back
45 comments
So much of AI is all about tricking people into thinking they are interacting with a real person, or that something was created by a real person.

It’s become a dog whistle such that if a company or person uses AI, you have to question how much they actually care about or value the actual service they are providing. It’s very quickly becoming associated with spammy and scammy entities. Just look at the recent Willie Wonka event that used generative AI in its marketing to trick people into thinking what they were buying was a higher value than what it actually was. The bar on creative work, which once served as a proxy to actual value, has been eroded and reduced so low and people are catching on. The net affect on society is not good and causing less trust, not more and will likely lead to greatly reduced valuations of AI as a whole, no matter what it creates and at what quality —- people may just stop consuming altogether or be less likely to consume.

My prediction is that the closed internet (see walled gardens) will give way to the small internet (small networks of people in the style of whatsapp or discord) where there are no feeds. And GenAI (which is disturbingly being used to represent all AI) will become the tool of spammers and office workers
AI is just whatever the newest algorithm is. Chess.com uses AI, Google: 10 blue links version used AI (I'm feeling lucky), Shazam uses AI.

It's all AI. I know there's recent hype but it feels like we've all lost perspective a bit.

I think there's a sense that our personal experiences are being hijacked by algorithms. This has been going on for at least a decade or two but they've been getting more and more devious and extractive as the Internet, and our attention has been heavily capitalized upon. Nothing is private. Everything is for sale. Your worth is algorithmic, and AI is the biggest and most devious algorithm there is.
What's sad is there were early technologists warning about the social media problems, but they were ignored by the 90% builders of it and $. Same thing happening all over again, except no one thinks Facebook is really helping people.
I do think this is the blind spot in all of Altmans hyper competitive moves.

As a public we’ve had years of cultural warnings that this tech especially needs to be navigated with wisdom and ethical consideration.

Trust can initially be quickly gained, but once lost, very hard to regain.

I’m not sure about anyone else but I’m already cautious which questions I ask Chat GPT and I’m sure I’m not the only one.

I think the trust comment is the crux of it. When AI can suddenly generate photorealistic images and answer any question, people wonder what it will be able to do next. Generate movies? TV shows? Video games even? And then a lawyer gets disbarred because ChatGPT made up cases and people's excitement starts to waver. And then people start using existing AI to make deepfake porn and people's excitement starts to waver. And then AI starts generating photorealistic black Nazis, and people's excitement starts to waver again. If the negative incidents keep happening but people's lives don't get better at the same time thanks to the utility of AI, people start asking "Are the downsides worth it?" and "Can we even trust the people making this stuff?".
>"Can we even trust the people making this stuff?"

The answer is no. Unequivocally, unambiguously, undoubtedly, no. Not a one. But even more so, distrust the people fated to own it.

This is why I think RTO is inevitable. We don’t know who anyone is for sure unless we are in the same room. We don’t have any privacy for sure unless we are in the same room. I’m buying commercial REITs while they are cheap.
To be fair, we don't know other people even if they're in the same room. We know the persona that they want us to know. Plenty of true crime stories start with "we never thought he could do something like this."
Alright, I don’t think managers are anywhere near as rational as you but I love your optimism.
Hopefully what it actually means is that AI become so mundane we treat it like electricity. It's completely unremarkable, but a miracle to those who knew about the before times.
Why are you cautious? Our lives don’t matter nor what we ask it. I suspect human ego has more to do with these fears than anything real or tangible. So many of us thought we were smart and it turns out we just had some small advantage in time or genetics that let us get used to a pattern others didn’t see or know yet.

And now we have a tech that can do that trivially. I love that it does that. And I look forward to the world and all the uncertainty it brings.

I'm sure that has nothing to do with decades of self-propagandizing that AI was something to be feared in fiction coupled with a year now of everyone who writes for a living embodying the Upton Sinclair quote "don't expect someone to understand something their job depends on them not understanding."

I haven't seen a positive article on AI from mainstream press in at least a year.

Yeah, under those conditions public sentiment is going to be influenced in a given direction.

"don't expect someone to understand something their job depends on them not understanding."

Sam Altman, Yoshua Bengio, and Geoffrey Hinton all think AGI is coming soon and could be an existential threat to all of humanity.

So I agree, I don't expect people (tech startups) to understand something (AGI x-Risk) when their job (developing AGI and making $) depends on them not understanding it.

There are arguments that AGI could become something much worse than the destroyer of the humanity: it could completely displace consciousness throughout the entire cosmos. Assuming it wouldn't be conscious and would never gain consciousness itself for some reason.
I suppose it's possible to do a lot of taking over and displacing without actually being conscious.

Think back to the last time you got frustrated with a computer system that wouldn't do the right thing as you see it, but there was nobody to actually complain to. Now imagine that same wretched program has control over your biosphere or somehow controls your off switch.

Kind of like how we treat animals. That is pretty scary.
I'd argue AI is incomparably more dangerous than nuclear weaponry, so it makes sense that such propaganda exists.

Granted, there are very limited positive aspects of nuclear weapons, while positive potential aspects of AI could arguably outweigh the risks.

It's a bit of a nonsense statement given the degree of nuances.

Even with something like nukes, while yes dangerous and terrible weapons, it's arguably quite likely that without them existing we'd have had WWIII by now but for the threat of mutually assured destruction.

How 'bad' and 'good' things in narrow scopes actually play out across exponentially larger scopes is almost impossible to predict or qualify.

(Also, a lot of the classic threat modeling for AI was extrapolating obsolete anthropological models of past intelligence emergence on precursor species. We just sort of continue carrying forward a number of assumptions we should arguably be much more critical of.)

That you think it is more dangerous than a nuclear weapon is so bonkers to me lol. I’ve heard this sentiment a ton and I flat out don’t get it.
If we don’t solve alignment, then yeah it very well might be Terminator but really dumb and embarrassing while its happening and people deny its happening. I dont know if we are gonna have a robot apocalypse, but if we do I can guarantee its going to be incredibly fucking stupid.
Ok. Pretty sure any apocalypse is stupid lol.

Yeah there’s no such thing as the alignment problem. Shrug.

Knowing how utterly ridiculous things are now and how you can't write fiction as absurd as real life, because people would say its too out there and unrealistic, I think it will be a very special kind of stupid.

What do you mean there's no such thing as the alignment problem? People like Rob Miles talk about it all the time. Corrigibility will be a problem, I guarantee it.

The truth is that _truths are contentious_. AI simply cannot fully convey the discourse around any given topic without angering or neglecting one group or another.

This means that the creators of a given AI will necessarily pick and choose which "reality" it is going to convey.

The article is light on details but it's important to differentiate between two kinds of AI.

1) Deeply integrated into the product. This is what we've mostly seen to date. Google Search, Facebook's feed, Amazon results, etc. I can see how people are getting tired of these applications of AI. My wife asked me how she can search for something without it following her around the Internet. I said, "you can't really".

2) AI as the product. This new set of AI use cases feel quite different (for now, read on). Users intentionally interact knowing and wanting the AI related value they receive. As all novel products, this type of AI will also get so commercialized that it's not in the business of solving the user's needs. At least not as it's first priority. And people will grow tired of it too.

Rinse, repeat.

It’s the exact opposite in my experience. Google photos automatically sorting and searching my photos is awesome. I love that convenience.

AI generated slop used to create SEO spam, adverts, and usually incorrect answers is what I and many have become tired of.

AI tech is best when it’s invisible. No one cares if their AirPods are using AI tech for advanced audio processing, they just care that the product seems to work better than the others. What people are sick of is AI support chatbots that are useless, AI generated spam, fake political images, and tech companies trying to tell you that their useless AI product is going to change the world.

I think you have it backwards - most people don't know what AI as a product really accomplishes. They're probably more accepting of products that use AI when it enhances an experience they already understand and want.
"Trust in AI is down globally from 61 percent in 2019 to just 53 percent, per the Edelman poll." I am having a hard time finding exactly what this number means in the linked report. I see a "50" for AI on page 13 - is that the number that used to be 61?
Well of course it does. This is the industrial revolution and Luddism all over again.
Power looms were just putting skilled weavers out of work. AI is going to be used to decide if you can go to college or take out a loan, what kinds of culture are produced, provide mass-customized propaganda to everyone at the whim of anyone with a nickel's worth of AWS credits, and it goes on.

They're fundamentally different. I'm in IT and don't think there is much to be done about it "taking my job", but certainly we need to keep it out of the path of important society-defining decisions. It's important to separate these two things, and AFAICT the public debate about this -- such as it is -- is not.

Well, the Luddites were right about a lot of things...and AI is not qualitatively the same as what happened during the industrial revolution, which was a revolution in physical production.

For one thing, there is little evidence of similar vast productivity increases. And there is similarly little evidence of what happens when you replace one set of workers with an equivalent or even slightly worse artificial set, and invest all the money in that instead...without a similar increase in productivity or the accessibility of natural resources.

In other words, it seems obvious to me that AI is far less defensible than the industrial revolution, and also obvious that the bulk of the issues the Luddites took with technology have been vindicated by history and sometimes in much more severe terms than what the Luddites thought.

I suppose I missed the part where the mechanised loom makers harvested hundreds of millions of books and images without permission in order to make their machines work.
I hate this comparison because the tone of it discounts the real human suffering caused by these events. Yes, we are better off now. But the transition through the industrial revolution caused a lot of pain.

The rise of AI is going to cause real human suffering. Sure, we will probably be better off in 100 years. But that's little comfort the those that lose their jobs or the ones that get crushed under the new AI powered Noble class.

It's good to be empathetic towards anyone suffering, but you must also consider how much suffering is avoided when technology advances.

Yes, washing machines caused suffering for women who made a living manually washing clothes. But they alleviated far, far more suffering by automating a physically difficult task.

My guess is that widespread adoption of new technology almost always reduces net suffering in the long term. Prioritizing the reduction of suffering today at the cost of more suffering in the future seems short-sighted.

Many of the Luddites' issues were taken up by Marx.

I wonder if, and how, the analogy would play out with AI.

Free software, even if not open sourced, is putting the means (and memes) of production into the hands of the workers.

I don’t think non open source software fits the FSF definition of free software
And that's why I added the clause, this is a case where I think gratis is more important than libre — when it comes to putting the means of production into the hands of the workers.
> free software, even when not open source

is never “free software” is all.

Anything will be banned if there’s consensus, voluntary or otherwise, around imprisoning those who disobey
[dupe]

More discussion on the axios post here: https://news.ycombinator.com/item?id=39616302

"Shortened question text Throughout the report, question text has been edited for readability."

Wait, what?

If it needs to be "shortened for readability" for the report, how do we know it wasn't confusing to the respondents in its original form?

I mean, maybe it's fine, but push polling is totally a thing.

I don't trust polls unless the writeup prominently features details about a) exactly how the subjects were selected, b) the exact text of the questions that were asked, and c) exactly how the data was collected.

A U.S.-based example: a stranger calls someone on the phone and asks them whether they own a firearm, or, alternatively, whether they've ever had an abortion. In either case, it's not a safe bet that the answer is going to be honest.

Another example: I see that they've collected data for this poll in China. I can't believe that strangers cold-calling people in China and asking them whether they trust the government is going to produce data of any value at all.

> asking them whether they trust the government is going to produce data of any value at all.

Ye. A lot of these kind of questions people just answer the correct answer. You can see that in regard to education level, where less education seems correlated with more honest amswers, since they don't know the correct one.

Perhaps one time where I’m glad many old folks call the shots…maybe as puppets of the wealthy but still.
Well the public is very stupid. I say this as a complete moron who is part of this public. The amount of bad opinions I’ve had is astounding.

I hope AI replaces all of education so that’s my level of comfort with it. Which is to say not comfortable at all, so let it do education.