I for one am somewhat glad we are going to be forced out of our shell when it comes to the nature of truth in media. All it means is that people are going to have to have a higher standard for what they believe. Many wont, just as many today think infowars or Joe Rogan are telling them the truth, but we shouldnt cater our society to the lowest common denominator. Instead we should create a society where the more informed people can do their best to guide and educate the people more easily fooled to develop the skills to be more discriminating in what they believe.
A large enough change in degree becomes a qualitative change.
If what once required nation states budget is suddenly free and in your pocket, it’s potentially a tectonic shift.
Fooling the easily fooled accounts for most social media and ad traffic already. So now they get more fooled or fooled more often? I think thats not good but also not catastrophic in the way that is being implied.
That works for you, but for everyone else who is less diligent in source evaluation, this now gets worse. Don't you agree that society suffers when a large proportion of people now get their information from fake sources?
People should strive to be more critical, if that includes evaluating the source of whatever they listen to then I think thats great. This is something we should be doing anyway.
I can be as critical as I want, if the people around me who vote, have influence on my environment, etc end up in misinformation, it has a huge impact on my life.
Advocating for critical and thoughtful consumption of media will become more important with the rise of AI and it will also lift up society as a whole. If society really takes this seriously we could make real progress and help many people who are currently vulnerable to misinformation become less so.
Honestly this an absurd take, and I like AI…
I personally have seen two or three cases of myself or someone I know getting duped by fake info, about meaningless trivia, but just the result of a quick google like "is so and so dead" or "did so and so break up" and got wrong info, because I don't care enough to do the deep dive to invalidate it.
This kind of argument rubs me the wrong way. It's like saying a printing press is just a faster scriptorium, or a nuclear bomb is just a change in degree from a hand grenade. It's misleading because it ignores the monumental difference in society these later inventions brought about, which were not caused by the previous inventions, and never would have been. It ends up being an argument to not engage with the reality of the situation.
The more complicated reality is that things qualitatively change at scale. We need to have honest sober discussions about the trade-offs and unknowns involved, instead of just YOLOing blindly into moving fast and breaking things.
With these image generation models, the only things filtered by human involvement are the creation of the prompt, and whether or not they believe any generated image is believable. The required time and effort is minimal, and on top of that you can generate hundreds by the time a single person can make a convincing Photoshop edit.
---
In terms of being able to sift through generated images across the wider internet, I've no idea what can be done. But I feel a potential defence against them would be for news sources to publish the exact images they were sent online, keeping the original resolution, removing cropping, and retaining Exif data.
Not to get too political, but covid and the 2020 US presidential election showed how bad we are at handling untrustworthy information.
Just like between a fork and a nuke, not sure it's a very solid argument
But not in an automated way, that is difficult to detect, and that could impact every facet of your life. Like, yeah, photoshop was a thing, but now AI filters can "nudify" virtually any picture, quickly, and in an automated way. Ditto for generating ads, political propaganda, etc.
We can deplore it, but must not delude ourselves into thinking that the old web didn't have a nonsense problem. You had sites that would go unlinked because the creators were crazy. At first the thing that curbed influx of nonsense somewhat is barriers to entry such as cost and technical competence. Then the curation was reputation based with Google's pagerank as an example.
Now we will find new ways. And it hurts. It hurts like spam hurts email.
Of course the problem lies exactly in the ability to automate incompetence.
It's not over, unless Google doesn't care enough about Google News to fix it.
Most media has been coercive throughout my lifetime through the capture of distribution channels. If you control the distribution of where people consume their content, you can sway public opinion/sentiment.
For example, you might set a "narrative target" for your news outlet to report "Tech = Bad" this quarter.
Or you might purge all rural content from your network, like in the Rural Purge.
The internet has blown that wide open. Many people, especially younger demographics, are curating their own content now. Instead of "news networks," many of my younger friends talk about individual citizen journalists. Those journalists go by many names, vloggers, podcasters, youtubers, etc.
I've noticed that, in the past few years, the people proclaiming the loudest that "Journalism is Dead!" are doing so because they've ignored where the craft of journalism has migrated to: the internet. They look at the hollow dying husks of organizations that were built around capturing distribution channels, husks that have burned their credibility over years of blatantly coercive reporting, and somehow feel nostalgia for those husks.
The legacy distribution channels have, by-in-large, lost. The only metrics they appear to be relevant in are the ones that conveniently exclude the distribution channel they've lost to: the internet.
With the rise of OSS AI, I'm _extremely_ bullish on personalized moderation and curation.
In the next few years, I want a personal AI that automatically mines all of my social media feeds, email, text messages, calendars, subscribed channels, podcasts, etc. and gives me a personalized curated view.
I want that AI to be highly-biased, and highly-aligned. But the measure of bias and alignment is _me_. I want it to have my bias. I want it to have my alignment. I want it to be fine-tuned on me. And I want it to go churn through a lifetime of content daily and bubble up the most valuable/meaningful content for me.
I think we are moving into a period where the ability for people to be coercive by controlling media distribution channels is coming to an end. The rise of personalized AI will kill it.
And this is before I get into generative AI disrupting the larger entertainment industry for the better!
Locally, on my devices. Where property law says I own the fine-tuned model and the device.
> And who exactly will create the content?
Anyone and everyone? I don't understand what you mean. What you just posted here is content. You produced content. I assume you will continue doing that moving into the future.
> Your personal AI will just consume a shitload of content that the powers with most compute generates
I think that assumes a random walk through the internet? Content discovery probably shouldn't be random?
As your AI curates content, you give it ongoing feedback about the quality of the content it gives you, (feedback based on your highly biased definition of "good" and "bad" content). Your AI learns what is "good" to show you and what is "bad" to show you.
As your AI finds "identities" and "content channels" that produce reliably high-quality content that your highly-biased and highly-aligned AI can curate for you, I'd imagine your AI would prioritize searching those content spaces for quality content first.
For example, I'd like to think my personalized AI would learn to surface up interesting conversations and content from Hacker News.
As content channels get "overrun" by noise, the AI is filtering out the noise, the signal-to-noise ratio drops, and the AI slowly lowers the priority of those content channels in favor of channels that consistently produce content you find valuable to consume.
The key here is that the AI is ultimately accountable to me, overseen by me, controlled by me, fine-tuned on signals generated by me, etc.
Imagine you had an army of 1 million personal executive assistants who could scour the world looking for information relevant to you, and give you a curated feed of useful information on an ongoing basis. That sounds like a pretty cool future to me.
And if someone wants to spend their time squandering their wealth to accumulate enough hardware to run 1 trillion personal executive assistants to generate a bunch of valueless (or value-negative), coercive, uninteresting dribble on the internet? Cool. Let them squander their wealth. My executive assistants will filter that out for me.
Personally, I want to be challenged regularly. I want AI's that are biased based on the direction I need to be challenged in the moment, which I would mark as "good." I want an AI that can challenge my assumptions as I work by surfacing relevant research papers that show I might be following the wrong path, which I would mark as "good."
I want AIs that can adopt the personas of someone I'm reading that I disagree with, who has long been dead, so we can have a conversation over beers and sort out some of my personal philosophy.
And I want all of that on my terms.
To ground this desire in classical terms, I want to pick the books I read. I have a bookshelf full of books. I have books I agree with. I have books I disagree with. But I only have books that make me think. Being able to pick my own books doesn't mean I want to live in an echo chamber. It means I want to cultivate my own personal philosophy on my terms.
I mean, you can always just not rely on the internet for news and narratives and such.
And yes it is possible. It is possible to work in this economy, raise a family, eat well, and not need to be bombarded by fake anything, whether generative AI or human authored.
I think the atomic bomb is more credible as a seed of human destruction.
By the way does anyone know anything about any launch codes?
Also - have you seen the whereabouts of John Connor?
That commenter went on to list examples where one’s personal usage of the Internet would irrelevant to one’s wellbeing. It won’t affect one’s level of care received from doctors or one’s safety when crossing a bridge.
it's not me I worry about
it's all the facebook users, who can also vote
Nuclear weapons absolutely are an existential risk.
>By the way does anyone know anything about any launch codes?
I don't have any top secret knowledge about current systems but have read about historical command and control systems and it's a terrifying slice of history. Now there is starting to be talk about integrating AI into those systems, which I find horrifying.
Ok, but what do I do about the other 159,690,456 people that vote in this country?
I choose to be naively optimistic: the glut of reasonable-sounding bullshit will increase the practice of citing sources when making claims. Then we can go back to good-old-fashioned lying with statistics.
This isn't hypothetical. We already have bullshit citations and bullshit articles today. The only thing remaining is to stick the two together.