All The AI News You Missed This Week

There's been a handful of exciting things that have happened in the AI world this week, so let's go ahead and break them down, starting with the fact that X, formerly known as Twitter, updated their Terms of Service to include the ability to train any of your data on AI. So they actually updated their terms of service under 2.1 operate, improve and personalize our services to say we may use the information we collect and publicly available information to help train our machine learning or artificial intelligence models for the purpose outlined in this policy. This article here on Stack Diary is making the prediction that at some point Elon is planning on merging X Corp with XAI Corp and using all of the data that it trained from Twitter in its new upcoming XAI platform. Also this week, Zoom rebranded what it was calling Zoom iq. They're now calling it the Zoom AI Companion. One of the most interesting things about this Zoom Companion is that you'll be able to ask questions of it, just like you do with ChatGPT and and it will reference all the conversations you've had in past Zoom conversations to answer your queries. In this chat, it says here, users will be able to query the AI Companion for the status of projects pulling on transcribed meetings, chats, whiteboards, emails, documents, and even third party apps. They'll be able to ask the AI Companion questions during a meeting to catch up on key points, create and file support tickets and draft responses to inquiries, as well as features that it was already able to do like have the AI Companion summarize meetings, automatically identifying action items and surfacing next steps. Zoom even put out this video showing somebody asking a question what has been discussed so far? And the AI Companion catches the person up as they join the Zoom meeting. So even if someone's late to the meeting, they can still get caught up to date on everything that's been talked about. That's something that we also recently saw Google talk about at their Cloud Next event when they showed off that they were rolling out similar features with Google Duet. The video also showed off a handy Chapters feature. You can see down here at the bottom where you can jump to specific parts of the meeting. A lot of these new Zoom features look pretty co, but as we've been seeing, a lot of these other platforms like Google and Microsoft are also rolling out pretty much the exact same features inside of their competing products. So it'll be interesting to see if Zoom can actually differentiate themselves and do something at least a little bit different from what we're starting to see Google and Microsoft roll out now. Probably the Biggest news in the AI world this week was that the open source Falcon 180B was launched, making it the largest publicly available open source large language model to date, completely eclipsing Llama 2's 70 billion parameters. The capab of Falcon 180B sits somewhere between GPT 3.5 and GPT 4, so you're probably going to get better results and better output than you will with like the free version of ChatGPT, but not quite as good of results as what you'd see with something like GPT4. Inside of ChatGPT, it kind of falls in between and across the benchmarking test that was done with Falcon 7DB. It actually outperformed Google's Palm in most scenarios. And on the hugging face leaderboard, Falcon 180B is the highest scoring openly released pre trained large language model, surpassing Meta's Llama 2. So it's really exciting to see a model this large and this powerful being open sourced so that anybody out there can build off of it and iterate off of it. And anybody, anybody can go and use it and play around with it. Right now. Because it's available on Hugging Face, you can go use the Falcon 180B demo. I'll make sure it's linked below in the description. But you would use it just like any other chatbot. Now I'm not going to get too deep into it in this video. I'll probably do a different video in the future where we break down this chatbot. Specifically, I want to try to rapidly get through the news. For example, let's enter the prompt what are the most exciting scientific Discoveries of the 2020s? And click Submit and you can see it quickly starts generating a response here. I do not have personal opinions, but I could provide information about some of the most exciting scientific discoveries that have been reported so far in the 2020s. Here are some examples. The development of MRNA vaccines, gravitational waves from neutron star mergers, CRISPR detection of water on the moon and then it cut off on number five. Probably because I have this token limit way down. Let's see what happens if I bring this token limit way up and try again. And you can see by cranking the token limit up I got a much longer output where it gave me a full 10 scientific advances of the 2000 and twenties. Speaking of large language models, the not open source anthropic Claude just released a pro plan and it seems very similar to what Chat GPT did with their pro plan. If we take a peek here, it costs 20 bucks a month to use it and what you get for it is five times more usage than the free tier provides. Which is interesting because I've actually never gotten to my limit on the free tier. I mean, I use Claude quite a bit, but I've never pushed that limit to where it says I can't use it anymore. So I'm not actually sure where the free tier limit is yet. You get priority access to cloud AI during high traffic periods, very similar to what ChatGPT did. And of course you get early access to new features that help you get the most out of Claude. Now, personally I use Claude more than Chat GPT these days. I use it for summarizing articles When I write my Friday newsletters, I actually have it double check a lot of the grammar and spelling for me on those newsletters. It's been a pretty valuable resource to me. But again, I actually haven't hit my limits yet of what the free plan is capable of. This week OpenAI introduced a Canva plugin for chat GPT plus subscribers to enhance content creation. Taking a peek here, if I come over to my chat GPT account, click on the GPT4 up here, click on plugins, click where it says no plugins enabled, and then click on the plugin store. Let's do a quick search for Canva. Install the Canva plugin and now that we have the Canva plugin selected, let's test this quick prompt here. Recommended by Paul Cuver. I'm launching my new AI related app Create a real Instagram template related to AI and new technologies. Let's see what happens if we use that exact same prompt here over on ChatGPT. Paste that in there and it actually gave us a bunch of templates here. Dark blue futuristic AI. The future Instagram post. I click on check it out, it opens up my Canva and this post is right here inside of Canva. What it did was it found an existing template for me and it just opened that template for me in Canva it was an existing one. Here's one called Abstract Artificial Intelligence Technology Free logo. Not a bad looking logo, but I have a feeling it's not unique. And right there you can see it's already a pre made template. So this plugin's not actually building anything thing for us. It's just essentially searching Canva for existing templates based on what we asked for. Still pretty handy because sometimes searching through Canva for the exact right template could be tedious. So chat GPT just bringing the 5 most relevant to our attention actually kind of useful. And speaking of OpenAI, they'll be hosting their first ever developer conference on November 6th. If we take a peek here at the official announcement, the developer conference will be in San Francisco and it'll bring hundreds of developers from around the world together with the team at OpenAI, to preview new tools and exchange ideas. So that'll be really interesting because there could be some new announcements and new functionality of Chat GPT that gets revealed during this event. In this quick post on the OpenAI site, Sam Altman said, we're looking forward to showing our latest work to enable developers to build new things. Now, I'm not clear if Sam Altman or Greg Brockman or Ilya Suskever or any of the, you know, bigwigs over at OpenAI are actually going to be in attendance, but that would be pretty cool if they are. If I wasn't going to be on a family vacation during that exact week, I'd probably make the trip. Up next, I want to tell you about the sponsor of today's video, Cohesive. So Cohesive is an all in one text generation platform, image generation platform, and it even does text to speech with realistic voices. And to make it even simpler than that, they have tons and tons and tons of templates that you can use, so you're not starting from scratch. You don't need to be some crazy prompt engineer. Just pick a template like a transactional email or a legal memo generator. Or maybe you want to create a romance audiobook. They got it in here. It took me forever to scroll to the bottom, but when I finally got here, some of the stuff that's hidden way down here, menu prompt calls, appointment reminder calls, a subpoena generator. But let's go ahead and use the one that I'm probably most likely to use myself here, a script for a YouTube video. It already starts the prompt off for me. Here. Write a YouTube video script for a five minute history of artificial intelligence. Target audience, Tech enthusiast video length 5 minutes. Sounds about right. I'm going to leave the reference links empty. Let's just see what it generates itself. And click. Go. It's doing the research for me. It found a few YouTube videos. So let's ahead and click. Generate. All right, so now it's actually read the transcripts of those YouTube videos and is generating a unique script based on what it learned. All right, now we got a script. Hey there tech enthusiasts. Welcome back to another exciting video on my channel. In today's episode, we're diving deep into the fascinating world of artificial Intelligence. Why am I reading this myself? Let's have Sam read it instead. Hey there, tech enthusiasts. Welcome back to another exciting video on my channel. In today's episode, we're diving deep into the fascinating world of artificial intelligence. Not bad. Not only do I have a script for a YouTube video, I could generate audio with all of this text, turn it into an audio podcast, and I can copy all of this text and turn it into a blog post. Look at this. It even gave me example shots that I can add throughout this. So shot A visualization of scientists at the Dartmouth Workshop. Let's go ahead and copy that prompt there. Do a little control slash and generate an AI image. And let's go ahead and generate a visualization of scientists at the Dartmouth Workshop. Obviously, this is a suggestion for a YouTube video, but why not use the same suggestion for the blog post portion as well? And there we go. Now, I'd probably take this image and crop out the funky text, but that kind of reminds me of one of those, like, sketches that they show in courtrooms. Kind of cool. So that's Cohesive, a platform that'll do research for you, write content for you, generate images for you, do the voiceover work for you, and even give you ideas for titles for the content when you're done with it. Once again, you can check it out over at Cohesive. So. And thanks again to Cohesive for sponsoring this video. You guys rock. Now, on the vein of using AI for audio, remember that song, that Heart on My Sleeve song that was like a fake Drake song that went sort of viral by this guy named Ghostwriter? Well, that guy or girl or whoever it was is back at it again with a brand new song, and this time they actually want to try to win a Grammy with it. Now, on September 6th, when I actually came across this article, it actually had the song playable right here on the website, but now it's saying, video currently unavailable. So apparently it was already taken down. I probably wouldn't have played it on this channel anyway, because I don't want to get my videos taken down. This time. The song was trying to mimic Travis Scott and 21 Savage in a song called Whiplash. And at the very end of the TikTok video, some text flashed on the screen that said that if Travis Scott and 21 Savage wanted to get in contact, he would divert all royalties made from this song over to them and wanted to figure out how to collaborate with musicians on actually moving into whatever the next wave of music is going to be. Now, I personally haven't heard yet whether Travis Scott or 21 Savage actually reached out or if if anything's ever going to come of this, because it looks like TikTok's already taken it down, but I'm sure a quick Google search will help you find it if you really wanted to hear the song Now Speaking of AI music, in a recent video I showed off Suno, which makes a tool called Chirp, and this Chirp tool was that tool where you can give it an idea for a song. It would generate the song for you with background music and everything. And I showed off a version that wasn't quite released yet where you can actually pick the genre of the song and it would make it in that genre. Suno just announced this week that Chirp version 1 is now available in Discord, which is the version that I showed off in that last video a couple weeks ago whenever I posted it. So if I jump into Discord real quick, if I go into the Suno Discord channel here and I go into one of these Chirp beta rooms, type slash to get this little pop up box and then select Slash Chirp, hit Enter and it should give me a little dialog. It gives me the option to describe the style of music, enter my own lyrics, or have Chat GPT make lyrics about so let's do bluegrass and then have ChatGPT GPT write me a song about why Future Tools is the best place to find out about new AI tools. It already generated the lyrics here for me and after 69.62 seconds I've got two versions back.

I was searching far and wide for AI tools that would guide two future tools I came where opportunities reside with their expertise and insight My worries were set aside the best place to find new AI tools I cannot hide Future tools All future tools and knowledge they impart they're the pioneers leading the way.

In the say I are cut off I was really digging that song. Here's the other version it did I.

Was searching far and wide for AI tools of A guide to Future tools I came where opportunities reside with their expertise and insight My worries were set aside the best place to find new AI tools I cannot hide Future tools oh, future tools the knowledge they impart they're the pioneers leading the way in this AI art Discovering innovations right from.

The very start they're actually really good. I wish there was a way to extend them but this is now available for anybody to play with and as of right now it's actually free to use. But if we take a peek over at their announcement they do have a free and a Pro plan. All users are automatically subscribed to the free plan upon on joining Suno's Discord. The free plan includes 250 credits or 25 chirps per month, and free subscribers will still be able to work solo in their DMs. The Pro plan, on the other hand, is 10 bucks a month and it gives subscribers a thousand credits or 100 chirps per month, the ability to buy additional generations priority in the queue and more features. That will come soon, but at the very least, go check it out, generate a couple songs on the free tier and see what it can do because it's kind of a blast now let's talk about the world of AI art for a second. Not a ton of huge revelations this week in AI art, however, my good friend Ali jewels it hang out on the Mid Journey office hours and as always reported back to X with what they talked about. It sounds like sometime In September version 5:3 is going to be released with new aesthetics. They also admitted on the call that when you are using the InPaint feature, it actually generates better images when you InPaint with the full canvas. So for example, if I take an image like this, click on Vary region and then literally select the whole image like that and then generate again. It actually uses a slightly newer model of Mid Journey, which often gets you better results. So if you generate an image that you like and it's fairly close, you can do a vary region, select the whole thing and just generate another image with the exact same prompt and it might actually come out better than what the original image came out as. They claim they've got an early version of version 6 ready, but they still plan on taking a while to release it. They also claim it should be able to generate text, but still no estimated release time. They plan on releasing the website in two parts. The new website that they're launching won't have image generation turned on at first, so it's probably just going to be a redesigned midjourney homepage kind of thing with better search and downloading features. Supposedly that part two that's coming out whenever they decide to release it will have the ability to actually generate images inside of the Mid Journey web app and finally get us out of discord. They're also working on a native mobile app. It sounds like there's going to be upgrades to the NI model, and it sounds like eventually we're going to have video and 3D and they'll probably be coming out about the same time, so pretty cool. And once again thanks to Alijules for making these updates on midjourney for those of us that don't have several hours to spend in the mid journey office hours Calls Speaking of AI art news, this week a bunch of artists signed an open letter saying that generative AI is good. Actually, there's been a lot of news and letters and things going around where people are trying to sue companies and shut down AI and say AI bad. But now it looks like a group of artists are stepping up saying, we actually like AI. Let AI keep doing its thing. It's actually making art easier and cooler and giving us new things to play with in the AI world. In this snippet it says, just like previous innovations, these tools lower barriers in creating art, a career that has been traditionally limited to those with considerable financial means, abled bodies and the right social connections. Unfortunately, the diverse pioneering work of individual human artists is being misrepresented. Some say it is about merely typing in prompts or regurgitating existing works. Others deride our methods and our art as based on stealing and data theft. Many individual artists are afraid of backlash if they so much as touch these important new tools. So it's cool to see that some artists are actually taking a stand going, yeah, there are some vocal artists over here saying we hate AI. AI bad. Get rid of it. AI should go away. But that's misrepresentation of the overall art community. A big chunk of the art community actually wants to see AI art succeed and get better and do cooler things. And in other kind of sort of AI art news, Microsoft is actually testing background removal in Microsoft Paint. So you can see a demonstration here of a little flappy dude with feathers where Microsoft Paint was just flat out able to remove the background. That'll be a pretty handy feature. Right now I use Canva or the Adobe Image Background Background Remover tool, but to have it straight inside of a desktop app in Microsoft Paint, I mean, could save me a few seconds in having to remove backgrounds. A lot of the AI world has actually talked about how Apple seems to be kind of lagging behind in the AI world. Everything seems to be all about Microsoft and Google and OpenAI and Elon Musk and all of that kind of stuff. But according to reports, Apple is actually spending millions of dollars a day on training AI. Apple actually believes that Ajax, its most advanced language model, is more powerful than ChatGPT. Right now, Apple's A team has around 16 members, including several former Google engineers. This all comes from an article from the information. And they claim that a visual intelligence unit is developing an image generation model and another group is researching multimodal AI which can recognize and produce images or video as well as text. Now, these are all just reports. This isn't anything specifically announced by Apple, but we definitely shouldn't count Apple out of the AI race. They are definitely doing a ton of crazy AI stuff behind the scenes. So this is pretty interesting. It's actually called Amazon one and it's a tool where you it actually scans your hand and it can use it for payments and loyalty cards and entry access and various ways to identify you by your palm. And they actually use generative AI to train this. In fact, they created a whole bunch of synthetic hands using AI and trained it on these synthetic hands to help better recognize real hands. And supposedly it's more accurate than scanning your retinas. The idea being that in the future, if you want to pay with credit cards, you need to show identification. You want to access entry to a concert or something like that. You no longer need to scan your phone or a ticket. You just wave your palm over a sensor and it knows that you're you and you're supposed to be there. This week an article came out in Time magazine all about Elon Musk's struggle for the future of AI. Now, it's quite a hefty read, but basically it sort of breaks down Elon's history with AI. It talks about how he's been deeply concerned about the risks of uncontrolled AI and how he actually had arguments with Larry Page, who called him a species for wanting to value lives of humans over potential consciousness of computers. In 2015, Elon Musk Co founded OpenAI in order to go head to head with Google's dominance in the AI world. Basically, Walter Isaacson, the guy who wrote the Steve Jobs book, is also now writing a book about Elon Musk. And this article breaks down some of the stories that Elon has about his relationship with AI and how Elon is feeling this increasing urge to try to make AI as safe as possible while at the same time get humans on Mars in case things go haywire. Also this week, an article was released at Fast Company about why Yann Lecun isn't buying the AI doomer narrative. If you're not familiar with Yann Lecun, he is one of the godfathers, one of the early OGs of AI. In this article he talks about how releasing these AI models to the public to get their hands on in sort of open source models and things like that will actually help improve these models faster. It was actually Lecun who urged Meta to start open sourcing a lot of their AI models. John LeCun compares AI to the invention of the car. You can have a car that rides three miles an hour and crashes often, which is what we currently have describing the latest generation of large language models. Or you can have a car that goes faster, so it's scarier, but it's got brakes and seat belts and airbags and emergency braking systems that detects obstacles. So in many ways it's safer. So basically he believes it's completely overblown. Large language models are prone to hallucinations and have no concept of how the world works, no capacity to plan and no ability to complete basic tasks that a 10 year old could learn in a matter of minutes. They have come nowhere close to achieving human or even animal intelligence, he argues, and there's little evidence at this point that they will. And in even another hot take that came out well at the end of last week, Mustafa Suleyman, the co founder of Inflection AI and one of the original founders of DeepMind, had a lot to say about where he believes AI is going. Suleyman believes that AI is about to completely reshape the world. This article is article literally starts off with the sentence we are about to see the greatest redistribution of power in history. The article argues that a lot of the white collar jobs are going to be eliminated by AI. And he's talking about things like doctors, lawyers, accountants, business VPs. We're getting to a time where technology is able to do more and more things for humans to a point where humans being the workers is essentially going to be the old way. These AIs will organize a retirement party and manage your diet. They will develop and execute business strategies whilst designing new drugs to fight cancer. They will plan and run hospitals or invasions just as much as they will answer your email. Building an airline or instead grounding the entire fleet. Each becomes more achievable. Whether it's commercial, religious, cultural or military, democratic or authoritarian. Every possible motivation you can think of can be dramatically enhanced by having cheaper power at your fingertips. These tools will be available to everyone. Billionaires and street hustlers, kids in India and pensioners in Beverly Hills. A proliferation of not just technology, but capability itself. The argument here is essentially that the role of humans, the role of businesses, all needs to sort of be rethought because inevitably, if computers and technology can do it better, faster, cheaper, safer, all of that kind of stuff for us, that's likely the route that humans are going to take. And as we take that route and get further and further down that route, the way we see ourselves, humans as the workers, the ones who do the things, needs to change. If you want to dive deeper into this article and its implications, there's actually a great video here by David Shapiro. I'll make sure it's linked up in the description. It's about 10 minutes and he really, really breaks down the implications and deep dives on this specific article. I highly recommend watching it. It's definitely worth the time. But that's three completely different takes on where AI is headed right now. Elon Musk, who's scared and wants to fix things and wants to get us off this planet if possible. Yann Lecun, who believes all of this is overblown and everybody's just kind of worrying too much about where AI is going. And Mustafa Sulaiman, who believes that this is about to dramatically change the world. Not necessarily in the negative, positive sense, in the sense that we just need to rethink everything. We need to rethink how we see labor and work and what the humans roles on this planet are. So three completely different thought processes around AI and hard to say which one I sort of lead to. I definitely think there's probably bits and pieces of all three of them that I could relate to. So I just love going down these rabbit holes and hearing what others think about it and getting my own thoughts spinning on where this is all headed in my mind. It's just fascinating. The smartest people that are out there actually building this stuff, the Elon Musk, the Yann Lecuns, the Mustafa Suleimans, these people are all on the inside. They're seeing this stuff being built. They know better than anybody else where this is heading. And so to get three articles within one week of all of their thought processes of where this is right now, it's just fascinating to me to sort of put them all together and have that sort of holistic view of where AI could potentially be heading. And finally, to wrap this video up, Time released their top 100 most influential people in AI this week. They broke it down into four categories. They also picked tops of each category such as Dario and Daniello Amodo. I'm not sure if I pronounced that or not, but they're the presidents of Anthropic who we talked about earlier in this video, of course. Sam Altman, Demis Hassabis, Elon Musk, Jensen Huang and Mustafa Sulaiman, who we also just mentioned. There's the innovators led by Ted Chang, shapers led by Alondra Nelson, but also including Eric Schmidt from Google, Elazier Yudkowski, someone who literally believes that at the pace AI is going, we have about a year left on this planet. And finally, the thinkers led by Geoffrey Hinton, the godfather of AI, also included Jan Lecun, who we talked about earlier this video, Ilya Suskiver, the co founder of OpenAI and many others. There are some omissions to this list that I found interesting. It's interesting to me that Ahmad Mustach from Stability AI didn't make the list. I also kind of find it a bummer that creators who are also sort of pointing people in the direction of a lot of the innovations didn't make the list either. But I can't complain about that too much. But there definitely are people out there that are really helping push the AI narrative forward. People like Robert Scoble and Rowan Chung and Ben Tassel. And maybe I'd toss myself in there as well. We're trying to do our best to keep the public informed about where AI is going, but again, this is just the top 100. Maybe we'd fall into the top 200. Who knows? Anyway, that's all I got for you today. I really appreciate you. If you haven't Already, check out FutureTools IE where I curate all the coolest AI tools that I come across. I keep the AI news up to date on a daily basis. I'm adding new tools every single day. Added 12 tools in the last 24 hours. That's all a little too much for you. Make sure you join the free newsletter and once or twice a week I'll send you an update with with all the latest news that you need to know. Just the coolest tools I've come across and a few YouTube videos to make sure you keep coming back and watching more. And once again, thank you so much to cohesive for sponsoring this video. I really appreciate it. You make it possible for me to keep on making videos like this for a living. So I really, really appreciate you. And thank you so much for watching. If you like this kind of video, maybe give it a thumbs up and subscribe to the channel. I'd really appreciate it. All right, thanks again. See you in the next one. Bye. By. Sam.