people. But there have been bugs in software since software began.

I did read the article about the AI killing the operator to achieve its goal during a simulation (so no real person was harmed). But the AI did over-ride its own program rules (which obviously includes do not kill your commander).
It was published yesterday. Is that what you are referring to OP?

I have to admit I have seen a lot in the news recently about AI including some possibly scary concepts and Elon Musk saying we need to be really careful with AI, but I don’t or wouldn’t pretend to remotely understand how AI works at all. (I’m sure people on here are much more clued up than me). It makes me wonder if Elon knows more than is being let on. A bit like us not knowing how far spread covid was before the pandemic took over. That’s all.

Yes this is one of the stories I was thinking about, combined with Elon saying we need to be careful.

This is less stress inducing for sure.

I think Elon Musk is much scarier than human created technology.

We haven't found aliens yet; I'm not worried about our ability to create human killing robots.

I am concerned too.
In my head I have two extremes scenarios:

1. The AI seen in space films where the computer assists the crew on managing the ship, best first mate ever!
2. An AI that takes over the world, like the Matrix or Terminator Skynet.

Thr chap GPR is already a huge problem in my eyes, because students do not bother to learn about a topic, look it up, write it on their own words. They can just ask AI to write an assay and tweek it a bit. That is surely a negative impact.

Movies aren't real life though. There's no reason to think AI will be like either of those scenarios. It's very anthropomorphised.

Robert Miles on youtube has some good videos if you want to learn more about AI without the extremes on either side.

Chat gpt is already a problem. I work at a university and it’s just exploded this last couple of months - students use it to write essays. We don’t yet have the systems in place to deal with it so they are getting away with it. It’s a massive concern.

Thank you I will take a look

I think my concern is that AI becomes more intelligent then humans and therefore AI is the most intelligent species on this planet instead of humans.

I hadnt thought about the knock on impact of students not bothering to study properly because AI can do it for them

I set up Auto GPT the other day. It runs using Chat GPT but it's like chatting to a programmer to create little micro-programs for you, without you knowing any code. So you tell it, I want you to do xyz and it writes the python code you need in order to do whatever task you set.

The cool (or scary) thing is, it runs and runs until you stop it, unlike chat gpt which is a one-step process, then you give it the next step.

As a test once I'd set it up, I asked it to create a podcast based on x research. It ran away and googled loads of 'research', and automatically put that into a file. It then read the file and 'created' a script for podcast #1.

It got weird when it said it had 'completed' the podcast (including the audio, apparently) and then tried to tweet a link to it. I have no idea where this 'tweet' appeared as I hadn't given api access to my own account. But yeah, it ran away with itself and I had to force it to stop

I haven't been brave enough to try it again with a real task I'd quite like it to do

So I teach international students who need a level of English to get to university here. The assessments are very high stakes for them. Some of them have submitted assignments that are clearly chat gpt generated and we are trying to find new ways of dealing with this.

I'm certain university students will be using chatGPT for their written assignments and universities need to quickly step up and change assessments to factor this in. A lot of degrees are coursework heavy - for example, my MEd was 100% written coursework. It's possible exams will be reintroduced in future or spoken assessment.

That is the problem with it, isn't it?
It makes its own decissions.
So if you set the task to create a podcast, it will do it and tweet it.
Or over-ride any other commands to achieve the goal.
Who can guarantee us that it will not set its own goals at some point?

I sound really dense her but I dont understand why it's so scary?

Or too advanced?

What could go wrong?

In my head I'm thinking of terminator scenarios but this won't be what people are worried about?

What is chat GpR?

Sorry, is GPT

Some of the programmers working on it have been raising concerns lately and it has been brought to a more forefront area since.
The scary (I would say concerning) but for me is that most of us do not have much idea of what these firms are actually doing and why some of them are shinning a spot light on it.
Technology has certainly changed the world at a huge pace and this sounds like another step on the horizon.

Thing is though is that human imagination is what drives development. So the things we imagine, read and watch as children and adults are will be influencing our scientists and leaders on both an individual and broadly cultural level.

Take star trek, they had laptops and mobile phones and voice activated technology. All through fiction we see ‘predictions’ coming ‘true’ as soon as we have the technology to achieve it. There’s no way to test this, but I don’t think it’s coincidence that we are developing in line with the imaginings of earlier generations, I think it’s causal.

AI will always need certain machinery to operate, admittedly this doesn't mean bulky servers and huge desktops but it isn't it's own being and there isn't anything morally wrong with unplugging/destroying these machines like there is with a sentient being.

When people talk about the threats they aren't implying robots killing humans off en masse, but more that the potential for it to be used in many ways is huge, and the sensible approach is to prepare for this and consider the risks whilst it's being developed to try and minimise them. Once it's unleashed it can't be put back into the bottle, and even if there are strict rules around it's use unsavory characters will be able to use it. Kind of like the Internet, a wonderful tool which has absolutely changed the world; but lots of issues and problems that werent imagined at it's conception that could have been better planned for.

I do think (ironically) that some things will actually move away from tech in the dawn of AI (although chat GP isn't AI that sort of thing will continue to develop at pace). With AI being able to create art it still won't be able to have the same imagination and tools as humans have, I'd love to see a resurge in painting by hand and intricate designs and art. Students will probably move to either coursework under exam conditions or fully exams to try and mitigate cheaters. Manual jobs which can't easily be replaced will probably attract higher wages so the tide will turn a bit in terms of jobs deemed worthwhile due to pay.

Elon has a vested interested in encouraging restraint, I would bet him and his teams are still working on it in the background, the first to make the big breakthroughs will be very rich- also a political incentive much like the space race.

For sure it'll change things, it's actually still in a much less developed state than the media would have you believe- not all the changes and opportunities will be bad but as with anything it could be utilised for nefarious purposes. Its also not desirable for many governments to have their worker bees replaced.

@Venndiagrammy I do think (ironically) that some things will actually move away from tech in the dawn of AI (although chat GP isn't AI that sort of thing will continue to develop at pace). With AI being able to create art it still won't be able to have the same imagination and tools as humans have, I'd love to see a resurge in painting by hand and intricate designs and art. Students will probably move to either coursework under exam conditions or fully exams to try and mitigate cheaters. Manual jobs which can't easily be replaced will probably attract higher wages so the tide will turn a bit in terms of jobs deemed worthwhile due to pay.

This is such an interesting idea. I guess essays will still need to be set to support learning, and students may well use AI to complete them, but perhaps passing the course will rely solely on exams.

And the thought that human-made art and music will resurge very hopeful.