You Probably Missed This AI News This Week!

So this week the biggest news in the tech world was that Twitter has officially rebranded to X and pretty much everybody on the planet can't understand why the X team seemingly threw away 17 years of brand equity in the Twitter name. I sort of understand why they did this, but it's a whole rabbit hole for a different day. We also learned this week that aliens probably exist. As I've stated publicly Already in my NewsNation interview, biologics came with some of these recoveries. Yeah. Were they, I guess human, human or non human biologics? Non human. And that was the assessment of people with direct knowledge on the program I talked to that are currently still on the program. And people have been freaking out by whatever the heck this trend is.

It's my birthday.

But that's not why you tune into this channel. You come here to learn about all the AI news. And this week there was some serious AI news. So let's go ahead and break it down right now. At the end of last week, a statement was released on the White House website State stating that the Biden Harris administration secures voluntary commitments from leading artificial intelligence companies to manage the risks posed by AI. Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI all voluntarily committed to practice safe AI. The leaders of these companies all promised to ensure products are safe before introducing them to the public build systems that put security first and earn the public's trust and shortly after, this White House announcement came out with One by one, each one of these companies dropped an announcement on their own websites. About the Frontier Model Forum this is an industry body focused on ensuring safe and responsible development of frontier AI models. The core objectives of this forum Advance AI safety research to promote responsible development of frontier models minimize risks and enable independent standardized evaluations of capabilities and safety Identify best practices for the responsible development and deployment of frontier models Helping the public understand the nature, capabilities, limitations and impact of the technology Collaborating with policymakers, academics, civil society and companies to share knowledge about trust and safety risks and supporting efforts to develop applications that can help meet society's greatest challenges. This is OpenAI's statement from their blog. But if you read the statement on Microsoft's blog, it pretty much says the same thing. And again, if you read the statement over on Google's blog, yep, pretty much the same message. Anthropic also released a similar message over on their blog, but they refuse to mention the other companies by name. If I do a quick search, you can see Google not mentioned, Microsoft Not Mentioned and open a Not mentioned. From my perspective Basically what this is is all of these companies are coming together to make sure that they align on safety, but it's also kind of them getting in the good graces of the government so the government doesn't try to overregulate them. They'll say, no, no, no, we'll all come together and work together and regulate ourselves. So while yes, AI regulations are incoming, these regulations are most likely going to be designed by the people that are building the AI models. Also, the same day that the White House made their announcement, OpenAI's head of trust and safety, Dave Wilner, stepped down from the company. I feel like a lot of the media tried to make this out like Dave Wilner was stepping down because he had conflicts with the way that OpenAI handled trust and safety. But in reality he was just stepping down because OpenAI was growing so fast and he was spending so much time working that he wasn't spending enough time with his family. Instead of actually being an employee of OpenAI, he's moving to an advisory role of OpenAI. So he's not totally leaving OpenAI, he's just stepping down from that specific role. But of course most media outlets are going to sensationalize everything and make you think this is a really big deal. Also, at the end of last week, Stability AI released Free Willy, their large and mighty instruction fine tune models. It gets the Free Willy name because it was Inspired by the Orca paper about progressive learning from complex explanation traces of GPT4. This model leverages the Llama 2 70B foundation model and at the moment is only available under a non commercial license. In these various benchmarks like Ark, Helliswag, MMLU and truthful QA, Free Willy 2 actually performed almost as good as Chat GPT. And if you have no idea what ARC, Hellaswag, MMLU or Truthful QA are, don't worry, neither did I. I literally had to look it before making this video. These are essentially like quizzes that they use on these chat bots to test how accurate the bots are. So when giving these various tests that sort of have different focuses in different areas, they'll score differently. So in this ARC test, ChatGPT outscored Free Willy. Where in this hella swag test, Free Willy actually outscored chat GPT3. If you'd like to test Free Willy 2 for yourself, there are a few hugging faces spaces available where you can give it a shot. Just go to spaces search Free Willy 2 and there's a very basic user interface. And if you really know what you're doing. The files are also available over on Hugging Face for people to run and experiment with themselves. Personally, this is above my level of expertise and I would have no clue how to run these myself, but maybe something I'll dig into deeper in a future video. This week, the godfather of AI, Geoffrey Hinton, made a claim that he thinks that AI has or will eventually have emotions. Hinton is one of the original AI researchers and he worked on AI with Google for a long time before recently stepping down from his position at Google because of fears that AI will be used for mass fake news and autonomous weapons. At a talk that he gave at King's College, when asked if AI systems have emotions, he said, I think they could well have feelings. They won't have pain the way you do unless we wanted, but things like frustration and anger, I don't see why they shouldn't have those. Now one thing I'm fairly certain doesn't have emotions are all of these cool AI video tools that we've been getting access to lately. Recently, Runway ML announced that with Gen 2 we can now use image prompts to generate videos. So if I log into my Runway ML account, you can see that there is now an image to video option here. If I click on this, I could take this image that I recently generated in Leonardo and pull it into Runway and if I leave this prompt blank, it will try to animate this image and here's what it generated for us. You can see the woman's action actually looking around and the light changes on her. Now it's only generating four seconds right now, but if you watched my previous video where I talked about Gen 2 a little bit, what you can actually do is take the very last frame of your video here, take a screenshot of it, and then take that screenshot as your starting image for your next video and start to string together a longer and longer video using the last frame of every single four second video. Realistically, you can make much longer videos and I imagine it's only a matter of time before tools like Gen 2 build that into the workflow where it just takes that last frame and then continues the video for you. Now, whenever I do an example like this, people always say, well, that worked with an AI image. Does it work with real images as well? Here's a silly headshot of myself that I give out when people ask for headshots, when I go on podcasts and stuff. Let's pull this in and generate with that. Now one thing I did notice is that if I also add a text prompt if you it changes the image completely. When you add a text prompt, it uses your image almost as like inspiration for the video, but it doesn't use it as sort of the starting frame for the video like you'd want it to. So if you actually want it to animate the specific image that you upload, you need to leave the prompt blank. And here's what it generated using my face. Oh wow. It really just changed the way I look. That's kind of creepy, but that's what it does. When I uploaded my own headshot this week, Kyber, one of my other favorite AI video tools, I announced text to animation. Now, they specifically asked me not to show off the user interface yet because the interface is going to change. But I'm going to enter the prompt Samurai holding a sword, ready to fight with the sunset behind him. And check this out. Here's the video that it generated. It only generated a one second clip, but I mean it's pretty impressive. And let's just keep on rolling with cool AI video generators. This is one I haven't talked about in a little while on this channel, but PlasmaPunk.com just added the new SDXL and Kandinsky 2.2 models inside of their platform. And I've always really loved the videos that Plasmid Punk generates. So let's walk through the process real quick here. Just for fun, let's pick a song. Let's go with rebellious.

Let's go with this Klein Sealer here. I'm sure I butchered that name. I'll just go ahead and use the little recommended section here for the music. Now it looks like to use one of these newer models here. It is a premium paid feature, but somehow I have 60 credits. So maybe you get 60 credits when you first sign up and then you need to pay beyond that. But I'm being totally honest, I don't exactly know how that works, but let's go ahead and try stable diffusion XL and 16:9. And let's click next. I'll go ahead and use this Sci Fi Digital one that it's recommending. I'll click next and then it asks for a prompt for the video style Future Punk, cyberpunk city, highly detailed painting, and then some of these artists. This was actually the default prompt that it suggested. So let's just see how it goes for the video story. This video is about a man standing in in front of a pub in the rain holding an umbrella. Or you can also give it some song lyrics let's go ahead and use this one that it sort of gave me by default. Looks like it's going to cost me 19 credits out of my 60. Let's go ahead and click Generate and see what we get. And here's what I got out of it.

Now I'm not going to play the full 20 seconds, but I'll toss it up on my Instagram if you want to take a look. Let's just keep on moving now. I know a lot of people that watch these videos are content creators like me, or they watch my videos to stay up to date with how to leverage AI for their own businesses or careers. Like me, you love to find new and unique ways to use AI and the latest tech to create things that wow people and to do it quicker than previously possible. So for this video, I partnered with HubSpot to get their latest free ebook into your hands. It's called Using Generative AI to scale your content Operations. With this ebook you'll learn about the tools and the specific prompts that you can use to generate ideas, research content, breakthrough writer's block, create powerful visuals and even help to market your content to get more eyeballs on it. I especially found this section about staying in Google's good graces even while using AI generated content particularly helpful because, well, I rank for a lot of great keywords for my Future Tools website, and a lot of content on my Future Tools website is actually generated with AI. So I think you'll find it super helpful as well. I put the link to download the free ebook in the description below. And thanks again to HubSpot for sponsoring this video and providing this free resource to the viewers. Now let's jump back into the news. This week, Microsoft's Bing Chat comes to Chrome and Safari in tests for select users. Basically, making Bing Chat only available inside of Edge was one of Microsoft's ways of trying to get more people to go use Edge. Well, it seems like now they just want more people to use Bing Chat whether they're using Edge or not, so they're making it available inside of other browsers. I checked to see if I got it and it says only available in Microsoft Edge, so doesn't appear they've given me access yet. Don't they know who I am? I don't know why I just did that. Here's some interesting news. If you remember a few months back OpenAI released their own AI detector where you can copy and paste text from any article or website and it would tell you whether it thought it was generated by AI or not. Well, it turns out it really wasn't too accurate and OpenAI shut it down without really telling anybody. Back in January they released this blog post new AI classifier for indicating AI written text. They launched the tool on January 31, 2023 and the only announcement about it being shut down is back on this original blog post. As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy. In other ChatGPT news, ChatGPT is now available for Android users. The mobile app was previously only available on I. This week they finally rolled it out to Android users. When the announcement came out, it was only available in US, India, Bangladesh and Brazil. However, today, inside of OpenAI's Discord, they announced that the ChatGPT app for Android is now also available in Argentina, Canada, France, Germany, Indonesia, Ireland, Japan, Mexico, Nigeria, the Philippines, the UK and South Korea. So now quite a few more countries have access to the app on Android. This week, Sam Altman announced worldcoin worldcoin is designed to help differentiate between humans and not humans. I suppose worldcoin consists of a privacy preserving digital identity, a World id, and where laws allow a digital currency or the WLD coin. You receive these coins simply for being human. To prove that you're a human, you visit an orb or a biometric verification device, and once you've verified your humanity, you receive a World id. This lets you prove that you are a real and unique person online while remaining completely private. Curious what an orb looks like? Well, here's one right there and I promise that's not me in the reflection there, even though it kind of looks like me. Now, if you're willing to verify that you're a human and not an AI, you'll receive a world ID and 25 of the crypto tokens that come along with it. Insider describes it this way. The orb's camera lens scans your face and iris for about 10 seconds and then beeps to confirm that you're a person rather than a robot. It'll then beep again once it has added you to the World Coins database. They also describe describe it this way. It was reminiscent of the photo ID checkers you'd find at an airport, only instead of staring at a screen, you're made to look into a vaguely terrifying miniature Death Star. However, they do say during the process they weren't required to hand over names, email addresses or phone numbers. It's just proof of humanity. More than 2 million people have already signed up for the project, but I'M still somewhat unclear of what the tokens will be used for. However, this article claims that Universal Basic Income is something that the OpenAI founder has hinted could be one of World World Coins future applications, so we'll have to wait and see. I'm curious, would you let one of these orbs scan your eyes? Also this week the CEO of Spotify, Daniel Ek, talked about potential AI capabilities on Spotify. Now Spotify already has an AI DJ that can give you some information about his song, but it sounds like they're planning on using these AI generated voices in part of the advertising platform as well. And Daniel Ek and Spotify have even filed for a patent for an AI powered text to speech synthesis system, which is interesting because there are already so many tools that use this sort of technology, but I know little to nothing about patent law. If you're looking to get a job in AI, maybe think about going to Netflix. Netflix is offering up to $900,000 for a single AI product manager role. The timing of this feels very interesting seeing as all of the writers are on strike in Hollywood, all of the actors are on strike in Hollywood, and Netflix is out there saying we need somebody to run AI platform. We're willing to offer up to $900,000 to have somebody help us build better AIs for Netflix. So if you've watched that Black Mirror episode of Joan is Awful, they're looking for the person who's going to come in and actually build that technology for real for them, most likely. I'm just guessing. I don't know what I'm talking about. This week Amazon held their AWS Summit in New York and the topic of the event was mostly about AI. However, most of what they talked about is completely over my head and it seemed like most of their announcements were GE towards engineers who knew what they were talking about. One thing I do know about though, is that they are using more Nvidia H100 GPUs inside of their cloud computers, making it easier and faster for models to be trained on Amazon's cloud computers. Also at the event they announced agents for Amazon Bedrock, which gives developers the capability to create fully managed agents in a few clicks. Now when they talk about agents, they're referring to AI applications that can perform multiple tasks and even do API calls to other services to continue to complete tasks. If you remember a few months back we were talking about things like baby AGI and various AI agents where you give it a task that you want it to complete and then it figures out what subtasks it needs to complete to get to that end goal of the final task. And it seems like with this announcement, Amazon's bedrock has some of that tech built into it. Also this week, a whole bunch of news publishers are coming together to try to sue the various AI companies for billions of dollars. Companies like iac, the New York Times, News Corp, and Axel Springer are very close to forming a coalition to fight AI companies that use their data to train the models without consent. They claim that search was designed to find the best of the Internet. These large language models or generative AI are designed to steal the best of the Internet. These news corporations are claiming that the value that they've been collected is running into the billions of dollars across the industry. Now, nothing has totally come from this yet. This is sort of an announcement of a potential announcement. But news companies right now aren't very happy with the various AI platforms. All right, let's talk about some more fun stuff that's come out this week. This week, Stability AI announced SDXL 1.0. We've had SDXL 0.9 for a little bit now. They're finally ready to Release the full 1.0 version, which boasts much more realistic images and also the ability to add text into the image. They also claim that you can use simpler language, so you don't need to necessarily, necessarily be a prompt engineer to get amazing images out of it. You should be able to use simple prompts to get quality outputs like they've done with the previous releases. They did make it available over on Clip Drop for us to play around with. So supposedly they're much more photorealistic and they work much better with using words. So let's test out an image that uses both a photo realistic woman standing on the beach holding a sign that says Future Tools. This one almost got the sign right. I'm not sure how realistic this woman looks. Here's another image that it gave me. This is probably the best one so far, although it does say Future Tools. And I'm not quite sure. This is like a half skirt, half jean shorts that says Future Tools. Here's some of the examples that they give. I want to see if I can generate a prompt similar to this one because this one actually looks pretty good. Photo of a brooding, serious, bearded old. I'm assuming the word man was in there as well somewhere. And these are much better generations, in my opinion. They definitely still have almost like a painting or digital art look to them. I wouldn't say they're super realistic, but it's Definitely a huge improvement over what we were getting out of the previous SDXL models. Now, before anybody jumps into the comments and starts saying I'm prompting it incorrectly, this is just my first impression of it. This is literally the first time I've tried SDXL 1.0. I didn't test it prior to this recording. So I'm sure there's all sorts of other prompting tricks and ways to get much better images out of it. I just wanted to show off the fact that it's available. You can go use it for free right now over at Clipdrop Co. I'll link it up below and you can probably figure out all sorts of cool ways to generate amazing images with it. And finally, the day that I'm recording this video, which is Thursday, the Adobe Photoshop Beta version just rolled out a new feature called Expand. Now, if you have Adobe Creative Cloud or you have just Adobe Photoshop by itself, this only works in the beta version. So if you go to the Creative Cloud app, make sure you're running Photoshop Beta and not the normal version of Photoshop, because the AI stuff is still in beta. But basically what this allows us to do is we can pull an image into Photoshop. I'll pull this jellyfish image in that I made in Leonardo in one of my previous videos. I'm going to zoom out a little bit, and then if I come up here to the Crop tool, I can actually crop to a bigger size here like this. And then there's a new button that says Generative Expand. In fact, let's set it on 16, 9, just to expand it to a fairly standard size. Try to center my image somewhat. And now I can click Generative Expand, leave this blank and click Generate. And it will try to fill in the gaps for me. And look at that. It just built on that existing image that I just had. Now, previously you were able to do that by doing a selection. This just kind of saves us a few steps and lets us expand it. And it does give us three options. There's another option. And there's another option. And I can expand once again. Let's go ahead and pull this out this way, Pull this down further this way and see how it expands it even further. And here's an option. Here's an option. And here's an option. I think that one looks pretty dang cool. So it started with this image here. Expanded to this. Expanded to this. I'd say that's pretty sweet. Also came across this Twitter video here from Howard Pinsky, where he shows that he actually pulled a video into Photoshop and it worked on his video. Now if you are going to do this with video you got to make sure you got it on a tripod so that the framing stays consistent. But they use this little trick in Photoshop beta to actually generate a video with expanded borders. Now anything could be easily Transitioned into a TikTok or reel sized video. Whew. What a week. That's all I got for you. Hopefully you learned something new. You found out about some cool stuff that you didn't know was happening. If you like this kind of video, give it a thumbs up and maybe consider subscribing to this channel. I put out new videos about what's going on in AI every single Friday and and try to put some other cool tutorials and experiments throughout the week as well. And I do my best to help people leverage all of the latest advancements in AI technology and tools. Also, if you haven't Already, check out FutureTools IE. This is where I curate all the latest cool tools that I come across. I keep the AI News section up to date pretty much every single day. It's one of the most up to date resources on AI news on the entire Internet and if you don't wanna keep up with it on a daily basis and you just wanna weekly tldr, join the free newsletter and once a week I'll just the top level what you need to know about the coolest tools and the most important news that happened in AI for the week. I also share a handful of YouTube videos with you and one cool way to make money with AI I send it every single Friday. You can find it over at FutureTools IE and click on Join the Free Newsletter Once again, thank you so much for tuning into this video. Thank you so much to HubSpot for sponsoring this video. These sponsors really, really help me keep this channel going and allow me to make it my full time job to stay in the loop with the latest AI news and AI tools. So HubSpot, you're awesome. Thank you so much for sponsoring this video. I really appreciate it and I really appreciate you for watching this video and wanting to stay up to date with the latest news and AI technology. I just love nerding out about this stuff and I love that there's so many other people out there like you that love nerding out about it with me. So thanks again for tuning in. I really, really appreciate you. I'll see you guys in the next video. Bye. By.

Sam.