Loading summary
Transcript1 lines
- [00:00]
A
Duckduckgo, the very famous privacy web browser competitor to Google, has just announced that they are blocking AI images in search. Now, there's a bunch of interesting things about this. They made this announcement. You're actually able to opt in or opt out. And there is some precedent for why this has happened. And they have been sort of making fun of Google in the way that they have announced this, essentially. So this is a search engine that has been famous for privacy for a long time. People love DuckDuckGo, it's a great competitor to Google. Based off of blind tests I've done, the results are pretty great in comparison to Google. So I think this is an interesting move. But beyond just like, oh, wow, look, they've blocked some images. I think this is going to have some huge implications for what everybody is doing in image, search, text, video, audio, a lot of different areas. So. So we're going to get into all of that today. Before we do, I wanted to mention if you've ever wanted to try the top 40 AI models without having to have subscriptions to all of the different platforms, which I know is something that has always sort of driven me crazy, I'd love for you to try out my startup, which is AI Box AI. So I have the. It's a platform where essentially it's like ChatGPT, where you have access to the. All the top AI models on there. You can switch between them all in the same chat thread. So you can talk to ChatGPT, or you can talk to Anthropics Claude or Google Gemini, or all the top models Grok, including image models and audio models, all in the same thread. It's super useful. It's $20 a month and you get access to everything, a bunch of cool features. It's in beta right now. It's $19 a month. And I would love to hear what you have to say if you want to try it out. All right, let's get into what's going on with DuckDuckGo. So I think the reason, like setting, kind of going back a little bit and setting the precedent for why this is happening, uh, Google got a lot of flack back in October of last year. Someone did a search, and I mean, this applies to like, a lot of different things, but specifically, someone did a search for baby peacocks, and they found in a search for baby peacocks, of the top results, there was three of them that were or two of them that were not AI generated, and there was like seven of them that were or eight of them that were AI generated. So of the top 10, two were real and eight were AI generated, which obviously is pretty insane and kind of concerning. When you start thinking about education, a lot of things, you, you're like, what the heck does a baby peacock look like now? No, now people are seeing this, this fake AI generated stuff. And I think this applies to a lot of different things, right? So Google got a lot of flack, which is the, the only reason I bring up this whole peacock example is because it's kind of funny. When DuckDuckGo made their big announcement, they were over on X tweeting about it. And, and in their example, like the example that they shared was the keyword baby peacock. They didn't specifically call out Google, but basically they called out Google. So what's interesting here, this is their exact tweet. They said a new setting, which is hide AI generated images. And DuckDuckGo, or philosophy about AI features, is private, useful and optional. Our goal is to help you find what you're looking for. You should decide for yourself how much AI you want in your life or if you want any at all, essentially what it is. When you're doing a search, you have a toggle you can drop down which is show AI images or hide AI images. It will completely hide them. How are they doing this? What does this mean in the future? This is what I thought was interesting. They said the filter relies on manually curated open source block lists, including the quote unquote nuclear list provided by Ublock Origin and you blacklist huge AI block list. So while it's not going to actually catch 100% of AI generated results, it will greatly reduce the number of AI generated images that you see. Or what's interesting here is they haven't generated specifically like a, they haven't gone through and like actually generated a technology that does this, which I found pretty interesting. They're relying on someone else that has kind of created these block lists or these ban lists of images, which means that like new stuff could get through and isn't going to get shown. But I think basically how this works is people have been scraping the Internet. They probably relied on a lot of manual humans to go and flag images, and then on top of humans flagging images, they train probably trained models to be able to understand that. And then they went through and make this giant data database of every single image that is AI generated, which is super, super useful for a company like DuckDuckGo to add. Now, in the future, I think DuckDuckGo is going to need to build something A little bit more robust, probably in house, maybe working with some open source libraries or tools like this. But at the end of the day I think they can't just rely on a third party company to do this as it's obviously not super reliable. Like if I make a brand new AI generated image of Trump with a machine gun on top of, you know, the White House and make some fake news website about how Trump turns into a crazy terrorist and I get that thing to, you know, for all the memes it gets SEO to the top of images. So when you search for Trump, that's the first thing that shows up and people are like, oh my gosh, is this real? Blah blah, right? So like it, it could get through very easily. And so I think this isn't like an end all be all but I think it does have a huge list of, you know, known fake AI generated images and hopefully we'll be training on more. So it should be kind of good at filtering but definitely not perfect. So just putting that out there. A couple interesting things about using this tool from DuckDuckGo so you can, like when you do a search you can go and toggle the filter off of all the images, but they also have an, basically like a, a URL which is N o a I no AI.duckduckgo.com so if you go to Noai duckduckgo.com and do any searches, it's automatically by default turning off all of the settings. Now this is image generation. And the reason why I want to do the story today is because I think there's way bigger implications than just image generation. First of all, this is fantastic. I'm actually kind of excited about this. I think this will be quite useful. But I think when you start looking at this for, for essentially what we could do with text, you know, blogs and stuff, Google has actually been doing a, I think probably a decent job of trying to find websites that are AI generated. I hear a lot of people saying that Google search is full of AI slop blog posts and my experience with this in my background, which I don't know, I'm sure some people will hate me or whatever or get mad at this, but when I was doing my last startup, which was self pauses before ChatGPT ever came out, I discovered an AI generating tool. It was Jasper and I was generating, I think I generated over 3,000 articles for that website. My website self pause and had them all posted. They're just like, like every question you could ever possibly think of around, you know, SEO around affirmations and, and meditation, that kind of stuff. And I had thousands of articles. Now inevitably what happened was a manual review came through. So some person reported my blog post and you know, back then I was trying to make them really good. This was like guys, this is before ChatGPT, so you got to cut me a little slack. I thought I invented the greatest thing. I was paying like people to go write these articles and generate them. You can only do like a couple sentences at a time. And so anyways, I had this whole flow and we're adding frequently asked questions and images and backlinks and all this crazy stuff. I thought I was a genius. But I mean inevitably it's probably low quality content, especially because this is like an old, old version of Chad GPT before Chad launched. So it's like, wasn't even, you know, it's like the oldest responses of hallucination that you could get. But in any case, at some point after ChatGPT came out and probably a year later, I was getting, you know, thousands and thousands of web clicks a day from Google and at some point someone manually reported the site and it got flagged and shut off. And that site hasn't really recovered in the Google search results. I haven't done much to salvage it because I've been working on my company AI Box. I haven't really been, you know, I haven't done anything really with self pause since then. However, it was an interesting point to me that Google is looking like they're trying to penalize some of these AI search results if they're low quality and, and they'll essentially just pull the website out of the search results if it has all that quote unquote spam or whatever you'd want to call it. So it's interesting to me, DuckDuckGo is kind of doing this thing where you can toggle on and off the images. Google inevitably is kind of blacklisting. But Google's blacklist was a manual review. So like if you just use a few here and there or you're using AI to help augment you know what you're writing or probably if you're just more clever than however, you know, I did it and you have a newer version of Chat GPT and you're getting articles rewritten when new models come out, give better results, I think that there's probably, you know, millions and millions of sites that are not getting penalized that are still showing up. So it is kind of a cat and mouse game. It is kind of tricky and I think maybe instead of, you know, I think when you start thinking about this, like, yes, it's cool that you can just turn off AI generated images in DuckDuckGo, but you don't want to by default never see AI generated image because there's plenty of viral ones. With blog posts, I think it's probably a little bit different. Um, except that a blog post, what is it? It's basically like a, a static Chat GPT response. So I, so I guess it's not necessarily that it's like an AI generated blog post is bad because it's just the same result that Chat GP would give you. And maybe someone's not using chatgpt, they just want to use Google for whatever reason, so they get that the only downside to the blog post is when ChatGPT gets an update, the response gets better, theoretically, and it, you know, responds better and it's more dynamic and it, you know, and these things get more and more intelligent, you know, with the latest, Grok4 is now as smart as a PhD in like every area. And so I think the blog posts, honestly I think are just an outdated model of consuming information when ChatGPT can, can change. Now, are there places for blog posts like I think are there. Is this still a useful medium? Yes, I think when you're working on something in your own home and you're taking screenshots of your workflow and what you're doing and like you can imagine these things being very useful in a case study on something you've done that ChatGPT has never learned. You're breaking a piece of news or sharing a piece of information that no one else has. Yes, I think there's a great place for blog posts then. But if you're just answering frequently asked questions about a topic that there's like 10 other websites, blog posts probably aren't very useful and you could probably get that same information from ChatGPT. So like, when do you go to blogs versus when do you go to chat GBT? Inevitably, blogs are all going to get sucked into ChatGPT. If you use a tool like Perplexity, it's going to go do their research and validate the information. So I think people still write, keep writing blogs and sharing that information, but if you use a tool like Perplexity, you don't really need it. So, yeah, anyways, it's just a fascinating time with AI to, you know, to decide what level. And the reason why this question is important is because we're doing this with images right now. But the next step will be video. And then of course we have audio and we've kind of already went through this with its own version of text. I think video may be somewhat similar to image, but at the same time, like there's viral AI generated images that are, you know, super funny memes that somebody made and you don't want to just like never see those anywhere, right? So it's like you can't just like block blacklist those all from the Internet forever because there's funny, interesting uses or reasons why someone might generate an image. Anyways, it is a very interesting time with content creation and with just everything on the Internet. I think at the end of the day it would be cool to have a complete toggle on your browser in Chrome or anywhere else. It just completely blocks all AI generated stuff. I think a lot of people would like that. But maybe that there's something, something somewhere in the middle a little bit more, which is like block, like most AI generated things. Unless it's like super, super popular. A lot of people are talking about it, right? Like, I don't know. It's going to be kind of interesting to see where, where it all lands on this. But definitely a question question that these big browsers and search engines. Search engines are going to grapple with, especially with DuckDuckGo taking the lead here. You know, Google's got to be thinking about it and a lot of these others are going to be thinking about it because nobody wants one of their competitors to come up with a feature that seems more useful and draws any of their users away. So should be interesting to see where all of this goes. Thank you so much for tuning into the podcast today. If you learn anything new, make sure to leave a rating and review wherever you get your podcasts. And and as always, make sure to go check out AIBox AI if you want to try the top 40 AI models without having to get subscriptions to all of those platforms. For 20 bucks a month, you can use them all in one chat interface. It's super useful. Switch between models while you're talking to them. I hope that you love it, but let me know what you think the link is in the description. I'll catch you in the next episode.