The AI Search Engine War (And Other AI News)

Well, it was another really busy week in the world of AI, and I'm going to do my best to break down all the stuff that I found really interesting from this week. We got a handful of announcements out of OpenAI, including the fact that they just introduced their new ChatGPT search feature. Now, this was a feature that they showed off months and months ago. They rolled it out to like a small select group of people and now I guess finally they're done testing it and it's being rolled out to all ChatGPT Pro users. Now, if you've used something like Perplexity at all in the past, this new ChatGPT search will seem very familiar. Head on over to your ChatGPT account and you'll notice a new little icon right here in your search box that says Search the Web. So you can ask it a question like who are the best YouTube channels to follow about AI? We'll make sure our search button is selected and we'll click the little blue arrow and after a few seconds we've got our search results. Two Minute papers. Oh my. Matt Wolf is listed number two here. That's pretty cool. Deep Learning AI AI explained Lex Fried. And you know, I would agree with a lot of these picks, except for that Matt Wolf guy. He's a bum. Now, if you want my opinion, I still kind of prefer Perplexity. It just kind of gives more detail with its response. If I ask the same question to Perplexity, you'll notice a little more detail. Like technical learning channels, Deep learning AI Send text, Two Minute papers, Programming and implementation. The PIE coach Nicholas Ronette News and Update channels. Matt Wolf all about AI the AI advantage in Depth Analysis. Lex Friedman AI explained. And you can see it's just got that context. The ChatGPT version doesn't seem to show as much information. It just kind of gives those results. But I do believe that that's going to improve over time. With Chat GPT. This is just the sort of very beginning of what they're trying to accomplish over there. But if you were to ask me, I'm probably still going to perplexity. First. ChatGPT also rolled out a feature this week that lets you search through your chat history. So if I go back to Chat GPT over here, you'll notice this little search chat icon in the top left here. If I click on that, I can search Quantum computing and you can see all of the times I've asked about quantum computing in the past. Apparently the last time I did was on April 13 and March 22 before that. So just a handy little quality of life feature that was added to Chat GPT. Now last week OpenAI rolled out Chat GPT on desktop for Windows. It had been out on Mac for several months already. Last week they rolled it out on Windows. This week they actually added the advanced Voice feature in there. Now you can download this new app on both Windows and Mac and when you download it, you'll notice that there's a little use voice mode here. This wasn't on the app when it came out last week on Windows and it definitely wasn't on the Mac app before that. This is a brand new feature they just rolled out. Now. If you downloaded the app last week, you might not see this icon. If you don't see it, make sure you update the app. So update it from the Microsoft Store or the Apple Store or wherever you downloaded it for. Make sure you have the newest version. Otherwise it might not show up. But you could use this just like you would use it on your mobile phone. If I click this button, we get our familiar little blue cloud bubble here and I can say, hey, how you doing today ChatGPT.

Hey Matt, I'm doing well, thanks for asking. How about you?

What's new in your world? I'm doing good. Just recording a AI news video to wrap up the week.

Nice. Any exciting AI news this week that's making waves?

Yes, there are, and I'm gonna keep telling you about em right now. A handful of the leadership team over at OpenAI didn't ask me anything over on Reddit, and we could see it included Sam Altman, Kevin Weil, the Chief Product Officer, Mark, the SVP of Research, Dharinivas Narayanan, the VP of Engineering, and Jacob Pachocki, the Chief Scientist, all participated in this ama. Now, I'm not going to go through all of the stuff they mentioned. I'm just going to point out a few of the things that rose to the top of the ama. For example, somebody asked, when will you guys give us a new text to image model? Dall E3 is kind of outdated. Sam Altman said the next update will be worth the wait, but we don't have a release plan for it yet. Somebody else asked Release date of ChatGPT5 or its equivalent? What are its features? Sam Altman said, we have some very good releases coming later this year. Nothing that we're going to call GPT5 though, so don't expect GPT5 in 2024. And then the one that I found the most interesting, somebody asked is AGI achievable with known hardware or will it take something entirely different? And Sam came back and said, we believe it is achievable with current hardware. So they believe that with the hardware that's out there right now, we're going to get to AGI. I think that's pretty fascinating. Oh, and one more thing here. Somebody else asked, will Chat GPT eventually be able to perform tasks on its own, like message you first? And Kevin Wheel here said, in my opinion, this is going to be a big theme in 2025. A big theme over at OpenAI for next year is AI agents. All right, moving along to Anthropic because they had a few announcements this week as well, including the fact that on the Anthropic mobile app you can now dictate with your voice. And in my opinion, this was the biggest downfall of the Anthropic app is that you could type the old fashioned way, but you couldn't speak to it like you could chatgpt. So when I'm using my mobile, I tended to either use ChatGPT or perplexity because I like the voice feature. But when I'm at computer I tend to use Claude more because I like the way it responds a little bit better. But keep in mind this is just a voice dictation feature. So you could speak your question to Claude, but Claude is still going to respond to you on the app back in text. It's just going to hear what you said and then respond with a text message. It's not like ChatGPT yet or even like Perplexity yet where it will actually respond in voice as well. In other Anthropic news, they actually rolled out a Mac and Windows app this week as well. Last week OpenAI rolled out theirs to Windows. This week Anthropic rolled out theirs to both Windows and Mac. I did download it. If I pop it open here, you can actually see that it is pretty much identical to the web user interface. We just have, you know, an icon to open it up now instead of opening it up inside of the actual browser. In some other interesting news related to Anthropic, you can now use Claude inside of GitHub Copilot. Not only that, but you can actually use gemini inside of GitHub Copilot as well. And the reason I find this interesting is because GitHub is a Microsoft company and as most of us know by this point, Microsoft is one of OpenAI's largest investors. And so it's very interesting because GitHub Copilot was for a long time using OpenAI's technology because, well, Microsoft and OpenAI have that relationship. But now Microsoft is really starting to branch out and even using competitor models like Google's Gemini model and competitors to OpenAI's models like Anthropic, I just found that really interesting. So if you use GitHub Copilot, you'll now have the option to choose which model you're actually working with as it helps you write code. Today's video is sponsored by Hostinger. Now, in my opinion, I think everybody should have a website and a company like Hostinger makes it super easy. And right now they're even running a Black Friday deal to make it even cheaper. So let me quickly show you how to set up Hostinger and why I think it's cool. If you head to the link in the description, you'll be presented with this special page. Go ahead and click Claim Deal. I highly recommend the Business Website Builder plan, because if we take a look here, this is the plan that has all of the AI tools we're about to get into here. And go ahead and choose that plan. Now, when signing up, click this little button that says have a coupon code and enter the coupon code Matt Wolf. There's an E at the end and click Apply. You'll notice the price goes down even more. Now, once you're inside of Hostinger, watch how easy this is. I'm going to click on websites here. I'm going to go ahead and click on Add Website to create a new one. And then we're going to use Hostinger's AI powered drag and drop website builder. Let's say I'm going to create a personal website about my various travels. I'll call it Matt Wolf's Travel Site. And then under Description, I can hit this microphone here and just speak out my description. I do a lot of traveling and I want to create a website where I can share photos and videos and even maps of the various places I've traveled recently. It says I have a great description, so let's go ahead and click Create Website. And then we watch as AI does pretty much all the magic to go build this website for us. And after just a few seconds, I have the bones of a website ready to go with a header, some example blog posts, my travel gallery with images, the various travel locations and the map that I just mentioned. Everything that I wanted from my website was created from just entering that simple prompt. When I'm ready to make changes, I can simply click on Edit Site at the Top here and now I have a drag and drop builder where I could rearrange this however I want and really dial in my site and check all of this out. An AI Image generator, AI Writer, page generator, section generator, Blog generator, AI Heatmap, AI SEO Assistant, and AI Logo Maker. Let's check out the AI heat map. It actually generates a sort of color coding showing us exactly which areas of the website will most likely get the most attention. And based on this information, this heat map information, we can design our website to make sure the most important elements on the website are in the areas where we know our eyes are going to be drawn to. Again, Hostinger makes this whole thing super simple. I showed you the whole process of getting a website online and it can literally happen in a matter of seconds. There's a special link in the description that will send you to the Black Friday deal. Use that combined with the Matt Wolf coupon code to take even more off and you are well on your way to having an online presence where you can send anybody to check out your stuff. Thank you once again to Hostinger for sponsoring this video. All right, so let's move over to Google now. If you've done any googling within the last several months here in the us you've probably seen the AI overviews in search. For example, if I ask Google what's the best taco shop in San Diego, you get this AI overview here. Responding to the question City Tacos is the correct answer. Well, that feature is now a global feature. It has now been expanded to over 100 countries. The chances are if you're watching this video, you're in one of those countries. And since we've been Talking about search, OpenAI has has search now the world has Google's AI overviews in search now. Well, now the Gemini API allows you to actually search. Now this isn't super helpful unless you're a developer, but if you are coding something using the Gemini API, you can actually have it search the web to help with the response in your app. So this basically opens up the ability to use Google search results for app builders combined with Gemini. So the same types of responses that we get with the AI overviews directly on Google, you can now actually get if you're developing apps with Google's Gemini API. And in the final note about Google, I found this to be really interesting. The CEO of Google, Sundar Pichai, made a comment about how more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers. They're not letting AI write all the code for them, but it's writing 25% and then passing through humans. So humans can just kind of do the final double checking work, I'm assuming. Massively speeding up the time to market with whatever they're coding. Meta had a handful of updates this week as well that are worth mentioning, including the fact that they released an open version of Google's podcast generator. So I've talked quite a bit about NotebookLM in previous videos. It's a tool where you can upload PDFs and documents and URLs and YouTube videos, and then you could ask questions about whatever you put in there. But it also makes these really cool podcasts with two hosts that sort of have conversations back and forth, forth with each other. Well, Meta thought that was cool as well and released their own version of it, but as open source. They have a GitHub page for it. You can find it over at GitHub.com meta-llama and well, this long URL here, if you know what you're doing, there's actually some instructions where you can download this to your local computer and run it locally, even being offline now, I did find a hugging face space where somebody actually replicated this right instead of hugging face and well, it doesn't appear to be working for me, it just keeps on loading. So maybe the hugging face space version of it doesn't quite work yet and you might have to run it locally on your computer, which is a bit of a bummer because I wanted to actually hear a sample of what this audio sounds like. But that might have to wait for a different video. Maybe I'll make a tutorial breaking down how to install it and run it locally and we'll get to sample it at that point. In other Meta news, they struck up a deal with the news company Reuters. This is the first time Meta has made a deal with a news content platform. Which makes sense because in other Meta news, it turns out that Meta is supposedly building its own AI powered search engine. So very similar to what we just looked at with OpenAI search and what we've seen with Perplexity Search, well, Meta intends to do the same thing. It sounds like now this says reportedly, meaning that it hasn't been confirmed. But the fact that they're partnering with some like other news companies and the fact that OpenAI, Google and perplexity have all really dug deep into the search game, it makes sense that Meta is probably going to follow in those same footsteps and create their own search product. XAI this week added the ability to understand images inside of Grok. So right now in order to use Grok, you have to have one of the X Premium accounts. But you can see we've got a little image icon here. I'll just drop a thumbnail in that I was working on for an AI news video and then say, describe everything you see in this image. Interestingly, I got this error that says currently GROK can't generate images or access real time knowledge once it has looked at an image. Please start a new conversation to remove this restriction. But then it goes on to describe what's in the image. The image depicts a person standing in front of a large digital screen displaying AI news in bold white letters, et cetera, et cetera. So it accurately could see this image. Something that we've had in pretty much all the other AI platforms already for a little bit now. Well, we now have it in Grok. Apple had a handful of announcements this week in the AI space. They started to roll out the new operating system with iOS 18.1 on it, and users are even starting to get 18.2 on it. It's these newer versions that they're rolling out that have all the features they kind of promised at their big WWDC keynote. Things like refining writing, summarizing notifications, mail and messages, a more natural and capable Siri, removing distracting objects from images, and a bunch more. So that is actually rolling out. But just as a reminder, I believe you need an iPhone 15 Pro or newer. So there's still very few phones on the market that can actually use this technology. There's a whole bunch of new hardware releases by Apple this week, all pretty much focused on using their new M4 chips, which are pretty much designed to use Apple Intelligence. So we got this new imac this week that has the M4 chip in it, which of course designed for Apple Intelligence. We got a new MacBook Pro this week with the M4 chip in it. And in their headline Ushering in a new Era with Apple Intelligence, I actually really do like the look of the new new black MacBook Pro. I'm a fan of that look. And they rolled out a new Mac Mini, which you'll never guess, but it was built for Apple Intelligence. It is pretty dang small though. I have a Mac Mini as well. And the Mac Mini that I have is probably four times the size of the Mac Mini that they're showing off here. You can see how small it is in comparison to everything else. The one kind of ridiculous thing about this Mac Mini is. Notice how you don't see the power button anywhere on it? Well, they put it on the bottom of the device. You literally have to like lift up the computer to press the power button. Not quite as poorly designed as having to plug in your mouse cable to the bottom of the mouse, but still pretty weird that you actually have to lift your computer up to press the power button. Anyway, most people that use Mac Minis never turn them off. They just let them use sleep mode anyway. But I still think it's a weird spot for it. And in the final bit of Apple news here, the Apple App Store is soon going to show summary of app reviews made by users. So we've seen this feature on Amazon already where towards the top of the reviews on Amazon. It will summarize what people think of it overall using all of the comments. Well, it sounds like we're going to get a similar feature pretty soon inside of the Apple App Store. This was an interesting story because there was this new mystery AI image generator that popped up on this leaderboard website and this new mysterious model beat out all of the other AI image generators. Now, the way these leaderboard sites work is that it will give you two images and the user votes on which image they prefer, which one's more aesthetically pleasing to them. And over enough votes and enough people choosing A or B across a whole bunch of images, it builds out a leaderboard and people are voting blindly. They don't know which image was generated with which tool. They just see two images, pick their favorite and over time it builds up this data of which model people prefer to use the most. And this red panda popped on there and nobody knew what it was or why it popped on there. But it beat Lux One Pro, it beat Ideogram, it beat Midjourney, it beat pretty much all of the existing models that were out there, winning 72% of the time it was shown for a vote. That mystery led to a lot more people talking about it and actually creating a little bit of buzz around this new model. And then later in the week, this company Recraft came out and said, yeah, that was us, that was our model. It was actually called recraft version 3. We could see in their press release here outperforming all competitor models proven by the hugging faces industry, leading text to image benchmark by artificial analysis. Which is what we were looking at at that last little leaderboard image, but just looking at their blog post at some stuff it generated, like this image of somebody in front of a sign that says Text Generation without limits. Recraft V3 is the only model in the world that can generate images with long text as opposed to just one or a couple of words. And that image is proof of that. Here's another image of three people either jumping or jogging. I can't really tell. A portrait of a young woman with long fiery orange hair flowing dramatically, wearing a turquoise turtleneck. A stack of three red cubes, etc pretty prompt adherent with what it was asking for as well. There's also an API for it. So other software developers could use this model in their software and you could use it over at this site. Recraft AI. So inside of Recraft here I opened up a new canvas. Let's go ahead and create an image. We could see our options here with this option for Recraft V3 RAW. So we'll go ahead and leave that set there. Let's go ahead and prompt a wolf howling at the moon and we get two pretty dang good looking images of wolves howling at the moon. So definitely an impressive model. Eleven Labs rolled out a new tool called X2 Voice, which you plug in like an X username into it and it will like create a character based on the types of stuff you tweet. I actually haven't tested this yet. Let's go ahead and try it right now. So if I go to x2voice.com Enter my own X username here, Mr. E. Flow and click Analyze. I can see an animation of me talking up here. Is it actually saying something? AI updates are quicker than cat videos, but just as addictive. Who needs holiday suspense when we have tech draw? It doesn't sound like me at all, but it used my profile pic from X and what it's saying must be based on my tweets because it knows I talk about AI and sometimes AI drama. Pretty interesting. I don't know exactly what it would be used for other than just kind of having fun right now. I could share it and copy the link and let other people listen to the little, you know, few second clip of fake me talking here. But other than that, I don't really know what this is used for. It's just kind of like a fun little, little thing right at the moment. It'll be interesting to see if this gets built out more or not, or if this was just somebody over at 11 Labs just playing around going, hey, look what I made. Let's share it with other people. Duno released a new feature called Personas. It says Personas let you save the essence of a Song, vocals, style, vibe, and reimagine it across your creations so you could take your favorite vocals or style and bring them to a new song. If I come to Suno and I look at my library, let's see, I've got a song here that says subscribe to Matt Wolf. Let's hear what that sounds.

Don't forget to check out Picture Tools.

Okay, so let's say I like that style and I want to create another song that's got that like folky style to it. I come over to these three dots here, hover over, create, and then come to make Persona. I'm going to leave it private for right now, but let's just call it Acoustic Bulk. And then the description, Acoustic Folk. Go ahead and click save. Now if I go to create a new song here, I've got this option under Persona that I can select for Acoustic Folk. So let's have it generate a song for us. I'll just do a song about a wolf howling at the moon. Not feeling very creative at the moment. I guess I'm going to go ahead and leave the style blank here and just make sure we're using our Persona. And now if I create that with our Persona, it should sound like a similar acoustic rock song, but about a wolf howling at the moon. And here's what we get.

I was just like, you're one of the ways the known and never guess it to me for will we sometime quite some time.

The lyrics make literally no sense. I don't know what happened there, but the style is pretty much the same. So something new from Suno to go and try and play around with did this week launched a new high quality avatar feature which is capable of real.

Time conversation by engaging with your audience person to person. Like you are doing now. Yes, I'm trying to engage with them. Isn't that what you created me for? Exactly. But let me do my thing. Talk to these guys instead of me. Sure thing, Gil.

So this is something I haven't played with yet and I don't believe it's available. It says that they announced it this week, but basically it lets you create an avatar that looks and sounds like you, but you can have real time conversations with. So theoretically I can go and make like a chatbot over on the Future Tools website that looks sounds like me and you can go and chat with a virtual me and it almost feels like it's real time. That's what they're going for at least. Should be fun to play around with once we have access to it. Wonder Dynamics showed off.