Why We Cover AI the Way We Do

0:00Hey, Joseph here. We are on holiday, as I hope you are as well. um this is a re-upload of an episode we've recorded over a year ago i think now but it was just for paying subscribers um and we just wanted to give listeners of our free feed a taste of what they can get if they do become a subscriber it's a discussion about how we cover artificial intelligence as i'm sure you've noticed we cover it in a bit of a different way from other outlets While they may be focused on the companies and the technologies, we're focused on the harms and the use cases and everything else that's happening to and by humans right now.

0:46So I hope you enjoy this episode and subscribe at 404media.co for more bonus content as well. Thank you.

1:10Sorry, that music went on for way, way, way longer than I thought. I was into it. Why does the bonus five music go so hard? I don't know. Well, we saved that exactly for the paying subs. Joseph here, the person you heard speaking with Sam. We also have Emmanuel and Jason. This is a subscribers only episode of the 404 Media Podcast. If you're listening to this, you're a sub. And thank you for that. We're going to get straight into it. You know, we've published a couple of stories over the past week that started or rather contributed to a lot of, let's say, debate around images and artificial intelligence and journalism and tech companies' responsibilities and all of that.

1:50You know, here we're just going to talk a bit more casually about what we think of that reaction, the questions it brings up, all of that sort of thing. Before we get going, let me just provide two very quick summaries of the two stories we're basically be talking about and around them. The first is one that we just spoke about on the weekly pod. You can go listen to it there as well. But that's by Sam. And that's called Bing is generating images of SpongeBob doing 9-11. Basically, people took Bing's AI image generation, which typically stops you from using prompts such as World Trade Center, 9-11, that sort of thing, because Microsoft doesn't want that sort of stuff on its platform.

2:29But people found work around by including sam doing prompts for like kirby flying a commercial jetliner at two skyscrapers all of that sort of thing uh and then the second is one written by a manual which was 4chan uses bing to flood the internet with racist images it is where it says on the tin members of 4chan instead of using photoshop or maybe you know more established tools they would go in they would use things, AI, to generate racist images. And then they were telling their community members, go post this on Instagram or wherever. Right? And then... If I read that on a tin, I'm not buying the product inside.

3:12I'm just saying it's accurate. I'm not saying it's a good tin. I know. If it's baked beans, I want to know what's in it. So we publish those pieces. And then there's some reaction. I think all of you You guys are more familiar with what the actual reaction was. So, I don't know, Emmanuel, do you want to jump in and summarize maybe what the reaction was exactly? Yeah. So, I think, first, it's important to say that the story, and in particular the tweets about the story, went very viral. A lot of people retweeted it, but namely Mike Solana, who is a venture capitalist who has a huge following, retweeted it.

3:58And then Elon Musk, the main character on the platform, responded to him. So it went mega, mega viral. And when you go that viral on Twitter, as you all know, as anyone who has experienced it before knows, it just becomes a mess that has a general shape that you recognize. Like, people make the point of the article to you, assuming that you don't understand it. People thought that we were being critical of the images themselves. We're saying that the images themselves were reprehensible, the SpongeBob image specifically. Right. And I don't think we care about some kind of cringe jokes, to be perfectly honest.

4:43Right, yeah. I mean, a funny image, like a great, great image. I mean, it's 2023. We're long past the time when 9-11 jokes were even mildly controversial. Yeah, I mean, they're controversial, but I definitely... I mean, I saw it and I was like, wow, we should use that image because it's very bravian, funny. And then, obviously, just the whole range of racism and anti-Semitism. I don't know, Sam, if you got any heat for being a woman, but it's like, whatever you are, it's like they kind of are mean to you in that specific way. So that's just the nature of going viral on Twitter. But the response that is specific to this story, I think, in order to understand it, we have to explain that the story is entering an ongoing debate in AI.

5:45And I want to try and map out roughly what the debate is. And there is a spectrum that on one end, there is people who are concerned with AI safety. And on the other end, there are people who are concerned with AI ethics. That's how people in the field describe the debate. I find that kind of confusing. Those are confusing terms. A better way to think about it is just open source AI development and close. That's roughly how the debate maps. And on one far end, like the very far end of the spectrum, are these effective accelerationist AI people. and accelerationism if you're familiar with the idea is whatever is happening you just want to like do whether it's good or bad right you just want to like pedal to the metal go faster reach the logical conclusion of the process right people may be familiar with that term more in the context of like you know extremism and that sort of thing but that's that's not that's not what you're saying here really it's just it's a similar dynamic it is related right so like the origin i think is actually like kind of like a lefty idea which is capitalism is this process we're not going to see the end of capitalism until we rush through the process and reach the conclusion and then we can do something else right a communist socialist kind of alternative and then as you say the like that dynamic was picked up by extremist people who are like we have to like this is kind of like the boogaloo boys thing right it's like and and the charles the charles manson thing as well it's like there is racial strife there's political strife let's have it out let's have a race war and we'll reach the end of the current state of america and then we can have something new in their eyes probably something like a fascist alternative right um in ai the effective accelerationist crowd, which has prominent members, like the head of Y Combinator.

7:53Which is a startup incubator. Yeah. And people are unabashedly effective accelerationists. They have the e-slash-acc tag in their Twitter usernames. And they just think that the only way through the upheaval of AI is through it. And it's just like, we have to just like develop it. No regulation, no restrictions, just let AI do its thing. And on the other end of that is some sort of like AI utopia, right? Because we're going to get all the benefits of AI. We'll go through whatever turmoil society is going to go through. And then we're going to reach some AI utopia. That's one end of the spectrum.

8:35And the other very far end of the spectrum is the like AI alarmist crowd. who believe that some form of AGI is going to kill us all imminently. And for that reason, we have to pause the development of AI. We have to restrict it with the number of GPUs that are able to work on it or ban it altogether. And also, critically, make open source development of AI illegal because it's too dangerous for people to have access to this technology. technology. Now, those are the extremes of each side of the spectrum. In between them are like totally good, cogent arguments for either side. So for example, like some AI researchers that I talked to who study the data sets that the AI tools are trained on and are trying to find the bias in AI.

9:38They say that it's much better to have open source AI because the outcomes might be bad, but at least we're able to see into the black box, understand what is happening, and try to mitigate. Rather than just one company being able to control all of it or private industry or whatever. Right. And behind closed doors, right? So it's like open AI, it's doing its thing. It's doing bad things and we can't even tell you why or do anything about it. So that's a totally good argument. And then a logical argument on the other side is the AI safety people who are like, people are doing harm with this right now.

10:16Actual harm is being done to actual people right now because the tools are open source. And this is something that we reported on, right? Like stable diffusion, text-to-image AI tool. People are using it as we speak to create deepfake pornography, non-consensually, of everyone under the sun. It is happening. Of specific people who did not give their consent. It's not we're going, oh no, they're making porn. It's they're targeting specific people. Right. So that's the argument. That's the spectrum. That's the debate. That's the large conversation that is happening. And then we publish our story.

10:51And I think the way people read it, despite, you know, both Sam and I making very clearly the argument that it's like, we don't know what the solution is. And we are not like saying that each that either side is correct. It's a very complicated question. And I don't have an answer for it. Like, I'm not going to give you my take on which is the correct side of the argument, because I don't know. It's also almost definitely somewhere in the middle. Oh, yeah, for sure. And I think we'll get to that also because the conversation is really stupid because people just assume that it can only be one extreme or the other.

11:31But people just assume that we are on a side that is kind of like wagging its finger and saying that AI tools have to be put behind some lock and key and average people shouldn't be able to use it. And then, you know, people trot out these metaphors around like, oh, would you ban the printing press? Would you ban word processors? Which is the tweet that Elon loved so much. With his little crying, laughing emoji. Yeah, yeah, yeah. His divorced emoji. So yeah, that's kind of the state of the debate. And that's how we entered it. And that's what people assumed about the stories, even though it's simply not what we believe and not what we wrote.

12:12Sam, what do you think about some of the reaction? Or did Emmanuel miss anything there? or what did you see from your end? Yeah, I mean, I saw a lot of people, like Emmanuel said, wishing for the articles that we wrote, begging us to write what we already wrote or just saying, I wish someone would tackle this in a way that's more nuanced. And it's like, you didn't read past the headline and you looked at the pictures. And that's very obvious. We did write quite a bit of nuance into these stories. We always do. We don't just post 200 word, look at this type blogs, we almost always try to provide some kind of context into why it's important to write about these things and also how it's affecting real people.

12:58I think my... And I wrote about this a little bit in our newsletter behind the blog thing, but my thing is these ways that these companies are using AI is impacting the way we use the internet like moderation affects the way people behave uh we that's very obvious good or bad um but you know it's an ai is being used in moderation pretty much everywhere you look at this point um whether or not they call it ai or they call it you know just like machine learning or whatever they want to say it is um that's it's ai is going after people it's going after more ai content. In Bing's case, it's AI going after AI, while Microsoft just kind of like some kind of third party with their hands up being like, we don't know, there's the users, and then there's the moderation tools, and then we're just like, we don't know what's going on.

13:55Or, you know, that's what I kind of, I inferred from their non-response to me about user safety. But yeah, you can't say sex on TikTok without getting banned. So people started saying segs with two g's now that's a thing people say out loud crazy um you can't say suicide on tiktok so people say s-e-g-g-s yeah like to be funny okay yeah like ironically but it's like for explaining to me what people do to be funny me like it's a joke it's to be funny but it's it's to avoid the algorithmic moderation yeah people say unalived instead of suicide that's the thing people say out loud in it's not just like they're avoiding like written bans or avoiding the ai scraping or like reading listening to them without you know considering context and then banning them all right um i think that's kind of the frustrating thing is like people just don't know what is going to get banned and what's not and the way that these rules are applied is so uneven um that who knows it's like i mean i said this in his piece but you can't generate a nipple on bing but you can generate like massive amounts like horribly racist shit that's something is something is wrong there like i don't think that's a controversial statement that like that's a lopsided view of moderating your own platform and this is one of the biggest companies this is like the second biggest company in the world has a huge vested interest in brand safety and it's dropping the ball massively because they wanted to roll out a shiny new toy yeah i mean there's a lot of just like in the word processing it's like it's so it's like it's one of those things it's like it's almost makes sense until you think about it for three seconds and then you're like oh this is a dumb argument to make just to dismiss somebody else's actual concerns so well on the nipple thing just very briefly like even if you don't take again i think the very justifiable position of like well these priorities are a bit out of whack right you you don't allow the generation of a nipple, but you will allow the generation of racist shit.

16:02You know, look, even if you put that aside and you don't even put a value judgment on it, you can just be like, Microsoft at some point made a decision that allowed, you know, and like, maybe it was an accident by Microsoft. Maybe it was a deliberate policy where they have a slide deck hidden in a vault deep inside Microsoft HQ where they're like, we're going to get rid of the nipples or whatever. Obviously, I don't think it's that. I think it's more messy and probably like, you know, a lack of foresight on how these tools are going to be used. But even at minimum, if you don't do the value judgment, it shows that decisions have been made that are impacting the output of this AI.

16:43You know, and it's that at a bare minimum. Jason, you were going to say something? I was just going to say both Sam and Emmanuel are geniuses and just said very smart things. One thing I do want to say is we're doing this podcast on the paid feed for a few different reasons. I think one of the reasons though is a lot of the people who are criticizing us and who are jumping into this conversation, they're very loud and they're influential and they're famous. And we can talk about that in a minute. But it's like, I am certain that tens of thousands of people saw our articles and they're like, that's fucked up or that's interesting.

17:23And then thought it was a good article and read it and continued on their way. And I don't think that we want to get into a position where we are defending ourselves against the loudest voices in the room every single time anyone criticizes anything. because as Emmanuel said, anytime you go viral, there's going to be some criticism, but it goes viral for a reason. And that reason is because people think it's interesting. Most times, we've been doing this for a long time. And most of the time that we have something go viral, it's usually our best stuff or when we've hit on a hot button issue or something that people are talking about.

18:08And I am certain that the vast majority of people read these articles and were like, good article. Thanks for doing it. But then there was a small number of people who were like, you're everything that's wrong with journalism. This is terrible. You want to censor us, blah, blah, blah, etc, etc. And I think because some of the people who were making that argument or have giant megaphones, it can seem like that's what the vast majority of people think. But I want to state for a fact, I don't think that most people are going to see this article about SpongeBob flying into the World Trade Center and think 404 Media is trying to censor Microsoft and the creative tools being used here.

19:02I just don't think that that's the dominant position. And so we didn't want to spend an hour talking about it in our main podcast. But it is an interesting debate and we're happy to have it. It's a springboard for us to talk about some stuff that we don't necessarily always get to talk about, which is more behind the scenes stuff that our subscribers can listen to. I think you just mentioned something about how lots of people read it and just go on with their day while others will almost read a motivation into it. And look, again, I didn't write these two stories, which is why I'm kind of like letting you guys just like more talk it out and you're a lot smarter than me on it.

19:39But I'm super interested in the journalism side of things. And that made me just think that, you know, sometimes I would publish a piece and people would be like, this isn't a scandal. Or they'd be like, they go on the assumption that you can only write about something if it's a scandal, which is ridiculous. You can just write about something because it's interesting. You can write about something because it incrementally provides more information. This, I don't think it's a scandal, you know, but I think it's incredibly interesting that Microsoft is allowing these things even when it actually goes specifically against their policies.

20:16And just... Joe, can I just like refine that? I agree with that. Let's refine it. Let's put scandal aside. Like, I'm not making any value judgment about scandal or not scandal. Microsoft is one of the biggest companies in the world and employs some of the most talented computer scientists in the world. It is a huge center of technological power. And when it deploys something, it has the ability to make millions of people take notice of it and use it. You earlier talked about Microsoft making choices. And I think what is interesting and I think what is important is that what Sam's story showed, and I think what the 4chan people have abused, is that it is not making choices as much as Microsoft think it is making choices.

21:26It is deploying the tools and it has written all these rules for how the tools can be used, but they don't work, right? Even according to their own policies. Their own policies don't work. And something that I really liked about Sam's story, like a really good observation, is that it's like essentially there's this new term that appeared in the past year, prompt engineer, right? And people are prompt engineering the tool to do things that Microsoft doesn't want it to do. Whether it wanting it to do it or not is a scandal, you decide, dear reader. But this giant, powerful company deployed this thing that it can no longer control.

22:15That, to me, is the interesting thing. And as Sam pointed out, it is not clear that they will ever be able to control it because it's not clear that while we can teach AI to understand language, it's not clear that we can make it understand cultural context, which is how people are hacking the tool, right? And it's already out there. Humans outsmarting the machines with our cultural context. Yeah, it's true. I mean, that is what's happening. Yeah. Let me just super briefly end that point, because, yeah, the scandal thing isn't coming from me. That's just what people have told me. But it's more just that people assume that when you choose something to cover, you're taking a side.

23:06And that's not what it is, is that we think there is a public interest in writing about that. Sometimes the public interest is very high when we publish an investigation. sometimes i i would call it a normal amount which is something here but we're like hey this doesn't match microsoft's own policies like that's kind of wild and sometimes it's like lower but still very beneficial in that i don't know you know jason often gets these really entertaining funny foyers about police officers doing stupid things and don't get me wrong they're funny but they also show, you know, maybe a systemic issue with the police or something like that.

23:45When you write an article, you're not going out to be like, I'm going to use this as ammo for my side. It's like, there's a public interest there. And like, that's almost like the start and the end of it. And I think people forget that. We're not going out trying to censor the AI. You know, we're going out because there's a public interest in highlighting whether this information is. Well, people say it about any article you do, no matter what. I'm making it super broke because I'm tired of this shit. Like the delivery robots story I did where this delivery robot was giving footage to the Los Angeles Police Department.

24:19Vast majority of response to that was like, wow, I didn't know this. Thank you for writing this. This is interesting. I see these robots all the time. There are so many people, though, who were like, you didn't know that? How is this news? How is this new? And I'm like, it's new because literally it was unknown. Like this is the, here's the specific case. It's the first time we could assume or infer that, yes, there's cameras on these things. Maybe the LAPD would want it. Now we're showing that it's happened. And there's this huge contingent of people being like, oh, you didn't assume that there was a, you know, if there was a camera, the LAPD is getting it.

24:58And it's like, you could guess, you could. Yeah. That's why I FOIA'd it. I foiled it because I thought maybe this is happening. And now we're showing the public that it's happening. And it's like, I'm very sorry that you personally do not feel that this was worth writing about. But lots of people did. Lots of people found new information from it. And I'm sorry if we write similar, as you said, incremental type stuff. Sometimes people don't get it until the 30th time you write about it. Sorry, maybe they didn't see the other 25 times that someone wrote about ring cameras and police or whatever.

25:39And that's fine. Not everyone consumes everything all the time. I don't know what they want you to write. That blog would just be like, I think that the cameras on the robots are watching me. We need to prove some things sometimes. And the way to do that is getting the actual documents or seeing other people do it in the wild. or we can't just make shit up. We can, but it's our website. We can, but our lawyer wouldn't like that very much. Yeah.

26:19As someone who's helping to run a startup, we've used all kinds of software and things quickly get siloed and confusing. Then we found Coda. Rely on Coda to keep your team on the same page by bringing the best of documents, spreadsheets, and apps into a single platform. Coda helps you collaborate by centralizing all of your processes and shared knowledge in one place. It cleans up the clutter with its all-in-one solution that replaces a host of siloed productivity tools. With Coda, you can stay aligned by managing your planning cycles in one location. Set and measure OKRs with full visibility across teams.

26:58Communicate and collaborate on documents, roadmaps, and tables instantly. And access hundreds of templates to get inspired by others in Coda's gallery. Coda empowers your startup to strategize, plan, and track goals effectively. Take advantage of this limited-time offer just for startups. Go to coda.io slash 404media today and get six free months of the team plan. that's c-o-d-a dot i-o slash 404 media to get started for free and get six free months of the team plan coda dot i-o slash 404 media this holiday season give a gift that lasts a lifetime it's masterclass with masterclass your loved ones can learn from the best to become their best masterclass is the only streaming platform where you can learn and grow with over 200 of the world's best.

27:50That's why Wirecutter calls it, quote, an invaluable gift. I'm trying to get a head start on my New Year's resolutions. That means I'm learning how to eat intentionally from Michael Pullen, fitness and wellness fundamentals from Joe Holder, and yoga foundations from Donna Farhi. I do a lot of my learning while I'm out and about, so I appreciate that I can take a masterclass on my TV, but I can also do an audio-only version while I walk my dog or exercise or watch these classes from my phone when I'm at the doctor's office. Plus, there's no risk. Every new membership comes with a 30-day money-back guarantee.

28:27Give your loved ones a year of learning with Masterclass. Masterclass always has great offers during the holidays, sometimes up to as much as 50 % off. Head over to masterclass.com slash 404pod for the current offer. That's up to 50 % off at masterclass.com slash 404pod. Masterclass.com slash 404pod. Let's face it, after a night with drinks, I don't bounce back the next day like I used to. I have to make a choice. I can either have a great night or a great next day. That is, until I found pre-alcohol. Z-Biotics pre-alcohol probiotic drink is the world's first genetically engineered probiotic. It was invented by PhD scientists to tackle rough mornings after drinking.

29:14Here's how it works. When you drink, alcohol gets converted into a toxic byproduct in the gut. It's this byproduct, not dehydration, that's to blame for your rough next day. Pre-alcohol produces an enzyme to break this byproduct down. Just remember to make pre-alcohol your first drink of the night. Drink responsibly and you'll feel your best tomorrow. I love seeing family over the holidays, but it also means that I drink a little bit more than I'm used to. I've been taking prebiotic before some holiday parties, and I definitely noticed a difference next day. Even after night out, I can confidently plan to live my life like normal because I wake up and I feel good and I just get on with my day.

29:53With the holiday season upon us, I know I'm going to be consuming a bit more alcohol than usual. With pre-alcohol, I can stay on track and not let the season throw me off course. Go to zbiotics.com slash 404media to learn more and get 15 % off your first order when you use 404 Media at checkout. Z-Biotics is backed with 100 % money back guarantee. So if you're unsatisfied for any reason, they'll refund your money, no questions asked. Remember to head to zbiotics.com slash 404 Media and use the code 404 Media at checkout for 15 % off.

30:35Let's bring it back around to AI a little bit, rather than the, here's 10 years of stored beef against people. And I don't beef anymore. I'm Namaste. I'm Zen. I can't wait to the next time you beef and I'm going to take this clip. No, absolutely not. It's not going to happen. so one reader who wrote in was Tim Sweeney founder and CEO of Epic Games you know they make Fortnite all that they have the Epic Games launcher competitive to Steam they have a lot of money Eman maybe explain it in a minute about what they do in AI because I'm a little bit unfamiliar on that but let me read out Tim's Twitter DM to me after he read some of this coverage because again it's a useful uh springboard to talk about some issues um he says the latest 404 ai articles have been irking me so i thought about why at the core they are written to be inflammatory while taking an implicit but not clearly articulated editorial position about who is to be considered morally responsible for content created by a human using AI prompting.

31:50And then he provides some positions they could take. The first would be, you know, it's the human who prompts the AI as responsible. It's the creator of the AI, or maybe it's the government through regulation. His message then continues, without clearly taking an editorial position, it feels like the debate is playing out in the Twitter court of bad analogies, capitalized. And he says he would love to see a 404 editorial on the topic of responsibility attribution, which seriously weighs the implications of those three positions he laid out. And he says, this might drive some productive discourse about principles rather than Microsoft Word comparisons and the like.