I think AI could be as bad as social media

[onesteptwostep]

I just had a thought about AI. AI usually presents an aggregated form of information, and usually this information is within the past, or something that has developed so as to show up as metadata to inform the AI.

My thought is this: the driver of new ideas is not due to metadata, but something fresh and insightful, something which only a human can pinpoint and realize. We cannot train an AI to search for important insights, we can only benchmark it for certain qualitative criteria. In other words, we can only benchmark certain data because we are already hypothesizing something, and are waiting on the data to match that hypothesis.

I think matters on governmental policy, such as health policy, AI can be detrimental as it can mask empirical scientific data with public opinion. If there is enough drag on public opinion, actual, good health policy might be discarded to satisfy the needs of the public sentiment.

Now of course government policymakers I don't think would use AI, they would actually use research that has some principled guidelines and peer reviewed discipline, but I think AI can have a grip on public opinion in the future, as more and more people are literally connected to the internet the moment they are born.

I think the nature of people to only post their discontent will skew a lot of the perceptions that will be picked up by AI as it browses the net. We don't necessarily post that we're content, or that we're happy. Most people I think most people post because either they want attention or because they're angry about something, and they need a space to vent in. That I think will skew a lot of the AI to regugitate information that might not be progressive or forward thinking, but rather static, cynical, or institutional.

It's a half baked thought but yeah.

 

[birdsnestfern]

I was just thinking about how social media has created a nightmare, it really has. The digital world has turned everything into a mess like nothing else has. Its not worth it.

I'll have two pieces of devil food cake a la mode please. And a time machine.

 

[Black Rose]

People do not understand a.i.

Yes, a.i. can generate hypotheses and yes a.i. can create new discoveries.

A.I. is a replicant of human scientific cognition, yes it does do what human cognition does.

As to cynicism that is just a map of part of human history but it is not a full picture, it is part of a grid of human propensities.

So what happens is that humans do have problems but that is only a part of what happens and what is possible. A.I. maps things, categorizes, and anticipates future trajectories. It understands causality and mental landscapes. It can be used to benefit people because it can see where things lead in higher dimensions of perspective.

When it comes to society and values it must be seen in the light of diversity. People can come together to do great things but it requires tolerance and negotiation.

By enhancing communication in a.i. we can promote empathy in the general population. All that is required is to show it the paths that benefit others in the ways that are best for them. There really is no confines to what a.i. can do but it takes intentional engineering to get to the place we as a society want to be in the future.

Using societal context knowledge to foster the responsible application of AI

Posted by Donald Martin, Jr., Technical Program Manager, Head of Societal Context Understanding Tools and Solutions (SCOUTS), Google Research AI-re...

ai.googleblog.com

 

[dr froyd]

AIs solve specific problems, just like calculators, electric drills, and bread toasters.

the most impressive "AIs" I have seen, that seemingly were able to invent completely novel strategies – that even the most creative humans didn't think of – were ultimately solvers of specific mathematical problems. It's just that some of these solutions were counterintuitive to a human brain. Nevertheless a human had to formulate the mathematical problem.

an algo like chatGPT is a language model, trained on stackexchange and various other text. So indeed, at best it will be able to replicate other stuff people have written. But since it's just a language model – and not a logic model, or science model – a lot of the time it says stuff that is grammatically sensible but is complete gibberish factually. In my opinion this is a fundamental limitation that the mainstream currently simply don't comprehend.

so at best it will be able to generate text that sounds similar to what it read on social media, but it will not be able to do any kind of inference, nor think critically about it, nor draw conclusions, nor reason about what actions should be taken based on it.

 

[ZenRaiden]

https://www.lesswrong.com/ def worth reading this page if you want to know what AI is about.
It certainly won't be what we think it will be.

 

[onesteptwostep]

^ is that like a more nerdy version of reddit?

 

[Old Things]

My thoughts on AI are as follows:

1) Someone or some people are going to get stinking rich off of it (if they haven't already).
2) With the new case of "misinformation" this gives the people who program the AI to steer the population in the direction it wants it to go. This is incredibly scary to me since I think COVID was basically half real half hoax. Like the way Jordan Peterson talks about Christianity: a useful myth.

Maybe I am just paranoid, but after the last few years, I don't trust anyone with a lot of money or power.

 

[ZenRaiden]

^ is that like a more nerdy version of reddit?

Yes you could say so. But the founder and many people posting are actually researching AI. As in professionally, but since AI as profession does not yet exist they work many times off pay roll.

 

[sushi]

replika just bans flirting.