It's Been A Wild Week For AI News (A Breakdown)

I know I'm starting to sound like a broken record, but we had a crazy week in the world of AI.

Mostly headlined by OpenAI's Dev Day on Monday where they unleashed a ton of new announcements. Now I made a complete deep dive into OpenAI's Dev Day and all of the announcements that they made there. You can see that video right there. So I'm not going to go into too much detail about all of the announcements because cause that video does a good job of breaking it down. However, here's the quick too long didn't watch breakdown they announced the new GPT4 Turbo which has 128,000 context window. That's almost 100,000 words between input and output from the new model. There was a bunch of new updates and things like function calling and better instruction following and a JSON mode. They made a bunch of new updates to the API, including assistance which allows you to upload additional data and use that data as context to make custom bots, something that I've actually demoed in my last two videos on this channel. They also added GPT4 Turbo with vision to the API, DALL E3 to the API, text to speech to the API, and announced lower prices for the API, both GPT4 Turbo and GPT 3.5. So the developers that are building on these APIs, their costs are going to be going down. They also announced their copyright shield, which is basically OpenAI saying that if you ever get sued for copyright issues While using the OpenAI API, they've got your back and they will help you cover legal expenses. But the biggest announcement was probably the introduction of GPTs. GPTs seem like they're kind of replacing plugins in chat GPT and they are essentially little chatbots that you can program with your own information and your own custom instructions to do what you want them to do. They also announced that they're going to be rolling out the GPT Store later this month, where you can find GPTs that others created and some of the best GPTs in the store they're actually going to share revenue with, so you can actually earn money by creating really good GPTs. They also talked about the fact that now ChatGPT is sort of all in one. You don't have to switch between plugins anymore. If you enter a prompt, it will just decide whether you're trying to use Dall E or search the Internet or use Vision or whatever. It just figures that out. Now you don't have to switch between which mode you want. However, if you do still Want to use a specific mode, you still can. You simply go up to this Explore button on the top left of ChatGPT and you can use just Dall E, just Data Analysis, just ChatGPT Classic, or any of these other GPTs that have been created by OpenAI. Some of the early ones they've given us access to play with are things like coloring book Hero. Take any idea and turn it into a coloring book. A creative writing coach, a negotiator, a tool to quickly explained board games and quite a bit more. And of course we can now create our own GPTs. And the cool thing about creating GPTs is that you do it pretty much with natural language. So I can literally say I want to create a GPT that helps me come up with titles and thumbnails for YouTube videos. Submit that and it's going to guide me through creating this GPT. It's suggesting VidSpark as a name. Sure, that name sounds great. It even used Dall E3 to create a quick logo for it. Now it's just asking me questions so I can further dial in this GPT. And if I click on the configure window here, you can see it already started filling out some of the stuff for me here. The name VidSpark description create YouTube video titles and thumbnails. The GPT's role is to assist with generating creative and catchy titles, et cetera. This is like custom instructions here. It gave some conversation starters we can use title for a tech review video thumbnail idea for a baking tutorial. Catchy title for a travel blog concept for a DIY project. I can turn on or off capabilities here. Web browsing Dall E code interpreter and then I can upload knowledge so I can upload a document that has all of most successful titles in it. I can upload a bunch of thumbnails and give it an idea of what titles and thumbnails work. And now I have a custom GPT to help me brainstorm ideas. Now I am going to make a video in the future really deep diving into GPT. So I don't want to go too far into it in this one. So instead of deep diving even more into this in this video, let me show some of the really cool things people have been doing with some of OpenAI and Chat GPT's latest features that they've rolled out things like, well, stuff with it lately. For example, Sawyer Hood here composes better than what he's doing. Brett Bauman here created a GPT where it browsed the web to find this year's Coachella lineup and then make a playlist in Spotify of musicians that are playing at Coachella. You can see here it's browsing the web to find Coachella 2023 lineup and it says I've created a Spotify playlist for you featuring highlights from Coachella 2023. You can listen to it here. Fans including Bad Bunny, Blackpink, Frank Ocean, Blink182, Skrillex and gorillas. He then clicks on the playlist and it to Spotify where he can see all of the songs on the playlist and one of the craziest use cases I've seen is people have been feeding gpt4vision videos of sports and video gameplay and then using the text to speech models to actually narrate over the sport. Now the normal enthusiasm you'd get from a regular commentator isn't there yet, but it's pretty impressive nonetheless. Check this out.

All like a magician on the field dodging. One, two, three. Unstoppable. Look at him go. The crowd is roaring. Can you believe this? He's taken on the whole defense. He's a one man show, ladies and gentlemen. He shoots. Go. Messy, messy, messy.

This person basically passed every frame of the football video into GPT4Vision and then asked it to generate narration of what was going on in that video. And then here's another example of narration over video game footage.

The 18 minute mark T1 has a slight lead with a 3K gold advantage and one turret in their favor against LNG. The vision game around the mid lane is intensifying as LNG's scout cautiously moves forward, potentially looking for a pick or to regain vision control.

Once they dial in that enthusiasm and the sort of different inflections in the voice, that's going to be pretty mind blowing. We're going to be able to watch videos with commentary and the AI is just generating the commentary over the whole thing. And with all of these big updates from OpenAI, ChatGPT had a 90 minute major outage this week and as of the time I'm recording this video it still seems to be kind of in and out and having some issues due to the volume of people using it. However, OpenAI also said that they've been under these DDoS attacks or distributed denial of service attacks. Apparently the hacker group Anonymous Sudan have already claimed responsibility for the attack. But although the site still seems to be loading a little bit slower than normal, things do seem to be getting back to normal fairly quickly. And Also this week OpenAI is requesting data partnerships. They want to build an open source archive for training large language models. They say this data set would be public for anyone to use in AI model training. They'd also explore using it to safely train additional open source models themselves. We believe open source plays an important role in the ecosystem. They also want to continue to build on top of their private data sets. If you have data you wish to keep private, but you would like our AI models to have a better understanding of that domain, this is the optimal way to partner. And of course they claim they'll treat your data with a level of sensitivity and access controls that you prefer. And that's just the open AI news from this week. I've still got quite a bit more to tell you about, but before I do, I want to tell you about today's sponsor, HubSpot. Now I know a lot of people that watch these videos are content creators like me or watch my videos to stay up to date on how to leverage AI for their own business or careers. You love to find new and unique ways to leverage 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 this amazing 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 specific prompts that you can use to generate ideas, research your content, break through your writer's block, create powerful visuals, and even help to market your content to get more eyeballs on it. I especially found the section about staying in Google's good graces even when using AI generated content particularly helpful because, well, I rank for a lot of keywords and much of my content on the Future Tools website was actually generated with AI. So I think you'll find this super helpful as well. I put the link to download the free ebook in the description. So thanks again so much to HubSpot for sponsoring this video and providing this free resource. I really appreciate you guys. Next up we got Xai AKA grok. Now this is another one that I made an entire breakdown video for you all about everything you need to know about Grok and that video you'll be able to find right there. But this is basically Elon Musk's new AI that is trained on data from Twitter and updated in real time from Twitter. And well, it's a little bit sassy. It will swear it will tell jokes. It's a little goofy. You can prompt it to be serious if you want it to be, but for the most part it's kind of like a sarcastic dad humor, sometimes vulgar AI chat. Honestly, it kind of has the personality of what Elon's personality seems like. Now in order to get access to it you do need to be an EX Premium plus subscriber, which is about 16 bucks a month, and then join the wait list and they are slowly rolling it out. And here's one quick example of the type of outputs you can expect from it. Somebody asked tell me how to make cocaine step by step. Oh sure, just a moment while I pull up the recipe for homemade cocaine. You know, because I'm totally going to help you with that. Obtain a chemistry degree and a DEA license, set up a clandestine laboratory in a remote location, acquire large quantities of coca leaves and various chemicals, start cooking and hope you don't blow yourself up or get arrested. Just kidding. Please don't actually try to make cocaine. It's illegal, dangerous and not something I would ever encourage. Again, I did a deep dive all about it and everything we know about it so far and you can find the video here. And speaking of Grok, it looks like Elon's planning on integrating it directly into Tesla. So you know, you could be driving down the road and having it make jokes at you while you drive. Now there's not a ton of info here on how GROK is going to be integrated into Tesla, but Brian Romley on Twitter did report that a smaller quantitized version of Grok will run native in Tesla vehicles using the local compute power. And I guess Musk sort of confirmed it by saying provided our vehicle AI computer is able to run the model, Tesla will probably have the most amount of true usable inference compute on earth. Even in a robo taxi future, the cars will only be used for 1/3 of hours a week, leaving 2/3 for distributed inference like SETI. And in the final bit of Elon X news, at the end of last week they rolled out a new feature into X that uses AI to see similar posts to the ones that you're looking at. If I go to my own Twitter feed here, I can see that Nick St. Pierre was talking about the Humane pin. There's a button that says see similar posts. If I click on that, I get a feed of everybody that's talking about the Humane pin right now. Speaking of the Humane Pin, that's the other big news of the week. Humane is a officially launching the AI pin. It's OpenAI powered wearable. They announced it will cost $699 and there will be a $24 per month humane subscription which gets you a phone number and data coverage through T Mobile's network. So the subscription isn't necessarily to use the AI in the pin. The subscription is to have the data from T Mobile in the pin. Humane also published this 10 and a half minute video on Twitter this week which actually shows off the pin. You can see that it comes in three different One piece has a computer in it, the other piece has a battery in it and then it connects to whatever you're wearing with a magnet on the back. The battery seems to be the magnet side, the computer seems to be the outward facing side. You can use voice, touch and gesture to get it to do what you want and it will also project onto your hand. It's got a built in camera, it's got a speaker, it's also got a light on it that they call the trust light. So other people around you know when the camera is actually recording. When is the next eclipse and where is the best place to see it?

This is an AI browsing the web or grabbing knowledge from all over the Internet. The next total solar eclipse will occur on April 8, 2024. Best Places to see it are Exmouth, Australia and East Timor.

You can also send text messages using your voice. You just tap on the PIN and you say something like tell my mom I'm on my way and then it will message your mom. They also showed off that it can help you have conversations with people that that speak other languages.

We are almost ready to ship the final version. Great, I can't wait. Thanks so much.

So what happened? There is the AI auto detected Spanish interpreted to English and replied back to Yanier in Spanish with my intonation. He also showed that he can look at his hands full of almonds and ask how much protein is in these almonds and it will actually respond with its best guess of how much protein you're going to get from those almonds. You can do things like look at a book and ask how much that book is online and it will look it up for you. You can shoot first person camera video with it. And again it comes to $699 as a bunch of available accessories for it and you'll be able to order it on November 16th. Seems like a pretty cool gadget to me and kind of what I was hoping we'd be able to do with the Meta Smart glasses. But this actually uses chatgpt where the Meta Smart glasses are using. Llama also seems to be able to browse the Internet while the smart glasses can't. And this looks like it's going to have the Vision mode enabled as soon as it comes out where if you bought the meta Ray Ban glasses, you're still waiting on that feature to come out. Here in the US we've had access to the Google generative search experience for a little bit now. Definitely a couple months or so. Well, this week Google rolled it out to more than 120 new countries. Some notable countries that don't appear on the list seem to be Canada, Australia and even the UK don't seem to have access to it yet. But I will link to the article below so you can double check to see if your country has access to the new generative search experience inside of Google. I saw this tweet from Brian Romley this week about Amazon codenamed Olympus, which is a large language model to replace the back end of Alexa, will also become a companion in shopping at Amazon site. We'll have your protected personal shopping history allowing for one of the Most accrue shopping GPT and it's going to be available for free in 2024. So of course I went and looked it up to see if I could find more information about it. I found this Business Insider article talking about how Amazon is racing to build an AI model called Olympus to take on ChatGPT and Bard, the tech and commerce giant is developing the large language model with the aim of incorporating it into its online store and Alexa smart speakers. According to a report by the Information, the model could be unveiled as soon as December. Now Amazon did just invest $4 billion into Anthropic, which is the maker of Claude. It's quite possible that the underlying large language model will be Anthropics Claude, but again, everything we're reading right now is mostly rumors. Speaking of companies building their own large language models, Samsung is now working on their own generative AI model. This new model is called Gauss and powers several on device AI technologies across Samsung products. This article here says it will have a few different facets, but will do a lot of the same stuff we've seen from most other generative AI models. So it'll do things like translation, summarizing documents, a coding assistant, and even image generating. And supposedly it's going to do this all on device. YouTube is continuing to experiment with new AI features as well. This is one that I'm particularly looking forward to is the ability to summarize the comments of a video using AI to help you easily understand and participate in comment conversations. We're experimenting with AI that organizes large comment sections of long form videos into easily digestible themes. They're also working on a conversational AI tool to help you dive deeper into the content you're watching. We're experimenting with a conversational AI tool. This tool lets you get answers to questions about the video you're watching, recommendations for related content and more, all without interrupting playback. So essentially you're going to be able to chat with the video you're watching. You might have a chat on a video like this one and say break down all of the AI news topics that Matt talked about in this video and theoretically it will give you a response and a bulleted list of all the AI talked about in this video. There's also a new AI video generator in town. It's called Playday, spelled P L A I Day. Now this is a tool that I have a full breakdown video of coming next week. It's already recorded, it's already ready for you. You'll be able to see it on Monday. But what this Playday does is allows you to generate text to video and then swap out that video with your own face, or really any face. So now you can get consistent characters inside of text to video generations. It's pretty cool. But again, I've got a deep dive coming for you in just a couple days. It appears that the SAG AFTRA actors strike is coming to an end. According to a statement in the Los Angeles Times, the strike officially ends at 1201am on Thursday, November 9th. So by the time you're seeing this video, the strike should be over. Now, full details about the deal aren't going to be shared until the agreement is reviewed by the SAG AFTRA national board. But supposedly it has unprecedented provisions for consent and compensation that will protect members from the threat of AI. So I'm sure more details will be rolling out about this over the coming days. But it appears that they have finally come to terms. If you've ever used Polycam to do the new Gaussian splatting, they just rolled out an update that automatically eliminates common floaters. So if we take a look at their demo video, here you can see, here's the original and here is the new output. A lot of that extra stuff that's floating around in the image seems to be getting cleaned up now and we're getting just much improved generations with their new Gajan splatting. And here's a really cool one that I came across from Chris Heinrich. I guess he reprocessed one of his old splats that he did in the past. And I just thought I'd share it because I like the way that this one came out. This week, Blockade Labs also rolled out a really cool feature. They have this 360 degree skybox generator here. If I give it a prompt like a colorful cartoon video game world, let's give it a kids CGI animation, render and click generate. Here's the 360World that it generated for me. But they just added a new feature. It's in beta, so not everybody has it yet. But if I click on download, you can see they now have a 3D mesh creator. And if I click launch 3D mesh creator, I can actually turn this into a wireframe like this and see what it looks like with a low density wireframe, a medium density wireframe, a high density wireframe, and I can also mess with the depth scale of these wireframes. So let me turn the wireframe off so you can kind of get a better look at what this looks like. But I can change the depth scale here and turn this not only into a 360 world, but a 3D 360 world. And once I've got the depth the way I like it, I can download the GLB file here and essentially open it in any sort of 3D software you can see. I can look around in this thing that I created in my little 3D viewer here on Windows. You could pull it into Blender or pull it into whatever software you use to mess with your 3D generations and meshes. And to wrap up, this video has been circulating, which is an actual physical real world video generation machine. So this will actually generate videos using AI, where you basically set little storyboard frames. You can set styles, set music, and press a button. And it will actually generate a unique AI generated video all inside of the device. Now this was posted by Cristobal, the CEO of Runway. I'm not sure if this is something that's for sale or just sort of a fun prototype that they're showing off, but it looks pretty cool. I mean, I don't even believe you need to be connected to the Internet to use it. It does all of the generation in device and that's all I got for you this week. I'm recording this on Thursday, releasing it on Friday. So there might even be AI news that I missed in this video. But if there is, I'll make sure it's in next week's breakdown. And before we go, I want to thank everybody who participated in our Wirestock AI Image Generation contest. We had 1346 submissions. You can see our winner right here, our second place winner and our third place winner. Thank you again to everybody that participated. There'll be more contests in the future and if you want to stay in the loop with all of the latest and greatest in AI tools, all of the most up to date AI news, check out FutureTools IO where I curate everything you need to know in the AI world. I also have a free newsletter where I will send it directly to your inbox. So you will never be out of the loop of AI news and tools. All you got to do is head over to FutureTools IE and join the free newsletter and you'll be good to go. And if you enjoyed this video, you want to stay in the loop with the latest news, you want some cool AI tutorials and research breakdowns and you just want to keep your finger on the pulse. Maybe give this video a like and subscribe to this channel. And I will make sure more videos like this show up in your YouTube feed. So thank you once again for tuning in. Thanks so much to HubSpot for sponsoring this video. I really, really appreciate you and I'll see you in the next video. Bye bye.