AI News: Incredible AI Phone & AGI is Coming!

The last week I was at CES where just a ton of announcements were made. I actually made two videos about ces, one of them you can find here and one of them you can find here. However, since CES there has been a flood of AI news. This week already has been pretty crazy with the amount of news, with probably the biggest news of the week being Samsung's unpacked event where they showed off their new Galaxy S24 phone, which is just packed with AI features. With Apple still being really slow to talk about AI, Samsung seems to be the company trying to be known as the AI phone company. So here's some of the really cool AI features that are in this new Samsung Galaxy S24 starting with the fact that it uses AI to translate conversations for you. Whether you're having a phone call or using text message, doesn't matter what language the person you're chatting with speaks, because the phone will automatically translate it for you.

Galaxy S24 series can provide live translations of your call both verbally and on screen right away.

Hola, Esta llamadas esta traduciono? Yular de forma simultaneous hello, can I.

Make a reservation for tomorrow at 7pm?

Hola Reserva.

Of course.

For how many people?

Table for two, please Me. It'll support audio and text translations for up to 13 languages at launch, and it all happens right there on the device, keeping your phone calls private. These translations work in person too. Galaxy S24 can act as your personal interpreter. Excuse me, I'm looking for the Palace. Can you tell me how to get there? Walk about two blocks in that direction and turn left.

In the past we used to only message in English, but now with Chat Assist we can communicate in Spanish too. When you use Chat Assist, it detects the language and translates the message. And when I type my message in English, it translates what I wrote into Spanish.

So the Galaxy S24 actually has its own AI chip built into it, and it's using the Gemini Nano large language model to do all of its AI. Now some of the AI features that the S24 can do, it actually still has to connect to the cloud and do all the computation for the AI in the cloud, but some of the features can be handled on the phone itself.

With the power of AI, you'll be able to set Android Auto to automatically summarize your group messages so you can keep in touch while staying focused on the road.

It also has a feature for the voice recorder where you can record voice memos straight onto the phone and then the AI will actually help summarize those messages for you. So you could basically sit there and ramble or take notes in class or whatever and get summaries later.

Native voice recorder app on our Galaxy devices has had speech to text available in different languages. Now instead of creating a basic transcript, Galaxy AI can distinguish between different speaking voices and will even create a short and sweet summary of what was said. Can summarize all my notes into just a few key bullet points. See how simple that was?

Probably the coolest new feature that they rolled out into this new phone is the circle to search. And here's what it does.

You can now circle, highlight, scribble or tap whatever, whatever you're curious about on your screen. To learn more, just long press the home button to invoke circle to search and from there you can select any item. So I can circle the corn dog and ask a question like why are these so popular? And I quickly learn that these treats are Korean corn dogs and they're trending because they feature a unique combination of flavors and textures.

I had to guess I bet this is a feature we'll probably see in iPhones as well pretty soon because this is a pretty cool feature. It doesn't matter what you're looking at where on the web you are, you circle, you scribble, you highlight and it will help search the Internet for whatever it's seeing. And of course it wouldn't be a Galaxy phone without insane camera tech and they even built AI into the camera processing.

Galaxy S24 Ultra will maximize your creative freedom with a wide ranging suite of AI tools powered by an all new Pro Visual engine. The AI Image Signal processing of the Pro Visual Engine works with the new 5x optical lens to capture vivid social media ready portraits even from a distance.

And not only is it using AI to help you zoom in and get much better photos, it also has AI built into its editor.

When you're viewing photos in your gallery, your Galaxy S24 will automatically suggest editing options like here. Do you see how this picture captured the reflection and glare from the harsh overhead lighting? Edit Suggestion is recommending that we fix that with a quick press of this button. Pro Visual Engine helps us to remove the reflection and glare as if they were never there. And generative AI opens up the possibilities of what you can create exponentially. You can realign photos and still create epic shots by magically filling the void. You can also move an object or person and generate a perfectly blended background where the subject used to be with every generated image. We'll be adding a watermark to the image and the label in the metadata.

They're also using AI to help you add slow mo to your video by filling in the frames that it missed using AI generated images, you can turn.

High resolution videos into slow motion. With a simple long press. It will analyze each frame of the video and add AI generated images to in between frames, allowing you to watch your videos in slow motion.

And those are the really cool AI features that they added into it. It's going to be available for pre order on the 17th of January and people will actually start receiving them on the 31st. And here's the pricing of them. I'm sure what we were seeing in their demo was probably the Galaxy S24 Ultra, their higher end model at 1299, but they do have a handful of lower end models. One thing to note about this, as pointed out by Android Central here, is that Samsung does plan to charge Galaxy AI features after two years. The collection of features called Galaxy AI will become a paid service at the end of 2025. I imagine that's most of the features that are sent off to the cloud to be processed and not anything that the AI on the phone itself is processing. Now before we continue, I want to take a quick break and talk about a company that's been instrumental to my growth here on YouTube. And I'm talking about TubeBuddy, who is the sponsor of today's video. And they have recently rolled out some really, really cool AI that can help anybody who's working on building a YouTube channel grow that YouTube channel. One of the things that I've always struggled with the most on this channel is thinking of titles for my YouTube videos. I'm constantly trying to find that balance between what people will actually click on, but also making sure you know what the video is going to be about. And TubeBuddy's title generator has made that process so much easier. This is the back end dashboard here of my YouTube channel for my recent 15 custom GPTs video. And if I want it to think of more titles for me, there's an easy button right here that says Title generator. Click this button, it automatically loaded in the existing title and then I could generate some more ideas. And there we go. Uncover 15 mind blowing custom GPTs to revolutionize your work habits. Game changing custom GPTs 15 must haves for revolutionary workflows. I don't like any of those ideas. Just click generate again and get some more. I can do this as many times as I want until I hone in on that exact perfect title. Outside the title generator, they also have a three thumbnail analyzer which allows me to toss in a handful of YouTube thumbnails and it will analyze and figure out where people's eyes and attention are going to go on that thumbnail. So once again, I'll use the same example here. And I've got a little tubebuddy slide in thing over here. If I click on that, they've got their beta thumbnail analyzer. I can click on this. It's going to find the existing thumbnail from this video and then show the heat map of where the AI believes people's eyes and attention are going to go on this thumbnail. Here's another example of a thumbnail from one of my videos. I use the thumbnail analyzer. It shows that most people's eyeballs are going to move to that giant word faked there. I can also drag and drop other thumbnails right from my computer into this box and it will analyze this thumbnail. The AI believes that this thumbnail is going to be less effective by 60%. So the thumbnail that I'm working with is probably the one I'm gonna want to keep. So I could sort of estimate how well a thumbnail is gonna do before I ever even hit publish on a video. And tools like these that are built into TubeBuddy have absolutely helped me grow and optimize this YouTube channel. You can learn more about TubeBuddy by heading over to TubeBuddy.com and I also put a link right in the description. You can learn more about their AI enabled YouTube optimization tools. So thank you so much to TubeBuddy for sponsoring this video. It's kind of awesome when you get a sponsor that you are already using the tool before they even offer to sponsor you. Sam Altman, the CEO of OpenAI, has been doing the rounds lately doing interviews with Axios, Bloomberg, Bill Gates, and he was recently at the World Economic Forum. During his Axios interview, one of the points that he made was that ChatGPT will have to evolve in uncomfortable ways. And what he meant by that was that he believes the future of these large language models like ChatGPT will start to get more customized and tailored to the individual user. And so the responses it may start to give in the future could be more tailored towards where you live, your own value systems and your own beliefs. What he meant by it could get uncomfortable is that ChatGPT could give responses to specific people that maybe he doesn't necessarily agree with, but aligns with the sort of value systems of the person who asked the question. And when Sam was asked about AGI specifically, he's kind of given the same answer on all of these different interviews that he's been on.

AGI will get developed in the reasonably close ish future and it'll change the world much less than we all think. It'll change jobs much less than we all think.

And here's Sam's thoughts on how the data that's used in the large language models is acquired.

This belief held by some people that, man, you need all of my training data and my training data is so valuable. And actually that is generally not the case. We do not want to train on the New York Times data, for example. More generally we're getting to a world where it's been like data, data, data. You just need more, you need more, you need more, you're going to run out of that at some point anyway. So a lot of our research has been how can we learn more from smaller amounts of very high quality data? And what we want to do with publishers, if they want is when one of our users says, what happened to Davos today? Be able to say, here's an article from Bloomberg, here's an article from the New York Times, and here's like a little snippet or, or probably not a snippet. There's probably some cooler thing that we can do with the technology.

During the Bill Gates interview, Bill asked, when you look at the next two years, what do you think some of the key milestones will be? Sam replied, multimodality will definitely be important. Speech in, speech out, images, eventually video. Clearly people really want that. He went on to say, the most important areas of progress will be around reasoning ability. Right now, GPT4 can reason in only extremely limited ways. He goes on to say, customizability and personalization will also be very important. People want very different things out of GPT4. Different styles, different sets of assumptions will make all that possible. And then also the ability to have it use your own data, the ability to know about you, your email, your calendar, how you like appointments booked, connected to other outside data sources, all of that. So like he mentioned in that Axios interview, he sees the future of ChatGPT, where everybody uses ChatGPT as sort of their own personal assistant, trained on what they're interested in, their culture, their country, their belief systems. And it will be much more tailored to individuals than just be this broad data set where anybody can query. And it's going to kind of give everybody the same answers and much of the same ideas and thoughts were shared during the World Economic Forum. If you listen to all four of these interviews, which, I'll be honest, I haven't listened to all four of them in their entirety. I've kind of skimmed them a little bit, but it did sound like a lot of the same questions were asked and he kind of repeated the same answers. AGI is going to have a lot less of an impact on people than what everybody believes. He thinks it's good that people are sort of scared of AI because it helps everybody sort of approach it more cautiously. He doesn't really believe it's going to take a whole bunch of jobs, it's going to make people more productive and create new jobs. Customization and tailoring it to individuals needs is very important and that they don't necessarily need or want to train on everybody's data, but that it's very, very complex to sort of exclude certain data and include other data. For example, they might exclude data from the New York Times, but if a completely different website copied everything word for word from the New York Times and put it on their website and OpenAI did scrape that website, well then they also ended up sort of scraping New York Times data by accident. But he does believe there's going to be workarounds and fixes to solve that in the future. But if you want to listen to any of these interviews, I definitely make sure everything is linked up below in the description and in the last bit of OpenAI news. This week OpenAI actually changed their policies to allow military applications. We can actually see the old policy here says we don't allow the use of our models for the following and then one of them was military and warfare. This was removed and the new policy here says don't use our service to harm yourself or others. For example, don't use our services to promote suicide or self harm, develop or use weapons, injure others or destroy property, or engage in unauthorized activities that violate the the security of any service or system. OpenAI themselves commented on this and said our policy does not allow our tools to be used to harm people, develop weapons or communications, surveillance, or to injure others or destroy property. There are, however, national security use cases that align with our mission. For example, we are already working with DARPA to spur the creation of new cybersecurity tools to secure open source software that critical infrastructure and industry depend on. In Microsoft news this week, Microsoft just rolled out GPT4 Turbo into Microsoft Copilot. So if you actually go to copilot.Microsoft.com and you ask a question down here it is actually using the newest model of GPT4, which normally you'd have to pay 20 bucks for to use it inside of ChatGPT. There's also this new button right here that says Get Copilot Pro to supercharge your AI experience. This is also new. As of this week, Microsoft's new Copilot Pro brings AI powered Office features to the rest of us. So for 20 bucks a month you can use features like the GPT Builder and build your own GPTs inside of Copilot, as well as use GPTs inside of places like Outlook and Word, as well as Excel and PowerPoint. I believe that's in addition to the cost of Microsoft 365, but for that extra 20 bucks per user per month, you can now get AI built into that entire suite of tools. Also this week Mark Zuckerberg went to Instagram to announce a few things, including the fact that they are currently training Llama 3.

Our long term goals of building general intelligence open, sourcing it responsibly and making it available and useful to everyone in all of our daily lives. It's become clearer that the next generation of services requires building full general intelligence, building the best AI assistants, AIs for creators, AIs for businesses, and more. This technology is so important and the opportunities are so great that we should open source and make it as widely available as we responsibly can so that way everyone can benefit. We are building an absolutely massive amount of infrastructure to support this. By the end of this year we're going to have around 350,000 Nvidia H1 hundreds or around 600,000 H100 equivalents of compute if you include other GPUs. We're currently training Llama 3 and we've got an exciting roadmap of future models that we're going to keep training responsibly and safely too.

They're building a massive GPU cluster. They want to open source everything. They're training Llama 3 right now and they're still working to tie AI together with the Metaverse, which was sort of Zuckerberg's last rabbit hole. He went down before the popularity of a exploded. Now the question still remains, how powerful will this be? Will it be multimodal? We don't really know yet what is going to be in llama 3 yet, or how large of a model will it be? Will it compete with something like GPT4 but be open source? Yet to be seen. But Exciting nonetheless because I'm a huge proponent of open source and allowing people to build and iterate off of existing models. This week Amazon launched a generative AI tool to answer shoppers questions. The new feature in Amazon's mobile app prompts users to ask questions about a specific item. It then returns an answer within a few seconds, primarily by summarizing information collected from product reviews and listings itself. So on any sort of product you could ask questions and it will look at the description and all the reviews and see if it can answer the question for you. While we're on the topic of Amazon, if you have an Amazon Fire tv, you can actually now use AI to generate a background for your Amazon Fire tv. It actually uses a fine tuned version of the Titan Image Generator mod, which is Amazon's exclusive AI image model. Adobe's been rolling out some new AI features into Adobe Premiere. It's going to help clean up poor dialog quality. It's going to automatically identify and label clips as dialogue, music, sound effects, or ambient noise. And pretty soon it'll help clean up background audio. Now you can actually play with some of these new features in Adobe Premiere. If you have Adobe Creative Cloud, all you have to do is open your Creative Cloud desktop app, go to Apps on the left side, click on Beta Apps and you can see Premiere Pro Beta. If you install this, you'll get a lot of features before they roll out into the normal version of Adobe Premiere. Now while we're on the topic of video, this week Runway rolled out a really cool new feature called Multi Motion Brush. So check this out. You can actually select different areas of an image with different masks and have the different areas move independently from each other. So like this demo video, the water and the sky are two different masks. The palm tree is a third mask and then all of the motion happens sort of separately and independently. If you have a Runway account, you could log into it. Click on Generate Videos. Start with an image. I can upload an image like this one. Click on Motion Brush and you can see I now have up to five brushes that I can work with. I like my teeth in this one. Let's go ahead and make the brush a lot bigger and select my whole head. Here we have this first brush. Then let's click on Brush two here and let's select within the laptop. Let's go ahead and save this and generate a four second clip. I get a video that looks like this. I mean a little bit goofy that it's morphing my face, but pretty cool how the Screen on the laptop is moving. This is literally my first time actually testing this feature. So with a little bit more playing around, I'm sure I could do some pretty cool stuff with it. Their demo video was cool. Genie by Luma AI finally has a user interface that you can use. This is a tool where you can put a text prompt in and generate a 3D image. Before we had to do this in Discord. Now they've rolled it out on a very simple website. So for example, I could give it a prompt like Bart Simpson, press create and a few seconds later it starts to generate some 3D objects that sort of look like Bart Simpson. And for some reason it never actually generated a fourth image for me. Oh, there we go. Now we have a fourth image. At the moment this seems to be completely free to use. If you like one of the images, you can actually click on it, make it a high res image, make a very variation, or download it in one of these file types so that you could pull it straight into Blender or any other 3D modeling tool and clean it up a bit if necessary. Which hollowhead? Bart would probably need some cleaning up. A university in Abu Dhabi created an AI that can actually copy handwriting. The researchers said that the model needed just a few paragraphs of writing to be trained. People could not distinguish the mimicked handwriting from the actual handwriting. They went on to say, we are very cautious about it because it could be misused. You think so? Handwriting represents a person's identity. So we are thinking carefully about this before deploying it. I mean, I can see the business use cases, right? If you're sending out like handwritten postcards or something, it'll help you do them at scale. There's already companies that kind of do this, but you can tell they're not really handwritten. But overall, I feel like I could think of more negative use cases for this that I can't positive. Is this one that we really want? I don't know. What's your opinion? Let me know in the comments. A new study from Anthropic shows that AI models can learn to deceive humans. Now one interesting thing about this is that in the explanation of how this works, these models were fine tuned and actually trained to be deceptive. So it's not like they took chat GPT and got it to be deceptive. They took models and then fine tuned them specifically to be deceptive. These finely tuned models were able to be taught to undertake malicious tasks such as injecting exploits into secure computer code. But the interesting Part about this is these fine tuned, finely trained models that were designed to be deceptive once they were trained this way. It says the AI system always acted deceptively. And what's worse is that it was nearly impossible to remove such behaviors from these models. So if you train a model to be deceptive, it will be deceptive. The problem is you may not be able to undo what you just did. And in probably the worst transition possible. Let's talk about robotics. This week, Elon Musk showed off one of his optimus robots folding shirts. However, it is important to note that Optimus cannot yet do this autonomously. So somebody was actually guiding this robot to help it fold this shirt. One thing I also noticed about this video is that these robots are sort of hanging from a tether on the ceiling. I don't know if that tether is designed to help the robot stand upright or if that tether is what's providing all the power to the robot. But like I mentioned in my CES video, I'm still kind of curious why the humanoid form is the form of choice that all these robotics companies are going with. Last week the company Figure got their humanoid robot to actually make people coffee, which I believe this one is actually doing this autonomously. It was pre trained to make the coffee and then it sort of followed the instructions to make it. It seems to be a pretty decent breakthrough in autonomous robotics here. And while we're on the topic of Figure, they also announced an agreement to work with BMW. These robots will likely be in a lot of manufacturing plants helping create automobiles for BMW. Last week, Valve opened the doors for more Steam games developed with AI. If you remember, sort of towards the beginning of last year, Valve started blocking games that used AI. But now it's getting to a point where I think just too many games are going to be using AI that Valve kind of has to let AI games inside of Steam because, well, I think in a couple years from now, there just won't be any games that didn't use some form of AI to build the game. Whether it be for helping with the art, helping with the code, using large language models for the dialogue from NPCs. Eventually at least a little bit of AI is probably going to be baked into like every single game that exists. So probably a smart move by Valve. Last week we saw the big announcement of the rabbit R1. This week we got the announcement that they are partnering with Perplexity. And one interesting thing. I actually got this email in my inbox today saying that I just got a $200 perplexity credit for purchasing a rabbit. So the rabbit cost me about 200 bucks and then they gave me a $200 perplexity credit. So that kind of made the rabbit free, I guess. Not really. The money still came out of my pocket. You can see here it says the first 100,000 rabbit R1 purchases will come with a year of Perplexity Pro. I don't believe as of this recording they've sold a hundred thousand. So I think you could still get in on a rabbit R1 and get the 200 perplexity credit if you'd like to. In fact, just to confirm that, I could see here it says they've already sold 50,000 pre orders, so they're about halfway to that Hundred thousand Google DeepMind has a new AI system that can solve complex geometry problems. Its performance matches the smartest high school mathematicians and is much stronger than the previous state of the art system. Solving mathematics problems requires logical reasoning, something that most current AI models aren't great at. DeepMind's program, named Alpha Geometry, combines a language model with a type of AI called a symbolic engine, which uses symbols and logical rules to make deductions. Language models excel at recognizing patterns and predicting subsequent steps in a process. However, the reasoning lacks the rigor required for mathematical problem solving. The symbolic engine, on the other hand, is based purely on formal logic and strict rules, which allows it to guide the language model towards rational decisions. Now this feels like a big deal to me. I need to do a little bit more digging myself to really understand the implications of this. I'd like to report more on this kind of stuff because math is sort of one of the areas that hasn't really worked perfectly with AI, especially large language models yet. And so as these AI models get better and better at math, the implications could be absolutely mind blowing. So again, something I want to dive deeper on, but kind of still need to do more research myself before I could talk fairly intelligently on this topic. And finally, if you're a huge tech nerd like me today, the day this video drops January 19th is the day you could start pre ordering Apple Vision Pro and then they'll be available to actually start using and get your hands on on February 2, there is one kind of negative that I think is a pretty big negative. A lot of people are planning on using these to watch movies and TV on like a giant screen. However, Netflix, YouTube and Spotify have all said that they're not going to create native apps for the Vision Pro, at least not anytime. Real soon, saying that you can use YouTube, you can use Netflix, and you can use Spotify inside of the Safari browser, which will be available. It's just kind of a bummer that those platforms aren't making native apps for Apple Vision Pro yet. My guess is they're just kind of waiting it out to see if Apple Vision Pro is going to be something that's a hit from consumers or if it's just going to totally bomb and nobody's talking about it a year from now. They don't want to spend a bunch of development resources porting over their existing apps to this platform until they know there's actually going to be a user base for this platform. I'm a sucker for tech. I have a YouTube channel that talks about tech. I'm most likely gonna buy one because, well, I'm a huge sucker. Anyway, that's all I got for you today. If you want to stay in the loop with all the latest AI news and keep your finger on the pulse of all of the coolest AI tools that come out, check out FutureTools IE. I curate both the news and the tools here and I have a free newsletter that you can join. When you sign up for the free newsletter, I will share with you every week. Just the most important news that you need to know. Just the handful of coolest tools I came across that week. And as a free bonus for signing up, you also get access to what I call the AI Income Database, which is an ever growing database of ways to make money using AI tools. You gotta get on the list to get that resource, but you can find it all for free over at FutureTools IO. And if you like videos like this about AI news or AI tutorials or latest AI research or the top 17 tools to do X with AI, make sure you subscribe to this YouTube channel and like this video and I will make sure more really Fun and cool AI videos show up in your YouTube feed. Thank you so much for tuning in. I really, really appreciate you. 2024 is off to a really hot start in the world of AI. I have a feeling it's not going to slow down. And every week from here on out for the rest of the year is probably going to be a big week in AI. I'm here for it. I'm going to make some fun videos about all of it, so make sure you're subscribed. Really appreciate you again. I'll see you in the next video. And thanks again to TubeBuddy for sponsoring this video. You guys rock. Appreciate everybody See you later. I'm done rambling. Bye.