AI News: We've Entered A New Era of AI Creativity

In this video, I'm going to break down the AI news that I think is interesting that you might have missed this week. So let's get into it.

There's actually been a ton of news in the world of AI image generation this week. There's been news around Flux and news around Midjourney and news around Ideogram and Freepix Mystic. And so much has been happening in the world of AI art generation that I actually made a whole dedicated video just to talk about what's been going on in the world of AI art. You can find it with this video here called Midjourney Is Free Again. And here's why. That video is not just about Mid Journey. It kind of breaks down the entire AI art landscape and what's going on there right now. However, I will give you a quick recap in this video, starting with the fact that Mid Journey opened up their free trial again. Back when Mid Journey first came out and it was only available inside of Discord, they had a free trial. However, unfortunately, a handful of people figured out how to game the system and a lot of people were figuring out how to get multiple free trials and just keep on using Mid Journey for free. So they eliminated the free trial and this week they brought that free trial back. And if you want to get good at using Midjourney now that you can play around with it for free if you want, I actually have a full playlist on Midjourney here with a ton of tips and tutorials on how to get the absolute best out of Mid Journey. So check that out if you're finally ready to try Midjourney, because now it's free. Now, there is a caveat to it. You only get 25 images. That's not 25 images a month, that's 25 generations. Once you've used Midjourney 25 times, done 25 images, they'll kick you out and make you pay for it. But Midjourney is still one of the best AI image generation platforms. I'm always still impressed with the types of images that I get out of Midjourney. So if you're somebody that's never played with it before, I highly recommend diving in now. Now, I doubt the timing of Midjourney offering their free trial is coincidence, because this week Ideogram rolled out a new model. Ideogram is a really good AI art generation model. In fact, it's most well known for the fact that it was pretty much the first AI generation model that was really, really good at generating text inside of the model. Well, this week they announced ideogram 2.0 and they also announced that it is now available to all users for free. So a model on par with Midjourney comes out, offers users the ability to use it for free, and on the exact same day, Midjourney goes and says, yeah, but we've got free trials now. So is that a coincidence? Maybe? Now, Ideogram is really pretty dang good, especially when you're trying to get text inside of the images. It's pretty decent at realism as well. It's hard to find images that it won't actually do. Now, they do say it is free for all. However, if I look at the subscription page here, this is the reality. You get 10 credits per day. Now, one credit generates four images. So for the most part you get 40 images per day. Midjourney, you get 25 images and then your trial's up. Ideagram you can generate with 10 prompts or 40 images per day. Once your credits are up, you either have to upgrade or just wait till tomorrow. Now, with all of this AI image generation craziness happening, we've seen a lot of sort of unhinged AI images, especially coming out of flux on Grok 2. And, well, AI is getting political. Donald Trump himself posted over on the Truth social platform these images that sort of imply that Taylor Swift and Taylor Swift's fans are endorsing Donald Trump. But of course these are AI generated images. Here's another image that Donald Trump himself posted over on X of Kamala Harris in front of a crowd with a communist flag out in front. So, yeah, the political landscape is getting really weird with all these AI image generations. Something that I think pretty much anybody that's been paying attention to AI image generation probably saw coming. Now, on the flip side of the AI image generation craze, the CEO of Procreate came out this week to say he's anti generative AI and that generative AI will never be a part of Procreate. Here's his little 40 second message.

You've been asking us about AI. You know, I usually don't like getting in front of the camera. I prefer that our products speak for themselves. I really hate generative AI. I don't like what's happening in the industry and I don't like what it's doing to artists. We're not going to be introducing any generative AI into our products. Our products are always designed and developed with the idea that a human will be creating something. You know, we don't exactly know where this story is going to go or how it ends, but we believe that we're on the right path.

Supporting human MKBHD actually retweeted this message from the CEO of Procreate saying, bookmark this. Such a fascinating announcement. Artists and users on social media celebrate. Take notes. Adobe. But then he goes on to say, but technically this is committing to never offering any of the features in their products, no matter how good or useful they may get in the future. An announcement to not add features. I wonder if they ever bend on this rule someday. And that's kind of the take that I agree with. I mean, I sort of commend James here from Procreate for taking a stand. And while every app in the world is basically saying, look, we've got AI in it, now, they're saying we're never going to do that. But also never's a long time. But at the end of the day, procreate is just one tool. And if you want to generate AI art, just don't use Procreate. Simple as that. Moving along to what's been happening in the world of AI video generation, Luma Labs released Dream Machine 1.5. Now it's got higher quality text to video, smarter understanding of your prompts, custom text rendering, and improved image to video. I jumped into my lumalabs account to test a few generations here. This was actually one of the recommended ideas here that I just had it generate an extreme close up footage of a young sailor woman with a concerned expression during a rainstorm and this is what it gave me for that looks pretty good. I mean, it's very cartoony, sort of Pixar ish looking, but I think it's a pretty decent video. But they mentioned that it's better at rendering text now and so I was curious about that. So I used one of the prompts that I was playing with when I was testing all the AI image generators and I asked it to do a mystical forest clearing where wisps of fog spell out the words Magic awaits. And in flowing ethereal lettering between ancient trees. My first attempt it says Magic Wanettes. So I went, huh, that's weird. Maybe I'll try again. The second time I tried it, it went and wrote Magic wa vats. It got magic perfect both times, but it struggled with the word awaits. Either way, I actually think that Luma Labs makes some pretty decent videos. I'm happy with most of the stuff that has generated for me so far, but I kind of have to disagree that it's great at text yet Maybe it's better if it's just like one word or something, but just trying to get it to say magic awaits. On two rerolls, I could not get it to work, but again, still makes some pretty decent videos. We also got another AI video tool rollout this week in a tool called Hotshot. And the creators of this tool basically compare it to Sora because their press release is titled How a four Person Team Built Sora. And admittedly most of the generations that they show on their homepage are are really freaking good. Like this girl popping out of the water, these corgis here, this girl sending a text message, this girl among like these wheat fields. These videos look really, really good. Unfortunately, they're cherry picked as hell. Now this tool lets you generate two videos for free every day. So if you generate two videos and you don't get one that you like, you have two options. You can either wait till tomorrow and try two videos again, or you can pay them a hundred bucks a month to be able to generate more videos. Me being a huge sucker, I upgraded. I paid the hundred bucks because I wanted to test more generations to see how well it does when I try my own prompts. So let's click over to mine and see what I got. And well, you'll see pretty quickly that nothing that I did looks nearly as good as the prompts they show on their homepage. This is supposed to be a monkey on roller skates down here. This is like a chameleon changing colors with the background. This was a sandcastle that was supposed to collapse when the waves hit it, but it doesn't collapse. These are sparks that are supposed to turn into constellations. This is a paper airplane flying through different cities. Yeah, nothing that I generated came out anywhere near as good as what they were showing on the homepage. So quite honestly, I think there's quite a bit of work to do. I've been much more impressed still with what we're getting out of Luma's Dream machine and Runway gen 3 than what I'm getting out of Hot Shot. And quite honestly, this is the most expensive of the options too. So hopefully it gets better soon. If you want to see a full detailed breakdown of all of my tests using Hotshot, I actually did a video called New AI Video Tool. Claims to be as good as Sora. And this one I do a much deeper dive into how I came up with the prompts, why I use these different prompts and the outputs that I got. So check that out if you want to dive A little bit deeper on what Hotshot is capable of and my experiments with it. In more AI video news, LTX Studio rolled out publicly this week, so now anybody can use it and it's got some pretty cool features in it. The coolest feature in my opinion is the ability to do this motion tracking so you can upload a video of yourself giving certain facial expressions, and those facial expressions will translate to the characters inside your video. To me, that's pretty dang cool. You can also do things like scribble a scene to inform the platform what you want your videos to look like. Give it a text prompt to go along with it and it will generate that scene based on your scribbles. And there's much more features like keyframe controls, enhanced collaboration, stuff like that. So it looks pretty cool. Also this week, DID launched an AI video translation tool that includes voice, cloning and lip sync. So if you're familiar with tools like hey Gen, this can now do what that tool does. Here's a quick example.

With DID Video Translate, a feature of our studio that automatically transforms your videos into seamless spoken translation. And since we're talking about AI audio right now, it's worth mentioning that 11 labs text to speech app called Reader is now available globally. We've had it in the US here for probably a month or so, but it's a cool tool where you can open up articles or PDFs or things like that and have them read to you in pretty much any voice you want, including your own voice. All right, let's move on to large language model news. And let's start with Perplexity this week because Perplexity is like my main go to when I need any sort of information from the web now. I go to Perplexity over Google search 99% of the time now. And Perplexity is starting to roll out some new features like ChatGPT's code interpreter. We can see Phil here says, wow, Perplexity's code interpreter can now install libraries and display charts in the result. This enables many more use cases compared to Chat GPT code interpreter, like creating stock market charts using Y Finance and then there's some screenshots here. Install a library to get the stock market returns of the day index over the past 20 years, then use it to make a chart and Perplexity actually goes and runs the code here and then generates the chart for them. So this is a really, really solid add on to Perplexity, which is probably going to make it more and more people's go to Tool to do research now. Everybody doesn't have this feature yet. Aravind, the CEO of Perplexity, says code interpreter and plot rendering are slowly being rolled out now in less exciting Perplexity news, but probably good for Perplexity, not as good for users. Perplexity AI is planning to start running ads in fourth quarter as AI Assisted search gains popularity. So it looks like Perplexity is going to kind of go in that Google model and start running advertising right inside of Perplexity when you ask questions. Apparently they're circulating a pitch deck right now that is promoting the app's reach and increasing usage. And down here it says with advertising, Perplexity will follow a model called CPM or Cost per thousand impressions. Perplexity said in its pitch deck that its key advertising categories initially would include topics such as technology, health and pharmaceuticals, arts and entertainment finance, plus food and beverage. Advertisers will be able to sponsor related questions below answers and and buy display ads to the right of a Perplexity generated answer. Now, I'm not certain from this article here on CNBC whether or not pro users are still going to have ads, because if you've used Perplexity, you know there's a free version and a pro version. The pro version you actually have to pay to get access to. I'm assuming the ads are only going to be in the free version and not the pro version. But we still have to wait and see because that's not specifically laid out in this article here. Now, if you're somebody that uses AI at work, or you're thinking about using AI at work, you're going to want to grab HubSpot's completely free bundle called 5 Essential Resources for Using ChatGPT at Work. And honestly, if you're not thinking about using AI in the workplace, just remember what Fei Fei Lee says here. AI won't replace humans, but humans using AI will. So if you're not using AI to speed up and improve the quality of your work, well, your competitors probably are. The link to this completely free resource is down in the description right now. And and trust me, this is something you're going to want to look through. It includes interesting flowcharts on when you should or shouldn't use ChatGPT. There's also a really cool template in there that you can use with ChatGPT to make sure that any content it creates for you follows your brand's voice. You've got an AI generated content refinement checklist to double check AI's work and ensure that you're putting something out into the world that you really want to be putting out there. There's a four page checklist that you can easily follow all about adopting AI in the workplace and a super comprehensive PDF guide on how to supercharge your day with ChatGPT. And what's really cool about this is if you scroll all the way down to the bottom of this document, they have a hundred ways to try ChatGPT today. It's got some pretty cool prompts to test out in it too, like providing recommendations for improving customer service and support, providing recommendations for improving website SEO, helping with email management and organization, and so much more. Again, it's super comprehensive, so super helpful. And once again, the link to this 100% free resource from HubSpot is down in the description below. And thank you so much to HubSpot for sponsoring this video Anthropic's in the news this week, but not for the best reasons. They're getting sued just like OpenAI. Just like pretty much every company making large language models. In fact, I'm probably going to stop talking about lawsuits in these videos because pretty much all of these large language models are getting sued these days, and I'll probably only start talking about lawsuits that are like concluded. And there's actually something that comes from it. But once again, a bunch of authors are suing because their works were scraped to get the data. For Anthropic's platform, there's a giant data set called the pile. And the PILE is this large scraping of information from all over the web that a lot of these large language model companies use as their sort of initial data set to kick off the training of their large language models. And it sounds like authors are kind of starting to go after any company that trained data on the pile because the PILE also included books from authors and paywalled content and things like that. Microsoft released a new model this week called Phi 3.5. This new model has 3.82 billion parameters. It is a smaller model that seems to be designed to run on mobile devices and things like that. And when compared to other similar sized smaller models like GPT4O, Mini, Gemini 1.5, Flash, Gemma 2.9B, Llama 3, 18B, Mistral, Nemo, et cetera, et cetera, et cetera, it pretty much performed better or on par in all of the various benchmarks that it was tested on. Moving over to OpenAI this week, they rolled out the ability to fine tune GPT4O so you can sort of custom train your own information into the model. Now, quite honestly, for most people and most companies, fine tuning is probably overkill. It's probably not necessary for you to do. You can use something like Retrieval, augmented generation or rag and just share documents and let your LLM have access to those documents to retrieve whatever sort of custom information you want. But if you do want to fine tune a model in GPT4O, you can now do that. They're offering 1 million training tokens per day for free for every organization through September 23rd. It's pretty cool that OpenAI is doing that and allowing companies to do that, but again, it's probably overkill for like 90% of companies. Also this week, OpenAI partnered with Conde Nast. I think I said that right. They own brands like Vogue, the New Yorker, Conde Nast Traveler, gq, Architectural Digest, Vanity Fair, Wired, Bon Appetit and more. And now content from those brands are going to be accessible inside of OpenAI and chat GPT and of course the new search GPT, which I personally still don't have access to because they seem to have cut it off after a certain amount of users and said no more until later. And if you're keeping score, OpenAI has now partnered with Conde Nast, Associated Press, Axel Springer, the Atlantic, Meredith, Financial Times, Le Monde, News Corp, Prism Media Time, Vox Media and others. So it seems like maybe they're trying to avoid all these lawsuits going the distance by just going and partnering with these companies and either paying them or giving them access to their tech in exchange for being allowed to use their data. That's what it seems like is happening here. And since we're talking about OpenAI, OpenAI finally spoke up about the Senate Bill 1047. I did do a podcast episode on the Next Wave with Anjanay from A16Z where we talked about this Senate bill and why AI companies are so opposed to it. Essentially, the bill wants to hold the companies that make the large language models responsible for any harm done using the large language models. So if somebody goes and uses an LLM like Meta's Llama LLM and uses it to create a bomb or something like that, not only will the person who used the LLM to create the bomb be held responsible for their actions, but Meta, who created the LLM, will also be held responsible. And that's a very oversimplified explanation of why all of these AI companies are so opposed to this new bill. They're worried that it will stifle innovation because how are they supposed to know what these models might eventually be used for? And how are they supposed to continue to make the models more and more advanced if they're worried that these more advanced models might be used by bad actors to do bad actor stuff? And the information who also reported on this same exact thing here did a quick breakdown that I thought was interesting of the friends and foes of California's AI safety bill. So TechNet is against the bill. Mozilla is against the bill. A16Z is against the bill. Google's against the bill. Hugging Face is against the bill. Meta is against the bill. Y Combinator is against the bill. They have OpenAI as sort of neutral, but now OpenAI is coming out as against the bill. And then Anthropic and Microsoft they're putting as neutral. But Anthropic basically said that they would support the bill if there was some amendments made. Some amendments were made, but then Anthropic hasn't really made another statement about it since. And then Microsoft has not taken a position at all about the bill. And then the proponents of the bill are the center for Human Compatible AI and Notion for some reason, which seems kind of odd because notion uses anthropic, OpenAI and other LLMs. A super quick edit here in the middle of editing this video. And it just came out today that Anthropic actually did take a stance on this bill. They said that Anthropic now says the benefits outweigh the cost for the California AI bill. So basically Anthropic is now saying they think that this bill isn't so bad. They didn't say they endorsed the bill, but they said that they think the benefits of this bill are going to outweigh the costs. Just wanted to make this little add on real quick and splice it in because new information came out about that specific topic. If you remember from back several months ago, Derek, during Microsoft's Build event, they announced their Recall AI feature, which was a feature that was like your search history in your browser but for your whole computer. It just kind of remembers everything you did on your computer. So you can go back and remember anything whether it was on the Internet in your browser or not. Well, a lot of people were really worried about the security implications of this. Microsoft sort of shelved it for a little bit and now it looks like we're going to be getting this in October. Finally, I actually had the opportunity to play around with this feature while I was at the Microsoft Build event. I thought it was pretty cool, but I also saw why it could be worrisome. I do Understand that it didn't blur out passwords or anything like that. So if you type a password on a screen that was saved in your history, and if somebody got access to your computer, they can go back, look through your history and see you typing in passwords and things like that. So there was definitely some issues. So it's a good thing they went back to the drawing board with it. But personally, I'm kind of excited for this feature. I thought it was really cool. Google made an update this week to Gmail with AI adding a polish option where you can type an email, hit a button to polish it, and it will sort of rewrite that email for you in a new polished fashion. They also added the ability to refine my draft. So you can see they wrote up a draft here and they're given the option to refine the draft and then they have options for formalize, polish, all that sort of thing. And then moving on to robotics, we've got some fun cool robotics news this week. The Unitree Robotics that we can see here on the screen the rolling out their G1 mass production version. We can see in the video what it can do. It can like walk up and down stairs, it can jump, it can do all sorts of cool stuff. And once it's available, it's going to go for $16,000. Now I know that's a lot of money, but quite honestly, that's a lot less expensive than I would have thought something like that would be at this current stage of where we're at now. According to this article on the Verge, Unitree hasn't revealed exactly when the production ready version of the G1 will actually be available. But when it is, potential customers shouldn't expect it to stroll into the kitchen and cook breakfast right out of the box. The robot is designed to learn through imitation using the company's Unitree robot unified large language model. And its initial appeal will be as an affordable humanoid platform for conducting robotics research, not as a real life version of Rosie the Robot from the Jetsons. And we got more robots this week too. This one is called AGI Bot and once again, it is another humanoid robot. To me, it kind of looks more like goat legs or something, but it looks like he's trying to challenge Elon Musk and their Optimus robot. Now, we don't have any videos of this robot yet, but we do have just the images that we've seen on screen. I'll probably talk more about this as more comes out from this company. This week, Best Buy introduced an AI Powered delivery tracking. Basically, people are impatient and when they order something from Best Buy, they want to know exactly where it is at the exact moment. And it sounds like Best Buy is using AI to accomplish that. Best Buy is rolling out an AI powered delivery tracking system that promises minute by minute updates. The company aims to satisfy customers who are increasingly frustrated with vague delivery windows and limited order visibility. And here's something I think everybody can get behind AI technology can help you win the battle over mosquitoes. There's this little device called the Bazigo Iris. I don't know if I'm pronouncing that right or not, but it's a smart mosquito detector. It weighs less than a pound and it uses AI vision and algorithms to detect, track and target mosquitoes. And it even works in complete darkness. And when it spots a mosquito, the Iris doesn't just sit there, it tracks the insect until it lands and then uses a safe laser pointer to mark the spot. It then sends an alert to your smartphone, which is compatible with both iPhone and Androids. So you know exactly where to aim your swatter. So it doesn't actually like blast it out for you. That's probably like V2, V1 just points it out with a laser and then you can run up and slap it. But I want one. I'm one of those people that is like a mosquito magnet. So whenever there's mosquitoes around, nobody else gets bit because they all come to me. So I kind of need this in my life. And then finally, a16z released their top 100 gen AI consumer apps this week, which breaks down the top 50 gen AI web products and the top 50 gen AI mobile apps. Taking a peek at the web products, no surprises here. ChatGPT is number one character AI is number two and perplexity is number three, followed by Claude and Suno. I'm actually kind of shocked that Sudo is this high up, but good for them. It's a cool, fun product that I know more and more people are catching on to. Funny thing is Janitor AI number six. I've never even heard of that one. I'm going to need to look into that and figure out what it is. I'm excited to see Leonardo here at number 17 and a lot of familiar apps on this top 50 list. And then under top 50 gen AI mobile apps, once again, no surprise, Chat GPT is number one. Microsoft Edge is number two. That one's a little surprising to me. I think most surprising to me is to see Perplexity all the way down here. At number 50. For me, that's probably my top most used AI app on my phone. And the fact that it's all the way down Here at number 50, this one deserves to be a lot higher. I think this one should be up in the top 10 because it's a really useful app. Anyway, that's just my personal opinion on it. I love seeing reports like this and seeing where tools are ranking. So that's what I got for you this week. Lots of lots of AI art news in the world of image generation and video generation. More lawsuits, more regulation that the AI companies are against, and some cool robots that we can look forward to in the future. The next several weeks are going to be fairly crazy for me. I'm going to be at VidSummit the first week in September, HubSpot inbound the third week of September, Meta Connect the fourth week in September so the videos might slow down a little bit on this channel. I'm going to still try to get the AI News videos out once a week to loop you in on the latest AI news like this one. But unless some really cool, insane things happen that I just feel I have to share with you, there may be a few less videos over the next, you know, four to five weeks. But again, when something major happens, I'm definitely going to take the time, step away from whatever I'm doing to make a video for you and keep you in the loop. So if you want to stay looped in on all the latest AI news, latest AI tools, fun AI tutorials that you can do yourself, make sure you're subscribed to this channel. Like this video. It will ensure that more of these videos show up in your YouTube feed and also makes me feel good and helps my channel out. So I really appreciate it. And if you haven't Already, check out FutureTools IE where I curate all of the coolest AI tools. I have a newsletter. I keep the AI news page up to date on a daily basis. Lots of good stuff there. If you're into AI like I am, it's all free at FutureTools IO. And once again, thank you so much for hanging out with me. Nerding out with me. I record these videos on Thursday, so if any big news came out late Thursday evening or on Friday, it'll probably be in next week's video. But I love that you just come and hang out with me once a week and we can talk about fun AI news. I really, really appreciate you. Thanks again to HubSpot for sponsoring this video. I'll hopefully see you in the next one. Bye. Bye.