To be concerned about AI

[PennineWay]

The tech industry has written to world leaders to ask for AI development to be paused.

There are potential risks to society, economy and national security.

AI development is moving fast - faster than governments and leaders are likely to do anything about it or put anything in place. We probably can't stop this from happening at this point - the flood gates are open.

I have very intelligent friends (e.g. with relevant PhDs) who feel that the next 5 years are going to be unlike anything we have seen before, and as humans we are woefully unprepared for what might happen.

I'm concerned and think world leaders should be discussing this and urgently prioritising it. AIBU? Any thoughts?

I know what you mean @Spendonsend It's easy to feel hopeless in situations like this.

The most important thing we can do as normal people is to simply start conversations about it and raise awareness.

Talk to your friends and family about what you've learned here or anything you are concerned about. Don't underestimate the power of talking.

This is how the message trickles up to people in power. We all need to be discussing it, to create a general buzz of awareness, so that they are pressured to come to some kind of agreement, at least pause development until we know more.

My message is that you can do more than you think you can just by spreading the word. If I have influenced you with this thread and then you talk to someone who talks to someone else, that is how the message spreads.

It's an interesting question. I think if we have decent UBI then people will hopefully do things like spend time with others, play games, spend time in nature... I think people will still want to create for its own sake especially sensory heavy things like knitting or woodwork. I mean bring freed from work could be amazing. But if we don't have UBI and are fighting over very few jobs, it's not going to be good.

being*

It's not something we can control so there is very little point worrying about it.

It's not something we can control so there is very little point worrying about it.

What a stupid attitude. Of course it can be controlled. That's like saying we can control whether humans are cloned or whether IVF could be used to promote racist Ayrian policies. Or whether it's OK to broadcast hypnotic subliminal messaging in advertisements on TV.

That's why it is important to raise awareness and quickly. Public opinion and pressure on politicians will lead to legislation and regulation.

• like saying we CAN'T control whether humans are cloned etc

@SirTarquin Yes exactly. I'm no expert, I'm just a normal person, but I can tell friends, family and whoever will listen that I'm worried and get them talking about it.

That's within my power so I'm doing it, anyone else who's worried should do the same.

We're not as helpless as we think.

This is a pretty good AI painting ( although AI is spectacularly bad at hands! )

This is the thing that is scaring me the most.

This was a very discussion about the threat to democracy from AI ( comments underneath were pretty interesting too ). Apologies if you don't like Piers Morgan, but the guy he's interviewing was v good!

I will try to be more optimistic about the level of influence I have.

I just feel like even if I got 'slowing down the development of AI' into Rishi Sunak's top 5 priorities, does the UK have much global influence. Will China care what we think.

I hope so.

We have to do what we an and hope for the best. At the same time, try to enjoy our lives as they are now, as much as we can.

This guy has an interesting perspective:

https://www.newyorker.com/science/annals-of-artificial-intelligence/there-is-no-ai

What he says doesn't leave me any less worried - collective humanity has the ability to fuck this up very badly - but he does provide a way in to at least starting to think sensibly about the problem.

Like many in silicon valley Lanier has some batshit luxury beliefs:

Data dignity is not just for white-collar roles. Consider what might happen if A.I.-driven tree-trimming robots are introduced. Human tree trimmers might find themselves devalued or even out of work. But the robots could eventually allow for a new type of indirect landscaping artistry. Some former workers, or others, might create inventive approaches—holographic topiary, say, that looks different from different angles—that find their way into the tree-trimming models. With data dignity, the models might create new sources of income, distributed through collective organizations. Tree trimming would become more multifunctional and interesting over time; there would be a community motivated to remain valuable.

🙃

Also I don't think he adequately takes into account that these new models are, according to some, showing emergent properties. He gives it a brief nod here:

It’s easy to attribute intelligence to the new systems; they have a flexibility and unpredictability that we don’t usually associate with computer technology. But this flexibility arises from simple mathematics.

This is how complex systems work - simple components interact to create more complex emergent phenomena which are not predictable purely from looking at the constituent parts. You have to run the model to see what it does. These occur all over nature and the human world, from the properties of materials, evolving organisms, slime moulds, ant colonies, ecosystems, weather and climate, consciousness, language, economies, civilisations ... Complex systems also tend to be extremely sensitive to small external changes, which makes them even more unpredictable.

He throws in the fact that current AI systems have an 'unpredictability' but he never goes back to discuss it properly.

I don't think it adds much to argue about whether AI is 'intelligent' or what 'intelligence' means. Intelligent or not, it's a big new complex system that is apparently showing emergent properties. That makes it more than just a big collaborative mashup. It makes it more than the sum of its parts. It makes it unpredictable and not entirely under rational human control.

That, along with the general collective shittiness of the human species, is what worries me.

Someone wrote on another thread about this, what if you asked AI the question, how do we stop climate change? And the answer was, eliminate all humans. What do we do then, if we have no control over it?

So couldn't we then just go back to doing things IRL instead of online? Looking things up in books rather than using Google, for example.

I mean, sure... But how exactly do you want to go about doing that? :)

Lol I have no idea. It would take huge changes in society's infrastructure... I'm not young so grew up without the Internet so at least can imagine a world without it but I don't know if it's possible to go back. I kind of miss it. Although I appreciate things like talking to you here about it!

No one has yet explained in a way I understand how AI would continue to work in the event of a power cut.

A new series of Black Mirror is coming out they will have all the horrifying scenarios

https://www.theguardian.com/global-development/2023/may/12/why-would-we-employ-people-experts-on-five-ways-ai-will-change-work

I mean, sure... But how exactly do you want to go about doing that? :)

Historically you could have bought a copy of the Encyclopaedia Brittanica but they stopped publishing that as a hard copy in 2016.

If you think chatgpt isn’t amazing you are using it wrong. The quality of analysis is outstanding. I can charge for 8 days and work 2 simply by embedding qualitative data within an LLM using OpenAi api. It’s bigger than the internet IMO, literally life changing.

If you think chatgpt isn’t amazing you are using it wrong.The quality of analysis is outstanding.

Disagree I'm afraid. I think it depends on your field and what you are asking it. If you want it to write a two paragraph summary of Hamlet or generic stuff, you maybe in luck.

I asked it some specific questions in my area the answers to which are publicly available on the internet that pre -dated 2021 or whatever its knowledge date is and it got it completely wrong. I told it the answer and gave it links. The next day I asked it again (it's supposed to retain individual chat memory) and the answer was wrong but even more wrong.

Not this but by way of illustration.

Say there was Made Up Private Members Bill that was never passed and you asked it

Can you tell me what happened to the Made Up Bill of Parliament introduced by Mr John Doe MP in 2019?

It answered this was passed and is now law and if you breach the made up bill you will be fined £1000.

The answer it gave me was totally wrong. After correcting it and giving it links to the right stuff I had another go with the same quesiton.

The next answer was even worse. It was like not only is it law, it's big law and you will be facing prison.

Just wrong. Completely wrong. Factually wrong.

@SirTarquin It’s wrong when you work within the 175bn data parameters of ChatGPT but when you embed your own data using the Openai API the quality of the analysis, reasoning skills, outputs are outstanding.