Hacker News new | ask | show | jobs
by rjf72 2684 days ago
Wow. I am curious how "specialized" the training was because those sample responses are beyond remarkable.

I think we're going to face a lingering question with AI. We're imminently reaching the point where AIs will be able to generate fake everything. In the near future (if not present!), I could be fake for all you know writing lots of otherwise coherent posts, only to secretly jam in some sort of agenda I've been programmed to advocate for. And there could be millions, billions, an unlimited number of "me". Or the latest hottest site trying to sell itself on its own 'buzz' could be full of millions of people actively engaging on the platform, except none of them actually exist.

So do we try to keep these AI systems secret, or do we make them widely available and rely on a rapid shift in public consciousness as a result? It's one thing to try to tell people to engage in sufficient scrutiny over text, images, audio, and increasingly even video. It's another when people see that such fakes are trivially produced by anyone.

I do realize that the 'chaos scenario' sounds... chaotic... to put it mildly, but I think the underlying issue here is that these tools will reach the public one way or the other. By keeping them secret the big difference is that the public will be less aware of the impact they're having, and the players operating such tools will be disproportionately made up of people trying to use them for malicious purposes - be that advertising, political influence, or whatever else.

3 comments

I think the sooner this kind of thing gets out in the open, the sooner sites being propped up in valuation based on fake users today can change.

On another note, think back to how general people responded to things like Eliza and other chatbots, or the sims and other emergent storytelling games. What if there was a "social media" platform where all your connections were purposely AI, like your own personal TV sitcom/drama/comedy/whatever. Surroubd yourself with people who think like you that you can finally have "intelligent discussions" with without heated arguments from all the idiots who disagree. People love gossip; what if you had an endless supply from fake people? "You'll never believe what Frank said to Janice!" "I thought Bob and Alice would be together forever."

A social media that is part Her, part Westworld. I think that could be a multi billion dollar idea.
Arguably, this is the best and the worst idea I've ever heard on HN.
A novel by William Hertling titled "Kill Process" follows the story of a startup building such a social media. It is depicted in the positive light.
> And in her ears the little Seashells, the thimble radios tamped tight, and an electronic ocean of sound, of music and talk and music and talking coming in.

-- Ray Bradbury, "Fahrenheit 451" (1953)

> Having lived through the horrors of the pre-smartphone days

> the agonizing pain of [..] standing in line at a grocery store

-- HN, 2017 ( https://news.ycombinator.com/item?id=15912378 )

"intelligent discussions" with things that have the biases of their creators? no matter what you do that will filter in, guess we truly are heading towards an age where the engineer decides what society [will] think
I'm not saying it's a good idea, but from my experiences on social media (forums, mailing lists, chat rooms, and the modern facebook/twitter/etc), most people want an echo chamber, surrounded by people who can discuss whatever event but generally agree with their opinion.
What does fake even mean? The AI exists. What’s the difference between a human doing disinformation and an AI doing it?

You should already be skeptical of every comment. Looking for what it’s agenda is.

The cheapness and thus commality of AIs doing it will just wake people up to fact they should not have been so trusting all along.

The difference is between 50% online being disinformation and 99.999999% being disinformation as content is overrun with cheap bots. Think of what happened to Usenet except worse. Instead of forums being overrun with ads for V1agra, social sites with be overrun with fake members posting reasonable sounding stuff that is laced with coordinated tactical falsehoods.

I think eventually every piece of information will have to be digitally signed and our devices will by default limit what we're exposed to based upon whitelists.

You've more or less described online discourse right now.
The difference is that ai is scaleable.
We are lucky in that we can still read a comment such as yours and know it is a human. I think. Maybe not.