AI News Update - ChatGPT Costs How Much?!?

I just got back from the artificial conference in LA and it was an awesome event, but I'm ready to get into the news. This week a report came out that ChatGPT is costing $700,000 per day. Most likely, Microsoft's $10 billion investment in OpenAI is what's keeping the company afloat at the moment. But since ChatGPT's peak in May, the amount of users has steadily declined each month. However, those have been the summer months when students are out of school. So now that schools are getting back into session, it'll be interesting to see if we see a rise of ChatGPT again. However, this report that's in business standard.com claims that OpenAI could end up going bankrupt by the end of 2024 if they don't start to turn a profit, and if Microsoft turns off the faucet of cash flow to them. Amazon recently announced that they're going to continue to improve the customer reviews experience with generative AI. Now, this is rolling out to a subset of mobile shoppers in the US across a broad section of products. We can see in this screenshot here a AI generated review that seems to be a summary the overall customer reviews on a specific product. They're also adding some buttons for some commonly searched things about various products like performance, ease of use and stability. And you can click on it and get an idea of what users are saying about that specific point of the product. So if you click on ease of use, for example, it will write you up an AI generated summary of what people said across all of the reviews about ease of use. That seems like it'll be pretty handy. However, if people keep on gaming the Amazon reviews like they do now, we're just going to have a gamed summary, so we'll have to see how Amazon deals with that as well. Also this week, Google updated their generative search experience, making it that if you hover over certain words, it'll actually give you a definition of the word that you hover over, and also allowing you to generate summaries of articles that you're reading to just get a quicker look at what the article says. They've also updated the coding capabilities. Now, if you do have access to the generative search experience already, you should have access to these new features as well, although you will have to make sure you update to the latest version of Chrome to get them to work. However, this summary feature is only available on the Chrome, iOS and Android apps. It's not available on the Chrome browser yet. In other Google news, they've added an AI update to Google Photos. Now, if you use Google Photos at all, you've probably seen it resurface memories from this day in the past. Well, they just added a new MEM Memories View. It's a home for your memories that is automatically curated and organized with the help of AI. It lets you easily relive, customize and share your most memorable trips, celebrations and daily moments with your loved ones. It started rolling out this week in the US and will be available globally in the coming months. You can actually rename the memories using AI so it will look through the photos and actually suggest a title for your memories, like a desert adventure and should just overall make it easier to resurface photos that you've taken in the past because I mean, how many photos on your phone have you taken that you never ever go back and look at? Something like this will bring them back to your attention again. This week, OpenAI announced a new GPT4 API feature where you can actually use GPT4 for moderation. Typically, content policy changes can take months to roll out because they require training a large number of human moderators. In contrast, GPT4 reduces the cycles to hours and is significantly more flexible. Our vision is a world where AI guided by humans can create a safer environment as content scales more and more. So the idea being you can use their API to build moderation tools for communities and things like that, freeing up time and catching things that human moderators might not otherwise catch. Also in OpenAI news this week, OpenAI acquired Global Illumination. They put out a short press release on their website about how the entire team has joined OpenAI to work on core products, including ChatGPT. Now we don't actually know what the Global Illumination team is going to do for OpenAI, but what we do know is that they've created games in the past like this Biomes game, which if we take a peek, looks very similar to like Minecraft or maybe the sandbox game. It's leaving the AI world speculating. Is OpenAI getting into the gaming world, or was OpenAI just a fan of their development team and wanted their development team working on ChatGPT instead? Speaking of gaming, Take 2 shuts down an AI GTA 5 mod now there was a playable mod for GTA 5 that used InWorld AI's character speech generation tool so that you can walk around inside of Grand Theft Auto and actually talk to other characters and those characters would respond to you using AI and you can have conversations with characters in the Grand Theft Auto world. Now take two is the company that owns Rockstar Games. The creator of Grand Theft Auto and Take two games apparently just took this mod down without any rhyme or reason, the creator of the mod said. As far as I'm aware, the mod is not violating grand theft auto or YouTube policy, referring to the fact that people have made YouTube videos playing this mod and some of those videos have even been taken down off of YouTube. Still, as of right now, there's no official word as to why the mod was taken down or what rules were actually broken by the creator of the mod. Also this week, Adobe Express with Firefly built in is now available worldwide and for free. If you head over to adobe.com express you can click Get Express for free and use all sorts of creation tools like their Text to Image generator, their text Effects generator, and much more. They even have things like AI background removers. One of my favorite features of Adobe Firefly is this sort of text texturing feature that uses AI to generate textures. So for example, I put some text like this on the screen and type shiny black metal and click Generate. It changes all the text to a shiny black metal effect. But you can also just use Adobe Firefly completely for free and generate any images you can imagine right with their text to image generator. So pretty cool that they made that available for free right now worldwide. Also this week, Runway, the company behind Gen1 and Gen2, the cool video to video effect creator, as well as the now text to video and image to video creator that I've talked about in past videos, they just rolled out a new feature called Watch. If I jump into my Runway account here, you can see there's a new button over here on the left that says Watch. We can check out some really cool videos that have been made with Runway ML. For example, this Echoes of the Valley video made by my friend Amar, which I highly recommend checking out. It is just, just absolutely amazing and really, really well done. But if you need some inspiration or you want to see some of the cool types of things people have been making with Runway, log into your account. I believe this Watch feature is available for free. This week, the Associated Press also set some AI guidelines for journalists. Journalists that are part of the Associated Press should follow these guidelines that say although you're allowed to use OpenAI, you cannot use it to create publishable content. Any output from generative AI should be treated as unvetted source material. So whatever the A say, they still need to vet it, which is probably pretty sound advice. Regardless, journalists aren't allowed to alter any elements of their photos using generative AI. They're not allowed to transmit any AI generated images that are suspected or proven to be false depictions of reality. However, if an AI generated illustration or work of art is the subject of the news story, it can be used. Their worry is that generative AI makes it easier for people to intentionally spread misinformation. And overall, reading these guidelines, I'd say they're pretty solid guidelines for anybody that's creating content online that they want to be taken seriously. Don't try to pass off fake images as real images, don't intentionally try to spread misinformation. And if you are going to use AI to help create content, use it as a rough draft and definitely check all of the information to make sure it's actually accurate is probably a good advice for anybody. This week there was also a string of announcements from ex company employees starting new AI companies. For example, Google's old CEO Eric Schmidt is launching a new AI company. His new company is aimed at modeling what OpenAI originally stood for in that it's a nonprofit designed to use AI for scientific breakthroughs. He's hoping to recruit some of the top AI scientists to come over to this company that might have had those initial ideals that OpenAI started with before they became a for profit company. Two ex Google researchers also created a new startup called Sakana AI. The company was founded by Leon Jones who actually co authored a breakthrough paper called Attention Is all youl need back in 2017. It was also co founded by David Ha, who was previously the head researcher at Stability AI and also formerly worked at Google Brain. They founded the company in Tokyo and they're hoping to take a different approach to generative AI. They claim that this new model, instead of using a single large language model, they want to create numerous smaller AI models that collaborate much like a swarm to deliver complex results. It sounds like over time they could be an alternative to companies like OpenAI that create models that generate text, images, code and multimedia content just from this newer approach that they're suggesting. Keeping with the theme of people leaving companies to go start new AI companies, the ex CEO of Machine Zone is launching a new AI based social media app called Be Fake Now. Machine Zone was a company best known for making mobile apps games that were freemium that had all sorts of paid upgrades in the games. This new app is designed to take pictures of yourself, but then actually transform them with generative AI and post fake pictures of yourself instead of all the constant selfies of your real life. It's a social media App where they encourage sharing your fake life. I guess some interesting research this week, it was found that AI bots are so good at mimicking the human brain and vision that CAPTCHAs are now useless. Now, captchas are these annoying things where it asks you to click the images of buses or bikes or whatever and then you click them all and then it takes you to another page where you got to do it again. And you know they're annoying. And in the near future they're going to be completely pointless because AI can solve them for you. And if you didn't know, captcha actually stands for Completely Automated Public Tiering test to tell Computers and humans apart. So if you're like me and didn't even know that that was an acronym, now you know. Now the fact that these AI bots can actually beat CAPTCHAs is sort of a huge risk. If people create autonomous bots that can brute force crack into websites and solve CAPTCHAs, it could wreak havoc on a lot of websites. We're going to have to come up with some sort of new proof of humanity really quickly. And something that I kind of considered non news, but I see other AI people talking about it, so I might as well talk about it as well. Snapchat's my AI chatbot glitched out this week and actually posted a random story that looked like a corner of a wall in a ceiling. And when people actually tried to chat with Snapchat's my AI, it just wasn't responding to them. But a spokesperson for Snapchat basically said, yep, it experienced a temporary outage, but it got resolved real quick. So again, in my opinion, sort of non news. A few weeks ago I talked about Neuro Angelo, a technology from Nvidia that is not quite a Nerf, but it uses very similar technology to a Nerf where if you take a whole bunch of images around a location or an object or it will actually turn that into a 3D image. And with Neuro Angelo, it actually creates a real smooth, cohesive image. This Neuro Angelo technology has now been made available by Nvidia and you can actually download it and run it on your own local machines if you want. And along a similar vein, Luma AI just launched Flythroughs. This is a mobile app where you can actually just walk around, film a video of a location, and it will actually turn that location into a 3D nerf and make it look like a drone camera is flying around that location. Now, as of right now, it's very limited. You can't really control the camera or the speed or you know, set up keyframes of what the video looks like. It's actually just going to show whatever you showed on your camera and just turn it into something that looks like a drone shot. But over time this is going to be very interesting because I imagine they're planning on turning this into something where you can actually virtually walk through the houses and move a character around in first person view in whatever 3D room you create. Now you can already do stuff like this with Luma Labs regular AI tool. However, this fly throughs tool is at the moment being sort of angled towards real estate and showing off houses and stuff. But the technology is really cool. That'll be a fun one to watch and see how it plays out. You can actually download it right now on the Apple Store and use it on your iPhone. I imagine it'll be available for Android in the near future as well. But if you have an iPhone, you can download it for free and play with it right now. And that's what I got for you today. Over the last several weeks, I've spent a lot of time at various conferences. I was at Siggraph, I was at the Artificial conference, I've been on a lot of podcasts, I've been taking a lot of meetings and I feel like I've sort of fallen behind on this YouTube channel, actually only posting two videos a week for the last few weeks. I want to ramp it up again. I have so many ideas for videos. I've just been traveling so much and going to all these conferences, but I've completely cleared my schedule. I'm getting better at saying no to conferences and podcasts and all these other things that people are asking me for. I'm going to put my head down and just focus on creating cool AI content, staying deep in the research, deep in learning about the new tools, updating the future tools website and making better and better videos and AI tutorials and news and research breakthroughs and all the cool stuff that I was doing a few months ago. I'm going to get back on track with making those style of videos and I'm excited to share some of the stuff and some of the ideas that I've got for videos. So if you haven't already subscribe to this channel, give this video a thumbs up. And one thing that would really, really help is if you click the little bell notification button. That'll make sure that whenever I put a new video out, you will see it on your YouTube feed just by subscribing YouTube sort of changed the rules. You don't see the videos automatically if you subscribe. You got to hit the bell if you want to see them for sure. But I am planning on ramping up the cool AI videos and there is a lot of cool stuff in the works from all of these conferences I've been to. There is some amazing, amazing AI tools and technology that is about to be released really soon that I'm excited to deep dive in and show off for you and play around with and make tutorials around. So expect that in the near future. And if you haven't already like and subscribe. I really appreciate you. Check out FutureTools IO. Lots of cool stuff over there if you're into the AI stuff and nerding out. And again, that's all I got for you. I'm done rambling. Appreciate you. See you in the next video. Bye bye.