AI Not Allowed To Use Nukes! (And Other AI News)

Well, it's been a little while since I've made an AI news breakdown video. In this one, I'm going to walk through all the stuff that's happened in the last few days. Let's start by talking about the biggest news that's happening in the AI world as of the date that I'm making this video. May 1, 2023. Today, Geoffrey Hinton, who many people refer to as the godfather of AI and someone who won the Nobel Prize of computing for actually developing a lot of the technology that we use today in AI. Now, if you're curious about what the Nobel Prize of computing even means, it's. It means he won the a.m. turing Award. Google actually acquired Hinton's research company back in March of 2013, and Hinton has been part of Google ever since for the past decade. Well, today, May 1st, the big news is that Hinton has actually stepped away from Google to focus on speaking out about the risks of AI. Now, he did an interview with the New York Times. The interview sort of made it sound like Hinton was stepping away because he wanted to talk crap about his former employees employee over at Google. But in his own words, he said, actually, I left so that I could talk about the dangers of AI without considering how this impacts Google. Google has acted very responsibly. So he has no bad words to say about Google. But his worries really revolve around the pace at which this AI stuff is really picking up. He thought that Google was doing a really good job at holding back, at releasing a lot of this AI technology up until Microsoft and OpenAI said started putting things into Bing Chat and this sort of AI war began to escalate, forcing Google to escalate the release of a lot of their technology. Now he has concerns over things like deep fakes, the spread of misinformation, and how it's negatively going to impact the job market and people's careers in the future. In his interview, he said, the idea this stuff could actually get smarter than people, a few people believed that, but most people thought it was way off. And I thought it was way off. I thought it was 30 to 50 years or even longer away. Obviously, I no longer think that. So the idea that one of the people that essentially invented the technology that we're seeing used by Meta and Google and OpenAI and Microsoft and all of these companies is now worried about the pace at which this technology is picking up, speed should raise some eyebrows. Now, in completely unrelated news, the company Palantir introduced an artificial intelligence platform for defense they released this demo video showing how it can take into account a whole bunch of information from a battlefield of where the enemies are, how long it's going to take for various vehicles to get to an enemy, how many weapons does any one military unit have? And it takes all of this data and suggests multiple courses of action on the battlefield and then sends them to a commander for a commander to make a decision on about which of those actions to take. You can see in this screenshot of their demo video. The captain was given three potential courses of action. Target with air assist, target with long range artillery, target with tactical team. With the time required for each, the assets needed as well as the weapons needed, distance to target and more information. You can see over on the map, it's got a breakdown of the landscape and where the various assets are located on the battlefield. If I'm honest, it does freak me out a little bit. Also, in totally unrelated news, lawmakers propose banning AI from single handedly launching nuclear weapons. No autonomous system without meaningful human oversight can launch a nuclear weapon or select or engage targets with the intention of launching one. Any decision to launch a nuclear weapon should not be made by artificial intelligence. They hope by publicizing this bill, it will spur similar commitments from China and Russia. Now don't get me wrong, I love all of this AI stuff. I love the AI tools, I love the creative elements. And it's been able to open up new pathways of creativity for people that maybe didn't have the skills or weren't capable of creating in the ways that they can create. Now, I love all of those elements of it, but somebody that was leading the charge in bringing this AI technology to the world is now backtracking and saying this is something we should all start to be a little bit concerned about. The fact that companies are starting to make map out their plans for the militarization of this technology, the fact that we even need to have conversations about whether or not AI should be allowed to launch nuclear weapons, that stuff starts to get a little bit scary. The deeper I go down the rabbit hole and learn about the technology underneath it, I think the bigger my tinfoil hat starts to get as well. I don't want to go all doom and gloom with this video. So I'm going to dive into all of the other fun stuff that's happening in AI right now. So let's go ahead and just change gears real quick. Also, recently, Google's DeepMind released this video of two autonomous robots playing soccer with each other. The robots are capable of trying to score goals they're capable of if they fall down quickly getting back up. And they're also capable of learning. The more they actually play soccer against each other, the better they actually get at it. And here's a video showing the ability for these little robots to get back up quickly. Now, I wouldn't bully them too much because we just got done talking about how they're learning to launch missiles and how we need to ban them from launching nuclear weapons. Maybe we should be a little bit nicer. But this does demonstrate their ability to quickly get back to their feet and run around again. Now, personally, if this was a sport on tv, I think I would watch this. I think it would be a blast. Put five on five in there playing soccer and maybe let them knock into each other and make it more full contact. I think that'd be fun to watch. You can learn more about this research over@sites.google.com view OP3 soccer. But I'll also make sure it's linked up in the description below the video if you want to look into it more. So we all remember the metaverse that Mark Zuckerberg had planned for us. The one that gives us amazing graphics like this with floating torsos with heads on them and no legs, and super cartoony looking conference rooms that nobody can take seriously. Well, it seems that Mr. Zuckerberg and Meta have upgraded their game a little bit and maybe we should start to change our expectations about what the metaverse might actually be. Now, this is a video that Mark Zuckerberg released that is actually super realistic. You could see the lighting change on him and as it moves around, the shadows on his face change. It's starting to look ultra realistic. And I think this is closer to the metaverse that we all want. You can watch this demo video by checking out my friend over on Twitter @AI breakfast. He shares a ton of great news in the AI space. So if you're not following him, definitely make sure to follow him. Now the real question is, are we going to get legs? All right. Now I wanted to get into some stuff that you can actually play with right now, today. So recently Microsoft opened up their Microsoft Designer tool and and they've made it available for everybody now. Now, this tool is similar to something like Canva, but with some additional functionality built in. And if you head over to designer.Microsoft.com you can get in and play around with the preview version of it Right now to start, you can create your image by just typing a prompt. So let's say I want an Instagram image to represent that. AI technology is moving super fast and let's just see what it generates for us ultra quickly. We have a whole bunch of images that it generated. Some of these are even little video clips if I preview them. AI technology moving fast. Keep up with the pace. Or a human shaking the hand of a, you know, a vector hand here. AI technology moving fast. Now, these images appear to be stock photos and not AI generated photos. But let's go ahead and pick an image here and click Customize design. You can see now we have a sort of canvas style editor here where we can choose a different template if we like. You can upload your own media. You could use some of the visuals that they have built in here with various stock photos and stock graphics and stock video that you can use, or you can generate your own images. There's a brand kit here where you can change the color scheme. So if I select this, you can see it sort of change the colors to a whole different palette. Or if I select this, it'll change the colors to a whole different palette there. Let's pick a different template real quick. Let's grab this forest image and let's say I want to put Bigfoot in the image. If I come to visuals and I click generate here, let's put Bigfoot. Now, I believe it's using dall E from OpenAI because OpenAI and Microsoft are working tightly together. And so here's a picture of a silhouette of Bigfoot. I could pull this image into my image here, make it larger. Click Remove background here. And now you can see it. Removed the background and I want to send it to the back. So let's send backwards. Do it one more time. So the text is in front of it there. And now I've got an image of Bigfoot walking through this forest here. And over here you can see. It'll give us alternate ideas that we can try. When you're done, you can download it and it's also got some additional social media AI features where you can have it automatically write captions and hashtags using AI. But again, if you've used something like canva before, this will feel very familiar to you and you can currently play around with it for free. Also, within the last few days, 11 labs introduced multilingual V1, which is a new speech synthesis model that allows you to create text in other languages. Now, it doesn't actually do the translation for you, but before everything was kind of in like a American accent. Now you can actually have it speak out in other Languages in sort of the proper accent, using any voice you want. So to do that, by default, it's set on 11 monolingual v1. If I come down here and select 11 multilingual, I can type anything I want here in any one of these languages. English, German, Polish, Spanish, Italian, French, Portuguese or Hindi, and it will speak it out in that language, but using my voice. So let's go ahead and type something in English here using the monolingual version. Hello, my name is Matt Wolf and I love to make videos that show off cool tools that use artificial intelligence. Let's generate that and it should play back in my voice now. Hello, my name is Matt Wolf and I love to make videos that show off cool tools that use artificial intelligence. All right, now let's go ahead and use Google Translate. I'll paste in the English version here and let's do it in Spanish. I'll copy the Spanish version. Let's jump into 11 labs here and let's switch it to multilingual version. And now it should say it in Spanish using my voice. Hola, mi nombres Matt Wolf. Imencanta a servideos que muestern intelligencia artificial. And you can see it was sort of my voice, but I was speaking in a Spanish accent. Now let's translate it to German. Hallo, mein ist Matt Wolf, und ich liebe ist Videos zumagen die. Now, I have no idea if that translation was correct or not. I'm assuming it was fairly okay. I just find it interesting that I can speak in my voice now in other languages. And so I can do things like make YouTube videos that are completely in German that still sound like me. I can take the transcript of the entire video, paste the whole thing into a translator tool like Google Translate, take the translation, pull it into 11 labs, and then export the audio file of me speaking the whole video in German, dub it over my existing video, and now I have an entire video of me speaking it in German. So probably something I'll test out in the future. Here's a tool that has been open sourced and made available called Espanol Love. And it's basically a real time translation tool Here it's using your OpenAI API key and it's also using the 11 Labs API that we were just looking at so that you can in real time speak out something in English and then it will translate it and then using your own voice, speak it back out in Spanish. So if you're ever on a trip and you don't know the language, you can use your phone. I'm not gonna argue that this type of thing hasn't already been around for a while. But now anybody can make this sort of app themselves with open source technology that's available and tools like 11 labs. So let's see what he made here. Where can I find the coolest toys in San Jose? Press translate and within just a couple seconds, hopefully I can figure out how I can find some Legos. Donde puedo en contralos julietes mas chevrez. Perfect. You can find this demo over at this GitHub URL here. But again, I'll make sure it's linked up in the description below so you can check it out. I just find it really cool that it will actually speak it back in your voice. It's not like a computer generated translation. It's speaking using your trained cloned voice from 11 labs. All right, now, finally, let's wrap up by talking about audio GPT. This is sort of an all in one open source audio tool. It could do things like text to speech style transfer, speech recognition, speech enhancement, speech separation, speech translation, mono to binaural text to singing, text to audio, audio inpainting, image to audio, sound detection, target sound detection, sound extraction, and talking head synthesis all in a single tool. And there's a hugging face demo where you can play with it. Right now I've messed around with it a little bit. You do have to enter your OpenAI API key here to be able to use it. And I've gotten sort of hit and miss results. Some of this stuff has come out really cool. Some of it has been a little. All right, that was funky. So let's start by having it generate some speech here. Generate a speech with a text. Subscribe to Matt Wolf on YouTube. Go ahead and run this and let's listen back. Subscribe to Matt Wolfe on YouTube. So we've obviously heard some better text to speech generators out there. I think bark is quite a bit better than what we're hearing from this right now, but not too bad. But we can also have this sing. So what if I type? Sing me the words. Subscribe to Matt Wolfe on YouTube. Let's see what happens if we tell it to sing instead of generate speech. Okay, so it basically stopped working on me and it just refreshed the page. Every time that happens, I've had to refresh, enter my OpenAI key again and start over. So let's try this again. All right, so I entered the prompt. Sing me a song with the word subscribe. To Matt Wolf on YouTube. And this is what I got out of it. So yeah, it couldn't figure out the words that I actually told it to sing. Now, it also supposedly can recognize an image and generate sound from that image. But if I upload an image, like for instance, this image of lightning striking here and then tell it to generate an audio of this uploaded image, I get something that sounds like this. I don't know about you, but I don't think that really sounds like lightning or thunder. So I like the idea of what this is capable of, a sort of all in one audio tool. But so far I haven't really gotten it to generate what I would expect it to generate. But it's open source, it's being built upon and I'm expecting at some point it'll be pretty cool. Anyway, that's all the cool stuff that I came across this week. I've been away from my computer for the last few days and spent some time catching up. I actually just celebrated my 40th birthday. You can see this cool little gadget that I got going in the background that was probably being a distraction in this video. This was one of the presents that my in laws gave me. It's really cool. Shows a bunch of memes and stuff on it. It's called the Tidbit. Also shows my subscriber count. My wife and I also got this cool trailer here in the last few days. So we're gonna kind of take the show on the road a little bit. And I'm excited to take some fun cameras and some cool toys and gadgets and start figuring out how to use some of this cool AI tech out in the world away from my office. So that should be fun. So some new videos coming about that. But if you want to stay in the loop, you love all this AI stuff. You love nerd. Not about all this with me. Check out FutureTools IE. This is where I curate all of the coolest tools that I come across in the world of AI and augmented reality and virtual reality and that sort of stuff. I also update the AI news page every single day. So you can stay in the loop there. And if this is all a bit too overwhelming for you, a bit too much to keep up with, join the free newsletter. And every Friday I'll send you the TLDR of what's going on in the AI world, I'll send you a handful of news articles, just the five coolest tools that I came across, a few YouTube videos, and one cool way to make money with AI. It goes out every single Friday. You'll get the just what you need to know about AI, and you get it only once a week so you don't have to stress about keeping up with it. And if you like videos like this, you want to see more of them in your YouTube feed. Give this button a like and maybe subscribe to this channel and I'll make sure you see more AI videos just like this one. I like to keep you up to date with the news and make some tutorials from time to time and experiment doing cool AI challenges. So if you like that kind of stuff again, maybe click subscribe and I'll make sure I show you more of this kind of stuff in your feed. Thank you so much for tuning in. I really, really appreciate you. I'll see you guys in the next video. Bye. By. Sam.