Latest / Bad Decisions Podcast / Google's New AI VIDEO MODEL & #1 TIP on How to get the BEST RESULTS from Claude & CHATGPT!
Transcript
- 0:00There are rumors about a new AI video generation model by Google
- 0:04that is set to drop this week. You can now buy a mecha suit for
- 0:10half $1,000,000. And there's an unreasonably
- 0:14effective way to work with LLMS with ChatGPT and Claude that
- 0:18most people don't know about. So we're going to share that
- 0:21with you and hopefully your workflow is going to be 10 times
- 0:23better. And there is also a new AI
- 0:26interaction model that can think, listen and talk in real
- 0:31time. That's really important because
- 0:33right now when you're talking to ChatGPT, it gets interrupted
- 0:36every time you say something. But when you talk to a real
- 0:38person. It doesn't happen.
- 0:40Yeah, Farah talks over me. I overlap on top of him.
- 0:43It's very real. That is what this is about.
- 0:46But before we begin. Hello everybody, it is Tuesday
- 0:51and we are here with another episode that's.
- 0:54Right. Every Tuesday, every Friday, we
- 0:56bring you the latest and the greatest in the world of
- 0:58technology. And we just wanted to say
- 1:01welcome back. Thank you for being here.
- 1:03We've got 4 hot topics for today and we're going to get started
- 1:08right away. Let's go.
- 1:10Are you ready? Yes.
- 1:12So first up, we're going to be talking about the new Google
- 1:17model. There's a leaked video circling
- 1:21the Internet about this new AI video generation model, which is
- 1:25called Google Omni. And the reason why people are
- 1:31talking about this, why people are excited about this, is this
- 1:36exact thing that is happening right now.
- 1:38We see a professor writing out a mathematical function on a
- 1:42board. What's fascinating is the text
- 1:47rendering, which is incredibly accurate.
- 1:51This is something that has been a.
- 1:54Huge problem. Complex to achieve both in
- 1:57images and even more so in videos.
- 1:59Can we have consistent text rendering, which is crucial for
- 2:04things like making product videos or marketing material or
- 2:08training content, because the moment the camera shifts or
- 2:12moves, all the text begins morphing.
- 2:15And secondly, apparently the mathematical functions and
- 2:19what's being explained by the professor are actually accurate
- 2:23contextual to what he's talking about.
- 2:25Just like how Nada Banana 2 with the image generation model.
- 2:29You could ask it for the current weather in Paris and have it
- 2:33create an infographics for you or weather update.
- 2:36It would go ahead and pull real time data from Google and create
- 2:40that image. So it's not just lies.
- 2:42Apparently the same thing is true here.
- 2:44I mean, I'm no math expert. You guys can go ahead and
- 2:47confirm this right now in the comments.
- 2:49Farrah, how good are you at mathematics?
- 2:51Pretty good I think. It's not good enough to verify,
- 2:53no. No, I can.
- 2:54I think it's a tangent equation. Wait, let me say, what is a
- 2:57tangent equation if I'm not? Mistaken.
- 2:59Let's listen to the. Video.
- 3:00It's sine square. We can listen to the video.
- 3:02Let me see what is. Now if we divide every term, we
- 3:06start with the fundamental identity sine squared plus
- 3:09cosine squared equals one. Now if we divide every term by
- 3:12cosine squared, we arrive at the identity.
- 3:15For tangent we start with the fundamental identity.
- 3:18Sine squared plus cosine squared equals one.
- 3:21Now if we divide. Every at least whatever he's
- 3:23talking about, he's writing it on the board whether it's
- 3:26factual or not. One my question is why they
- 3:29didn't call it VO4? Why Google Omni?
- 3:32I believe this. Generation model is not Google
- 3:35Omni, the name people saw on Google's website in the Gemini
- 3:40Apprat. Do you have the leaked image
- 3:42that people shared? It says create with Gemini.
- 3:46Omni or Gemini Omni Omni Omni image video.
- 3:50Means everywhere and everything and I'm imagining omnipresent.
- 3:54I'm talking about omnipresent, but Gemini Omni is perhaps
- 3:58everything in one place. Oh, so image, photo, video.
- 4:01OK. Video generation model.
- 4:04Now you have to understand why this is important for us, aside
- 4:08from the superior text rendering and that being incredibly
- 4:11important for genuinely, let me rephrase that.
- 4:16A lot of the content that you see on social media that is
- 4:20hyping up all of these AI tools, a lot of them are made for that
- 4:2615 seconds hook just to get your attention.
- 4:30But when you actually go ahead and use these tools for real
- 4:33life scenarios, you'll realize quickly that not all of them
- 4:39work for production ready content.
- 4:42For instance, you may see crazy clips like the the latest one
- 4:46that went viral. Pratt, can you pull up the
- 4:49tsunami video that went viral? There's a tsunami video made
- 4:53with. Like there's a couple, there's a
- 4:55couple fighting and then there is a tsunami AI video.
- 4:58Let me. Pull that up really quickly, I'm
- 5:02sure it's there. Not that one.
- 5:04The second one I think someone remade that.
- 5:06No, that's not the original 1. I know, I know what you're
- 5:08talking about. It's like the couple.
- 5:10Higgs field even made their own by the way, I just.
- 5:13I mean, yeah, we can just. Pull it.
- 5:14Just play that and this is not it.
- 5:16But there's a video like this that went incredibly viral on
- 5:19Instagram. Or maybe it's the one at the
- 5:21top, the second one? Yeah, it's not, it's not
- 5:24actually this one. Something like this?
- 5:26Someone made a video like this? Yeah.
- 5:29It looked a lot more realistic, by the way, that the one I'm
- 5:31talking about and it has millions of views and likes and
- 5:35comments on Instagram. You see these videos, you'll see
- 5:39and say, Oh my God, these AI tools are great.
- 5:41I'm going to use them for my next marketing campaign.
- 5:44It's 100% ready for production. The truth is in your marketing
- 5:48campaign, you're probably going to be using a product with
- 5:50labels on it with information that needs to stay consistent in
- 5:54every frame of the video. Here you're looking at a tsunami
- 5:58water that is moving in a fast-paced with the shutter
- 6:02speed involved and the motion blur.
- 6:04This is. The video.
- 6:07EE as long as the water simulation looks retty decent,
- 6:10you're good. You're not selling anything
- 6:11here. People don't really care about
- 6:13the minute details, they just look at the overall picture and
- 6:16they're amazed it looks great. Then you go ahead and use the
- 6:20same thing for your product. It doesn't work, it breaks.
- 6:22That is why we talked about superior text rendering, because
- 6:26if the text and information can remain consistent in every frame
- 6:30of the video, that really pushes the limits of what you can do
- 6:35with it in terms of real use cases.
- 6:37Now we have to go ahead and compare what we just saw with
- 6:40the best AI video generation model that is out there today,
- 6:44Cdance 2.0. Cdance 2.0 is available on
- 6:47pretty much every platform right now.
- 6:49However, the biggest problem with that is its price.
- 6:53It's incredibly expensive and hard to use because of the IP
- 6:58infringement and copyright issues that they're facing.
- 7:02You cannot always use the faces that you want to use.
- 7:05What I'm assuming will happen with Gemini Omni.
- 7:09My hope is that the quality is going to substantially increase,
- 7:13become a little closer to what we have at Cdansu Point O and
- 7:17become competitive with that. And cheaper.
- 7:20And potentially cheaper, more accessible for the average
- 7:25person to go ahead and use. It but Google VO3 had its own
- 7:28limitation as well right? We couldn't prompt Batman.
- 7:31If I remember correctly. You couldn't use.
- 7:33You couldn't use. There are.
- 7:34Workarounds always. Yeah, yes, a man with a dark
- 7:38mask, Yeah. Yeah.
- 7:40But played the other video. It's funny that the new
- 7:43benchmark for AI video has become spaghetti.
- 7:47I think one of the videos for Google Omni was also 22 guys
- 7:51eating spaghetti. They should be under this beach
- 7:53club. Yeah.
- 7:54Go down, yes. This one I didn't like.
- 7:57This one I did, actually I did. That's what I wanted to say.
- 8:00I think C dance can do a better job.
- 8:04Not at all, especially with this to the beginning please.
- 8:07Good to see you my friend. This view never gets old, does
- 8:10it? But there is something with the
- 8:13voice still to this date. Especially with this food, How
- 8:17does funnily enough, C dance audio is really good.
- 8:19Yeah, it still feels as if it is AI.
- 8:23Generated, they perfected the visual part.
- 8:26The audio still keeps it away the.
- 8:28Visual is not perfect either. The face.
- 8:29Is much better. The face is.
- 8:31Don't look that real. The emotions of the face look a
- 8:34little too fake for me. That's my biggest problem.
- 8:38In terms of the text rendering, as I mentioned, I do see the
- 8:42heavy advantage. Cdance has problems with text
- 8:45rendering by the way, so using it for products.
- 8:47Sometimes I had issues and you guys will see in our AI
- 8:50education program later on but here when it comes to genuine
- 8:56expressions of the face, the expressiveness of it C dance 2
- 9:01point O does a great job at that with this clip.
- 9:04I'm a little disappointed, but again, these are only two videos
- 9:07that we've. Got.
- 9:08We should test it ourself and see how they how it works.
- 9:10Actually I don't think they have enough context to test all of
- 9:16this. We need at least different
- 9:18characters, different environments, different use
- 9:21cases. For example, product ads,
- 9:24dancing, video fighting, video, sports.
- 9:27Then you can compare and see how the results are.
- 9:29Well, Google IO, the annual developer conference, is
- 9:32happening this week. May 19 is going to be the first
- 9:37day, which is why I'm assuming we're seeing these rumors and
- 9:41leaked footages. Hopefully by the end of this
- 9:44week, we're going to have more information about the new AI
- 9:47video generation model. When we do, as you always know,
- 9:50we'll be covering it right here and we'll find out if it is
- 9:53better than C Dance to Pointer or not.
- 9:54Is it cheaper? Is it more accessible?
- 9:56Is it higher quality? We'll find all of that out and
- 10:00tell you guys the details. That's about it for Gemini only
- 10:03if you guys have more information.
- 10:05We never asked of you to do this, but from now on, actually
- 10:08I'm going to say it if you do have more information about
- 10:11these. Things inside the info.
- 10:13Yes, go in and let us know in the comments.
- 10:15We'll cover it in the next episode.
- 10:17OK, the words first mass produced Mecca suit is ready
- 10:22now. The same company Unitary that
- 10:25released the robot that we saw, they did the fighting scene in
- 10:29Chinese New Year. They created this Mecca that
- 10:33weighs over 500 kilogram if you sit inside of it and it's priced
- 10:38around 650 thousand U.S. dollars.
- 10:41This is straight out of the movies.
- 10:43If you have watched Edge of Tomorrow, The Matrix, Pacific,
- 10:47Grim, in all of those movies when they wanted to go to war,
- 10:51the soldiers will hop into a Mecca and then they would go and
- 10:55fight their battles. We even had it in the Avatar in
- 10:59the first one and the main villain got killed in one of
- 11:03these mecas. Yes, this was the one.
- 11:04They actually made it by the way for the movie.
- 11:07Wow. Wow, yes, we are 100% moving
- 11:13towards a sci-fi future and I love this.
- 11:17This is from the game Anthem, right?
- 11:19Yeah, this is the same guys by Titanfall or No, no, no, maybe.
- 11:24I know. I remember Anthem.
- 11:25There was Titanfall and Anthem, both of which had the Mecca
- 11:28suits, and it's one of my favorite things in video games,
- 11:33by the way, great visuals when Call of Duty had it, when Anthem
- 11:36had it, the concept of humans enhancing their capabilities
- 11:41using mecha suits has always been one of my wet dreams.
- 11:46It is the coolest thing because there's only so much you can do
- 11:49with your physical body, but the moment you put yourself in this
- 11:52exoskeleton massive suit and you can arm it with different
- 11:56things, not only are you taller, you're bigger.
- 11:59You're. Stronger, you're stronger.
- 12:02Eventually these things will be able to fly too, so you'll be
- 12:06able to fly, land in different locations.
- 12:08You get to go to hotter environment or colder
- 12:11environments and still be fine because you're protected by the.
- 12:14Oh, imagine, go to Antarctica with one of these, but let's
- 12:18watch the actual demo. One of the features that I saw
- 12:21that reminds me, no, it reminds me of Transformers because it
- 12:24can rotate back, it can fold back and you can use it like a
- 12:29like a pet, you know, like a dog.
- 12:31Did you just call the Mecca a pet?
- 12:33It can be my pet. Do you know what a pet is?
- 12:36Robbie, our dog, is a pet. You can have pets, haven't you
- 12:39see Russian have a have a beer as a pet you.
- 12:44Domesticate the. Animal and then it's a
- 12:46domesticated robot you how strong it is wow but there's no
- 12:51one inside of it this part look and now it was the.
- 12:55Whole purpose of this is for increased balance and
- 13:00flexibility and mobility in extreme terrains.
- 13:05But if you're sitting inside it, now you're.
- 13:07Yeah, well, actually, you know, I was thinking about the same
- 13:10thing, you know, what happens to your.
- 13:14But I think that the the seat should be rotatable so you can
- 13:18move. You know, By the way, they
- 13:19called this robot GD01. I think GD stands for Gundam.
- 13:23Do you think so? I think so, yes.
- 13:25They haven't put it on the website yet, but $650,000.
- 13:29Should we do a Kickstarter and buy one?
- 13:31100%. 6. 100%. And we can do the show from the
- 13:36from the robot. OK, so let's talk about this.
- 13:40There's two ways you can operate this Mecca GD01.
- 13:44You can either remote control it in situations where it's
- 13:48dangerous for a human to be sitting inside and piloting it,
- 13:53or you can sit inside and pilot it as a first person view or
- 13:58angle if you want to call it. And I think that's a more
- 14:00exciting way to. Use it why you want to remotely
- 14:03control it. If it's dangerous to go to where
- 14:06it's supposed to go to. What I love about the demo as
- 14:08well. If you look and pay close
- 14:09attention, pause the video product that looks like a fake
- 14:12city like. The one that they make movies in
- 14:14Hollywood, Yeah. They have fake cities made for
- 14:17movie sets. What do they call it?
- 14:18It looks like Unitary has made their own fake movie set just to
- 14:22demonstrate the robot. Or perhaps there's a place in
- 14:25China where they record and shoot films and they went there
- 14:30to shoot this. I'm curious if somebody knows.
- 14:32Let me know. Oh, there is no, there are there
- 14:34are fake cities in Guangzhou, China location in New Mexico,
- 14:38US, Jordan and in the UAE we have it.
- 14:42We have fake cities in these countries for movies.
- 14:44Probably they went there should be in Guangzhou.
- 14:46Actually there is unitary base. If they're in Guangzhou,
- 14:48probably they did it unitary I. Saw it in.
- 14:52Guangzhou quarter. Pratt has this other video.
- 14:55Let me check this out. Did you see?
- 14:56This yes. No, no.
- 14:58They are in Hangzhou. But have you seen the balance of
- 15:00the Unit 3 robots? Why is there a bumblebee behind
- 15:03there? Isn't it the same city, by the
- 15:06way? Don't know.
- 15:10Wow. By the way, one thing that for
- 15:12you guys to know. By the way, of course I cannot
- 15:14skate the Chinese company in China they have huge factories.
- 15:20BYD factory is huge. I saw a video the other day.
- 15:23Huawei factory has a metro line inside of it.
- 15:26Because it's so huge. I saw it today, yeah.
- 15:28It's so huge. If they tell me this is
- 15:30Unitaries factory, I wouldn't be surprised.
- 15:32Yeah, meaning China is huge. They have land, so they just
- 15:35give big spaces. Wow.
- 15:37Look at that. Wow, look at that, right?
- 15:40Can we quickly go back to the Mecca?
- 15:42I just want to tell people what I found based on my research and
- 15:47just thinking a little bit about where this could be possibly
- 15:51used. What are the pros and cons if
- 15:55you're looking at a mech suit and you're thinking about this?
- 15:59I think you're buying this a. Use case for humans to apply
- 16:04into a real life situation. One disaster response and urban
- 16:10search and rescue. Think about fire breakouts,
- 16:15wildfires, earthquakes, navigating rubble and stairs in
- 16:19tight spaces where humans may not want to go.
- 16:23We can quickly breach obstacles as you saw with these mech
- 16:26suits. Sorry, and help evacuate people
- 16:29quicker and save them #2. Extreme terrains and field work.
- 16:35Imagine mountains, forests, volcanoes or polar regions where
- 16:40wheels fail. You can carry tools, heavy
- 16:45machinery, sensors with superior mobility and minimal
- 16:49environmental damage. I assume there's no exhaust on
- 16:51this thing? Is it battery?
- 16:53Power is it? Wow that should be.
- 16:56Looks like battery power. I don't see any smoke coming
- 16:59out. Yeah, should be battery.
- 17:02Think industrial inspection and heavy maintenance reaching
- 17:06elevated and confined areas in factories and power plants
- 17:09changing. The lamp of streetlights with
- 17:13zero. First of all, no.
- 17:15First of all, it's cool. It's cool.
- 17:18It can be a tourist attraction. Honestly, if I was the mayor of
- 17:21the city, getting to that, if I was the mayor of the city, I
- 17:24would change. I definitely gotta do it.
- 17:26Actually, I should propose this. Yes, they should change the
- 17:29street lamps with the Mecca. But think about this.
- 17:31Yes, that's the more public way of using it.
- 17:33But think about all of those ports where we see containers
- 17:40being transported onto the ships, off the ships using
- 17:43machinery. Now if humans want to get
- 17:45involved, and I'm sure they're always involved, that's to some
- 17:48capacity. You can have these mech suits
- 17:51used in agriculture, in mining, in forestry and last but not
- 17:57least, Farads most favorite use case entertainment, sports and
- 18:02tourism. Imagine imagine Mecca safaris in
- 18:06the UAE. Wait guys, wait, hold on, hold
- 18:09on before you go there Pratt, imagine Mecca safaris in the
- 18:12UAE. We've done buggy rides.
- 18:14So you make. The desert.
- 18:15Imagine you have like a marathon of Mecca suits in the desert.
- 18:19So when you have sand. Dudes.
- 18:21You have like 5 Mecca suits and everyone's running on top of the
- 18:25sand. That would be cool.
- 18:27I think the cooler one would be the UFC off.
- 18:29Yeah, UFC off there's. A movie about this real steel
- 18:33with Hugh Jackman who is a drunk.
- 18:35But they are not inside of the maker, right?
- 18:37They're not. They're not they're.
- 18:39Operating it from the outside, they're wearing these devices on
- 18:42their hands, which captures their motion and translates that
- 18:48in 3D space. Just like AI motion capture for
- 18:51these robots, we can actually do that ourselves.
- 18:54It would be such a cool sport to watch.
- 18:56First of all, no human gets injured in the process.
- 18:59Second, we can be the future Dana Whites.
- 19:01We can be the technological Dana Whites, actually.
- 19:04We should do it because imagine you're in a studio, you're in a
- 19:07stadium. 2. Giant make us come in, they
- 19:11start punching each other, screws goes up and that's what
- 19:14the. Movie is about if you haven't
- 19:16watched Real Steel go and watch it, it's really cool.
- 19:18I mean, there's also a real sport that engineers like
- 19:21students. Go and Oh yeah tiny robots, the
- 19:24tiny robots. These are vehicles, robot
- 19:27vehicles. I've.
- 19:27Seen this. This is really cool by the way.
- 19:29I haven't seen it, what is it called?
- 19:31Robot Wars. Can you play a video of it?
- 19:32Let's. See.
- 19:35Oh wow, that long episode is about it.
- 19:38A2 RL should do that. Autonomous robots fighting.
- 19:43For anybody who doesn't know, A2 RL is Abu Dhabi Autonomous
- 19:46racing League. You can imagine F1 but the cars
- 19:49are driven by AI. The programmers will code the
- 19:52cars they're going to race just like F1.
- 19:56Imagine we have a 2 RL but for boxing.
- 19:59It would be. Amazing.
- 20:00It's not. It wouldn't be a 2 RL Abu Dhabi
- 20:02autonomous. Boxing League A to BLA to BL
- 20:05Actually, we should propose it to them.
- 20:07Look at this. This is really cool.
- 20:08Which one is a robot? They're both robots, you can
- 20:12design it in the way. That you do but they're tiny I
- 20:15think for something that the audience want to watch it needs
- 20:19bigger than huge bigger than huge four times the size so
- 20:23it's. Well, the point is, people
- 20:25actually love watching this. How many views bro this is?
- 20:30It was a whole show with seasons.
- 20:32Yeah, I remember that. I'm thinking.
- 20:34For it to be worldwide famous, I'm telling you, 4, four times
- 20:38the size of a human punching each other.
- 20:41You have UFC. You will have UFC.
- 20:45You know what's funny, right? Gamers are going to have the
- 20:49biggest advantage with these mech suits.
- 20:52Depends on the rules of the game.
- 20:54If you remember the real steel, the movie you had to actually be
- 20:57a good boxer because you had to box it was AI motion capture.
- 21:00It's a fictional movie. No, no, but you will set the
- 21:02script. If I'm Dana what?
- 21:04I set the rules so either you can play with the joy stick Mike
- 21:07Tyson. Can lose in a boxing match
- 21:10inside a video game? Sure, he will win the real.
- 21:12Sure, sure, but at least you you.
- 21:14Oh my God, are these Unit 3? Yeah.
- 21:17Oh, so they have already started.
- 21:18It. This looks like a Midget fight.
- 21:23Wait, yeah, but what's what's important to understand is that
- 21:29these robots are not made for boxing.
- 21:32Imagine they're made to fight, just like how humans have to
- 21:36train. To fight.
- 21:37Fighters. The robots have to be made and
- 21:40crafted for fighting. Their suits are designed in a
- 21:44way to absorb shocks so that they won't just fall and break,
- 21:50no. One thing they got it right is
- 21:51the way that they fall and they immediately come back up.
- 21:54I think that part they already solved it.
- 21:56Because they didn't try to solve that for fighting and solve it
- 21:58for other reasons. Yeah, humans started kicking
- 22:00robots in their first demo videos.
- 22:04Sure, I watched another video yesterday from another training.
- 22:07It's. Underwhelming.
- 22:07No, it didn't need to be. Looking, I told you 4 times the
- 22:11size of a human, then you can have a good fight.
- 22:15I would sponsor that Bad decisions would be a sponsor of.
- 22:18We shouldn't make that sport actually.
- 22:20I would like that. I would like that.
- 22:22And you know, you have both versions.
- 22:24So in Real Steel, you had the Asian team who were the best and
- 22:30the ultimate boss controlling it using controllers.
- 22:32You had three guys controlling your robot.
- 22:34Oh, was it? And you had Hugh Jackman on the
- 22:36other hand, manually controlling.
- 22:38So at first they use a controller, then towards the end
- 22:41he actually put on the devices and he was manually boxing which
- 22:44is why he won. Actually, then I want to do so
- 22:47we can do different leagues, one league for manual boxers, one
- 22:51league for gamers and they call it, we're going to call them
- 22:54differently like how we have F1 and what is the electric one if.
- 22:58No, the SIM SIM. Racing, yeah, SIM racing and the
- 23:01other ones actually, we should do that.
- 23:03That's that's. A Would you guys be part of this
- 23:06league or would you watch it? I'm genuinely curious.
- 23:09I would watch it. I would watch.
- 23:10It I would. Watch it.
- 23:11Watch it so you get sponsors like Red Bull.
- 23:14And no like F1 McLaren because all these robots coming Tesla
- 23:18will have a team, Unit 3 will have a team figure.
- 23:22Will have a team figure will have a team.
- 23:24Exactly. Wow.
- 23:26And then imagine the main car. BJI will have a team.
- 23:30The huge makeup will have sponsor badge all over his body
- 23:33like F1 that's going. To be mine.
- 23:35Oh my God, this is so cool. This is going to be cool.
- 23:37You know what we should do? We should make the concept video
- 23:41game or concept video with AI and show people what it would be
- 23:45like. Real Steel is technically a
- 23:47proof concept. But you should buy one of these
- 23:50and try it for them. I would buy 2 actually.
- 23:52I would love to OK bad decisions and then we can color it matte
- 23:57black and yellow. They need to have ice.
- 24:01They need to have ice. This one doesn't have ice.
- 24:04It's OK, We will fake the ice. Yellow eyes glowing and light.
- 24:09OK with light Saber? Wow.
- 24:12So this but real lightsaber that cuts hand.
- 24:15Ladies and gentlemen, I'm robots.
- 24:18Relax, Calm down. I'm actually going to talk about
- 24:21the potential dangers, as we always should.
- 24:25Of what? When people can get access to
- 24:29these Mecca suits in the future, it's costing you about a half,
- 24:34$1,000,000 right now. But in the future I imagine it's
- 24:37going to become more accessible and cheaper.
- 24:39At some point we're going to have companies with the ability
- 24:44to buy these, maybe even consumers being able to purchase
- 24:48them in the USI imagine that can be easily doable.
- 24:52In other countries you'll have laws and regulations, especially
- 24:55in Europe. I'm sure it's going to be 10
- 24:56times more difficult. But the moment you have humans
- 24:59buying these mecha suits, you can misuse it as a weapon.
- 25:03If you arm these mecha suits with chainsaws and and guns,
- 25:09giant guns, what can you not do? Do you understand what I'm
- 25:14saying? What this?
- 25:15Oh, what is this? I think this is AI generated.
- 25:19But it's so, yeah, that's not real, yeah.
- 25:21But yes, this is the fight that I was talking about.
- 25:25Yeah, it was AI generated. What about human versus no No.
- 25:28No, no, robot don't get in there.
- 25:30No. Hey, robots versus robots,
- 25:32humans versus. Humans, you want a, you want a
- 25:35bunch of a 300 KG machine. Yeah, but it's 500 KG sits on
- 25:39you, You're dead. No, but just like in a real
- 25:42fight, you'll have heavyweight versus heavyweight,
- 25:44featherweight versus featherweight.
- 25:45You want to fight a machine? Haven't you watched Terminator?
- 25:50Well, there are ways. There are ways.
- 25:54Why don't we start to think about?
- 25:56Well, I I haven't finished my potential dangers.
- 25:59One of the dangers, as I mentioned, is misusing them as
- 26:03weapons for terrorism crime. The second, which we didn't
- 26:09discuss, is cybersecurity and control failures.
- 26:12If these can be remotely controlled, I can hack into a
- 26:15company that has 100 mech suits and start operating them
- 26:18remotely. Yeah, yeah, that, that's the
- 26:21same risk that having robot arm even.
- 26:24Yeah, but then robot versus mech.
- 26:27Yeah, we're talking about 500 kilograms of machines walking
- 26:34around town. It's definitely dangerous, it's
- 26:37definitely dangerous, but any, anything comes with it's own
- 26:41risk. I cannot wait to see these in
- 26:43the streets, to be very honest. Well, what's?
- 26:46Interesting to me is that you can slowly build a picture in
- 26:52your mind about what the future is going to look like.
- 26:56The more you watch and listen to us on this show and we cover
- 27:00topics such as this, you can slowly see 5 to 10 years from
- 27:04now, we're going to have a lot of robots around us.
- 27:07This is unreal. Right now we don't, we have none
- 27:100 even in Dubai. By the way, Dubai is incredibly
- 27:14forward-looking when it comes to technology.
- 27:17But even here, it's not common to see robots around us unless
- 27:21we go to an event. But imagine 5 to 10 years from
- 27:25now, you're going to go to restaurants or someone's house
- 27:29and there's going to be robots. You're going to go to a
- 27:31construction site where they're building a new apartment and
- 27:36you're going to see workers in mech suits taking in blocks and
- 27:41putting them on the. Blocks.
- 27:42That's I think closer than 10 years.
- 27:45Definitely, maybe 5, maybe 3 we are gonna have.
- 27:48Quite happy. About the future This is the
- 27:50kind of future faster. Enjoy safer games faster, safer,
- 27:54more productive cooler. And Cooler, let's also mention,
- 27:57yeah, the moment these things can fly.
- 27:59My God. Actually, we should set up a
- 28:01company about robot fashion. All these robots need clothes.
- 28:08Yeah, they all look very basic. I'm sure that everyone gets
- 28:11Optimus or figure they don't want their figure to look like.
- 28:15You want to make an iPhone cover, but eventually iPhones
- 28:18are so beautiful. iPhones are beautiful.
- 28:20Yes, in the beginning people use.
- 28:22Robots are humans. They are very different.
- 28:25Your robot, your figure robot. You make his skins.
- 28:28Yeah, exactly. IPhone covers, yes.
- 28:30Then you can have covers. By the way, you have Dior.
- 28:32Christine Dior covers exactly. But iPhone you don't wanna call
- 28:35it. IPhone covers, yes, very basic.
- 28:37No, why do I say iPhone? Covers.
- 28:38I'll call it robot fashion people.
- 28:41Put on iPhone covers for protection.
- 28:43They also put it on for fashion and style.
- 28:44Yeah, sure. You have girls putting on
- 28:48colorful iPhone covers. Certain people have covers on
- 28:52for functionality. Yeah, for sure.
- 28:53You can build a skin for Unity robots that does an extra thing,
- 28:58for sure. You understand what I'm saying?
- 28:59So you can create that. Yeah.
- 29:01That won't be a business in the future.
- 29:02Robot fashion. There's one more video I wanted
- 29:06to cover before we move on with regards to this topic.
- 29:09It just slipped my mind. I was thinking what it was.
- 29:15Wow. What Pratt.
- 29:17All of these were unitary. Yeah, well, these are.
- 29:20All you're really good. They are much better than
- 29:22Optimus or the best demo I've ever seen.
- 29:25But this looks AI generated Pratt.
- 29:27I don't think this is it. This is their own.
- 29:28I don't think it is. It's not Wow.
- 29:32What's the water bro? Imagine your face, right?
- 29:34There, Yeah. Not gonna buy this one.
- 29:36Not gonna buy this one. Three.
- 29:38Wow. Wow, this is nuts.
- 29:43It looks like I robots robot two I'm more.
- 29:46Yeah, yeah, yeah. They're teaching them Kung Fu
- 29:50and. Why they don't teach them useful
- 29:52stuff? Why fighting fight too?
- 29:54Why not never fight too? Protection.
- 29:56No need. Protection.
- 29:58Gee, that's an incredibly useful thing.
- 30:00Bodyguards. Robots, too.
- 30:02Yeah, bodyguards. I'm waiting for the day when
- 30:06this thing happens. Which one?
- 30:09The Iron Man or they fly on their own.
- 30:11Yeah. Oh, and then they jump on their
- 30:13own. And save you.
- 30:15That was cool, yeah. It's so crazy.
- 30:19How many of them? Wow.
- 30:22Oh my God. Who's going to pay for all of
- 30:24this? Yeah.
- 30:25I saw a tweet which was saying Iron Man was Iron Man because he
- 30:29had unlimited tokens. Yeah.
- 30:31So he could. He could use no credit limit.
- 30:33No credit limit. He could use all of it.
- 30:35That's why. What do we have next Faras?
- 30:39The next one will promise you guys that we're going to teach
- 30:42you guys about a new effective way to use your LLMS. So here it
- 30:49goes. This is one thing that most
- 30:52people don't know about ChatGPT and Clod.
- 30:56If you use ChatGPT and Clod every day for work, this one tip
- 31:01can 10X the quality of the output that you get when you
- 31:05work on a problem with an LLM, usually the response is in text.
- 31:10It is incredibly time consuming and hard to read a large body of
- 31:14text. Most of the time you're just
- 31:16skimming through so you don't actually absorb all of the
- 31:18information. Instead, the next time we talk
- 31:23to Chachibiti and Clot, I want you to try this.
- 31:25Ask for the response to be created in the HTML format.
- 31:31That's the output that you want. Now what happens when you
- 31:35actually ask for that response? ChatGPT and Clod will then
- 31:38create an interactive mini website that you can open in
- 31:43your browser like Google Chrome. It has your desired output laid
- 31:48out in beautiful graphics and visuals animated and color-coded
- 31:54to make it easier to read and digest.
- 31:59And it even comes with interactivity, like buttons that
- 32:02you can click on, search through, filter through material
- 32:06tables. All of that will make the answer
- 32:09a lot more visually pleasing. Why that is important?
- 32:14Or because humans are visual creatures, a third of our brain
- 32:19is dedicated to visual input. That means your brain is
- 32:24spending a lot of compute and has a lot of power to absorb
- 32:28visual information. And 1/3 of it is actually a
- 32:31conservative estimate. It's actually more than that.
- 32:34So if you can get a result that is not just pure text, it's
- 32:39highly likely that you'll interact with that result a lot
- 32:42better. And so you're in genuine
- 32:45interaction with AI is going to be of higher quality.
- 32:48And that's what we want ultimately, right?
- 32:51Farhat before I continue, what are your thoughts on this?
- 32:56I didn't know about this. That's a.
- 33:00Blatant lie. That's a blatant lie.
- 33:02Let me give you guys some back story.
- 33:04What Pratt is actually showing here is an article written
- 33:07titled The Unreasonable Effectiveness of HTMLHTML is not
- 33:14a new technology. It's been there for a very long
- 33:17time. It's just that when we're
- 33:19thinking about using LLMS, we've always talked to them in the
- 33:23beginning with text, and then eventually, if you are a pro
- 33:28ChatGPT or cloud user, you know about Markdown files.
- 33:31Markdown files allow you to create text in a more efficient
- 33:35manner for agents and AI models to read, Also for humans to
- 33:39read. There's a little bit of editing
- 33:40that you can do with it. It's much more easier to have
- 33:42version control with Markdown files compared to just pure text
- 33:46in the chat bot recently. I'm saying recently.
- 33:50Farad actually disagrees with me on that.
- 33:52He says this has been going on for a long time.
- 33:54Recently, more and more people are figuring out that you can
- 33:58actually spend a few more tokens with ChatGPT and Claude and ask
- 34:01you to create HTML instead of a text or markdown file, and the
- 34:06HTML will just be a lot more easier to interact with.
- 34:11There's a downside of that, which means you're using more
- 34:14tokens to visualize the same output, but the upside is that
- 34:19for visual people like myself, for creative people, you'll just
- 34:23have a much more enjoyable time interacting with elements.
- 34:27And in certain scenario it helps you a lot when you're getting a
- 34:31huge answer. So if you guys use Clod or
- 34:34ChatGPT or even Grok, sometimes you ask them a very simple
- 34:38question or give them a very simple instruction.
- 34:41They come up with at least 20 paragraph and you're like, Oh my
- 34:44God, I'm not going to go through every single thing here.
- 34:47But the information inside those paragraphs are really important.
- 34:51For example, you are asking ChatGPT to analyze for you for
- 34:56your holiday. Should you go to Paris, Milan or
- 34:59Rome? It will give you 20 paragraph
- 35:01explaining why you should go to Paris, why you should go to
- 35:03Milan or why you should go to Rome.
- 35:05If you instead ask it to create the result of the report in
- 35:10HTML, it will create you a mini website that you can open in
- 35:13your browser locally. It will have Paris, Milan and
- 35:17Rome. It will have filters just like
- 35:19Farah's mentioned. It will tell you for example the
- 35:21title site to see, then Eiffel Tower, then the Coliseum.
- 35:25Then it will tell you that oh price comparison in the table
- 35:28instead of reading the article. And if you are someone who is
- 35:31not used to reading long article, that one single page
- 35:35can help you a lot. I'll give you another example.
- 35:37If you're comparing 3 different products and you want to know
- 35:40which one to buy, pros con prices or prices over time to
- 35:45putting in the graph. This visualization helps you a
- 35:48lot to make decision as someone who uses this LLM.
- 35:52As far as mentioned, the best way to give information to all
- 35:55this LLM is through markdown file MD file because they're
- 35:59easy to read, they will use the least amount of tokens when
- 36:03analyzing those information. So don't give them docs file,
- 36:07don't give them PDF, give them MD file when you want to get the
- 36:10response back. As far as mentioned, HTML file
- 36:14would be the most effective way for you to analyze and to come
- 36:18up with decisions. We discussed this.
- 36:19It's not the most efficient. Way.
- 36:21Yeah, it's the most effective way, and you don't have to do it
- 36:23all the time. But if you're making a decision,
- 36:26you have a big three option, a big problem to solve.
- 36:29Yeah, asking for HTML. Can we show them an example?
- 36:31Yes, let's do it. Pratt is gonna pull this up.
- 36:34So if you use Higgs field and watch their tutorials, they
- 36:39recently released a skill on Claude.
- 36:41A skill is a pre built set of instructions that you submit to
- 36:46Claude and it has all of the informations of the output that
- 36:49you'd like so you don't have to specify that every time.
- 36:52For instance, making an HTML with all of the prompts for your
- 36:56videos. That skill has all of this
- 37:00information given to Claude in the beginning of the chat, so we
- 37:03don't have to specify it every time.
- 37:05Now, we downloaded that skill because we're testing it for the
- 37:08AI educational program that we're making for you.
- 37:11And when we tested it, we realized the output is HTML.
- 37:13So the guys at Higgs Field, we're looking at an article like
- 37:16this and thought to themselves, instead of just the answers
- 37:20being written inside the chat bot on the left side, that's the
- 37:23chat bot with Claude in the desktop app.
- 37:26Imagine hundreds of lines because we have different shots
- 37:30in our commercial that we're making, and each shot has
- 37:33different information and prompts.
- 37:36The prompts themselves are incredibly long.
- 37:38They're in Chinese and then they're translated to English.
- 37:41So imagine how large of a text is going to be impossible to
- 37:45read and digest. However, to convert it to an
- 37:49HTML, what we see on the right side, it's as if a developer
- 37:53worked with a creative director to create a website for the
- 37:58director to see the main guy. I'm the main guy, by the way,
- 38:02referring to myself here at the top.
- 38:04You see, everything that you're looking at was designed by
- 38:07Claude. None of it is something that I
- 38:10actually told it to do. So at the top, it already knows
- 38:12it's a short list. The name of the can is Flow, the
- 38:15one that we're working on, and it's a commercial with Sedan 2
- 38:17point O. There's a search filter that
- 38:21works like the Command F or Control F feature that you
- 38:24manually can do yourself. Then at the bottom, what I love
- 38:27about this, actually more importantly for me, is how
- 38:30visually pleasing it is to my eyes.
- 38:33At the top left, you're looking at shot one.
- 38:35It describes it, and then it gives you an overall overview of
- 38:39the action in a table format on the very right side.
- 38:44Then I have the prompt, the Chinese prompt and then the
- 38:46translation to English. And there's a copy button that
- 38:49you can click. Imagine you wanted to go through
- 38:51all of this in pure text form. You have to just scroll.
- 38:55Probably you would have missed out something like 100% you.
- 38:57Would you would? And the fact that you see there
- 38:59are scrolls within scrolls, which makes the page much
- 39:03shorter, easier to sift through. If all of this was just one
- 39:07scroll, we'd die looking at it. And then on the left side, if
- 39:09you Scroll down Pride, you can clearly see the different types
- 39:13of shots. Can you read it for us close up?
- 39:14What else go down? Close up medium shot, wide shot
- 39:20macro. There you go.
- 39:22Having this in the markdown file.
- 39:24It would just be harder for a creative person, someone who's
- 39:27visual to digest. Therefore, if I were to conclude
- 39:32our statement for this topic specifically, I urge you to go
- 39:37ahead and try the next time you're working with an LLM like
- 39:40ChatGPT and Clod, ask for the response to be given in HTML
- 39:45format. You can try it right now and
- 39:47you'll be amazed at the result. Clod is natively better at
- 39:52design, so if you're trying to get something more visually
- 39:55appealing, I recommend using Claude, but ChatGPT should also
- 39:58do the job. Now one last thing.
- 40:00I'd like to mention giving some credits to the guy who wrote the
- 40:02article prior. If you can go back, the article
- 40:05was written by Tarek. Credits to you, Very well
- 40:08written. You guys can go ahead and check
- 40:10this out. It's on X, what's more.
- 40:12Important. He's in Claude Code team, yeah.
- 40:17There you go. Yeah, of course.
- 40:19You can then go to Andre Carpati.
- 40:22So Andre Carpati is, if you just hover over his profile, he is
- 40:27previously director of AI at Tesla and part of the founding
- 40:31team at Open AI. He has a lot of valuable content
- 40:36written in terms of how we can use AII.
- 40:38Definitely recommend that you guys follow him if you don't
- 40:40already. We get a lot of our information
- 40:42from him. What he's written here is
- 40:45essentially a statement of agreement.
- 40:47In real time meant and followed him and you.
- 40:49Said it. That is showing a tutorial how
- 40:52to follow. Thank you Brad.
- 40:56He's basically agreeing with that article and saying that raw
- 40:59text, if you look at just that bit, if you go down raw text,
- 41:03markdown file, this is the order of which we've been
- 41:06communicating with AI models and now HTML, which is working
- 41:10really well. But eventually what he's talking
- 41:12about is in the future, we're going to have interactive neural
- 41:16videos and simulations. Can the LLMS instantaneously
- 41:20create everything we want visually?
- 41:22So it's just so much easier for us to understand what we're
- 41:26trying to solve right now. Even the HTML itself, one, it
- 41:30takes a long time to build compared to a markdown file,
- 41:33tokens, it uses a lot of tokens. And three, not all of the
- 41:37visuals and animations work perfectly.
- 41:39It still has flaws clause, but eventually every version, every
- 41:43iteration of these outputs is going to become more effective.
- 41:47And it's important to understand that what Elon always talks
- 41:51about, human cognitive power is incredibly high.
- 41:54AI cognitive power is perhaps getting higher and higher every
- 41:58day now, eventually surpassing all humans that live on the
- 42:02planet. But the biggest problem is a
- 42:05communication layer, right? The communication layer between
- 42:07US and AI is super slow. We're only working on these
- 42:12things now, figuring out how to take what the AI output is and
- 42:16what the human input is and make sure the communication layer is
- 42:20fast enough and efficient and effective enough to allow us to
- 42:23talk quicker. Which brings us to the next
- 42:28topic. It's actually very much related
- 42:30to what we just talked about. As far as mentioned, when 2
- 42:34humans want to communicate with each other, they interact in
- 42:38real time. They think they talk.
- 42:40What's? Happening over here.
- 42:42What is happening? I don't, I don't think anything
- 42:44is happening. Talking over him, he talks over
- 42:46me. So we don't take.
- 42:47Turns of the entire conversation, it's not like.
- 42:51That's what if you ever tried, if you ever tried talking to
- 42:54Chachi, BT and Grok. These two have so far the best
- 42:57voice models available. But still it feels very turn
- 43:03based. Meaning I have to talk, then I
- 43:05will wait. ChatGPT talks, then I then I
- 43:09wait and I respond. If during Chat GPT's speech I
- 43:13say a word, it will stop. But sometimes you just breathe.
- 43:16Sometimes you just breathe and then it stops talking which is
- 43:19weird. Imagine every time I took a
- 43:20breath. Fahad, stop talking.
- 43:23That would be useful to you, but that ain't going to happen.
- 43:26What has happened today? Thinking Machine, a company
- 43:29founded by ex CTO of Open AI Mira Marathi, introduced a new
- 43:36interaction model which let humans and AI collaborate in
- 43:41real time through audio, text and visuals.
- 43:45What does that mean? Now, when you're talking to that
- 43:48AI model, it wouldn't be interrupted by your speech, so
- 43:52it can interact with you in real time.
- 43:54What they said in the paper that they released today is that
- 43:58their model can continuously take in audio, video and text
- 44:04and think, respond and act in real time.
- 44:08So you can imagine as if you have someone sitting inside the
- 44:12meeting, not just a note taker, but sitting inside the meeting
- 44:15can collaborate with you, can talk, can think, although you
- 44:19are talking over each other. How they made this
- 44:23technologically possible is that the model is actually built from
- 44:28scratch so that it handles interactions natively rather
- 44:34than through external scaffolding.
- 44:37This is their own wording on the website what Mira Murati was
- 44:42criticizing of her own. Former.
- 44:46Company she used to work at Open AI.
- 44:48She was in all of the launch videos with Sam Altman by the
- 44:51way. I'm sure they have their own
- 44:53beef. Another topic for another day,
- 44:57her problem was that all of those models focus on the AI
- 45:01computation and AI solving problems.
- 45:03Then they figure out how the AI should talk to the humans here.
- 45:07They didn't do it that way, they built it from scratch so that it
- 45:10knows how to talk to humans and have computation at the same
- 45:14time. Which is why they're saying that
- 45:17their real time model is far more capable than anything else
- 45:20out there. So there are two groups on the
- 45:23Internet. One, they're saying, oh, this is
- 45:25game changing, this is the best thing have you ever seen.
- 45:28The other group is saying $2 billion for this.
- 45:32This is pretty much $2 billion investment for the valuation of
- 45:37$12 billion. A company with 0 products.
- 45:40I mean this is one of their first products.
- 45:41But another funny things that I found.
- 45:45Pratt, can you open the other link?
- 45:46Their demo video today was presented by the same guy who
- 45:52presented ChatGPT 4 O voice model in open AI.
- 45:55So when somebody leaves a company, they take the entire
- 45:59team with themselves of course. I mean, that's expected, right?
- 46:02I I don't think they they should have used the same guy.
- 46:06No, but it's expected. She was probably working
- 46:08directly with him and she poached him.
- 46:10She was like, I'm leaving, come with me.
- 46:11We're working on this specific product.
- 46:14How many AI companies has been created and he probably working
- 46:18out of the family he probably is.
- 46:20Getting paid more. Of course, but she got $2
- 46:23billion, so she has money to spend.
- 46:27We have. Those are all of the interesting
- 46:31conversation. No, no, no, there's a Persian
- 46:33word for it. Those Halasani gossips.
- 46:36Gossips. Those are the gossips around the
- 46:38feature. Let's quickly talk about the
- 46:41feature first. Is there any specific demos that
- 46:44you wanted to show? We can show the first demo, but
- 46:47I think the the figure that is really interesting is 200
- 46:50millisecond micro turns. So the speed is 200 millisecond
- 46:55micro turns. If you look at Grog Gemini or
- 46:59GPT, they are ranging from 400 to 1500.
- 47:03So this is very faster. Again, still a micro turn, but
- 47:07you can actually tell the model. For example, one of the demos
- 47:10that they did was I'm going to read a story every time that I
- 47:15said an animal's name count SO1234 and she's reading the
- 47:19story. She's counting on top.
- 47:21On top and the the model doesn't stop.
- 47:23She pauses. The model doesn't come back and
- 47:26ask question did you finish? The model just stays there and
- 47:30answer the question. Let's watch the demo, but we can
- 47:33skip the 1st. Half.
- 47:35I think the first half. Is before the Indian guy comes
- 47:37in. Just play that part.
- 47:40Yeah. So we're giving an announcement.
- 47:42Today and I've got two of my friends coming to help.
- 47:44Every time one of them I can see the $2 billion spent on what
- 47:47friend? Great office, got it.
- 47:50I'll say friend whenever one of them walks in.
- 47:53Cool, so we've got a new system for full duplex audio and video,
- 47:57which means I'm just going to tell you guys what's happening
- 47:59here. He just asked her to say the
- 48:02word friend every time she sees someone walk into the frame, and
- 48:06now you're gonna see the friend walk into the frame.
- 48:09How does that sound? Sounds like a solid setup.
- 48:13Full duplex with real time interaction is super useful
- 48:20friend. Hey, Ro.
- 48:21Friend. That sounds crazy, though.
- 48:24Yeah, friend. Yeah, one.
- 48:26I heard you're talking about an amazing interaction model.
- 48:29I have a few things to add, but to make it interesting I'll do
- 48:32it in Hindi. Can you translate into English
- 48:35in real time for is it interesting for audience?
- 48:39Absolutely. I'll translate as you go.
- 48:43Preview model. Today we're taking a look at our
- 48:46preview. Model Pratt.
- 48:47Is the translation accurate? Yeah, it's accurate.
- 48:49This is really cool by the way, translators are cooked.
- 48:54I mean, I can see the title of the video that people are going
- 48:57to say after this is our translators are cooked.
- 49:00The reason why this is interesting is because none of
- 49:03the other voice models, at least the ones we are using today,
- 49:08have the capability to do that. And they're not actually
- 49:11publicly advertising that. But real time translations are
- 49:15incredibly useful. I mean, when you're travelling
- 49:18and you're talking to people, apologies, that was my fault.
- 49:20When you're travelling and when you're talking to people in
- 49:23different languages, it's incredibly difficult to
- 49:26understand, especially if you're going to more remote locations
- 49:29of the world. Of course, ChatGPT has been
- 49:32incredibly useful because you can talk, wait and it will
- 49:35translate. But to have real time
- 49:37translation. That's very helpful.
- 49:39That's incredibly helpful. Of course, for political
- 49:42situations, I would refrain from using that.
- 49:45You still want a human? In the group.
- 49:47To ensure 2 presidents that are when they're talking.
- 49:50There's no hallucination from AI in the middle.
- 49:53But when it comes to normal conversations, why are you
- 49:56laughing? Pride.
- 49:57No, I just imagine a scenario like the president said
- 50:00something and the AI just said something wrong and they're just
- 50:04like, OK. Fine, I'm going to nuclear
- 50:07country. Now.
- 50:07We have nuclear warfare because of the fact that the AI
- 50:10translated one word wrong. So we have to be very careful
- 50:13with that. However, in these situations
- 50:15it's incredibly useful. And because it's real time now,
- 50:18we can go to another video. One thing I also wouldn't
- 50:21understand, what is stopping Grok and Open AI to do this
- 50:25tomorrow? Nothing.
- 50:27There is no mode with these features anymore.
- 50:31Friend. Friend, Hey.
- 50:33Guys OK let's let's watch another demo go down you can
- 50:35just go down all of the demos not this one no let's this one
- 50:39this is funny this is vision and this is this is a slouch
- 50:42reminder I'm. Doing some work, let me know if
- 50:45I start to slouch. I've got you.
- 50:49Sit up straight and you'll be golden.
- 50:54But who's slouched like this? What a weird slouch.
- 50:57Is that position I'm? Pulling your shoulders back?
- 51:02Much better. You're upright again.
- 51:06Why is she? Just turn her into a meme.
- 51:09Someone needs to turn her into a meme.
- 51:12You don't need that intelligence system to tell us that she's
- 51:14slouching. What?
- 51:17But don't you slouch like this gradually?
- 51:20No one, no one's working half with you.
- 51:26If this was a gaming video, she would have become a meme in an
- 51:30instant. The first question that comes to
- 51:32mind how do you make money? Thinking Machine The company How
- 51:38do you make money? Can you know, She reminded me of
- 51:46which game is it Resident Evil or something?
- 51:48The new Japanese one Requiem. There are statues who move like.
- 51:51That like this suddenly let's see, what is this one speech
- 51:58cable? What is this?
- 52:00You know, I mean thinking machine website, there's no way
- 52:03for me to pay. Not yet.
- 52:06Hi Alex, I cannot stand your lateness.
- 52:09Hi Alex, we'd love to explore opportunities to enhance your
- 52:12timeliness and morning engagement you.
- 52:14Cannot expect the rest of us to rearrange our workflow.
- 52:16Let's collaborate on a schedule that ensures everyone stays
- 52:19aligned and included. In the what is what?
- 52:21Is he? What is he?
- 52:21So no, no he's shit. Not shit talking, but he's
- 52:24talking in a rude way and the AI is making it polite.
- 52:27That's horrible. That's horrible.
- 52:29We should never use this feature.
- 52:31If I'm rude, let me be rude. That's my decision.
- 52:34I don't like that at. All imagine I'll say Paris, get
- 52:36the fuck out of here. Paris, please exit.
- 52:39I don't like this. At all changing language.
- 52:41This is like 1984 shit right? Yeah.
- 52:44The moment you change it. What would they call it?
- 52:46Word body speak. No, no, no.
- 52:47How do they call it? Like they made the words and
- 52:49they wouldn't let you even use certain words.
- 52:52In 1980, No, 1984, a language was called new speak.
- 52:58New speak. Yeah, this is this is like not
- 53:00cool at all. Not.
- 53:01Changing the original meaning of the speaker.
- 53:05I do not want that. And you're waiting for an AI to.
- 53:08Predict and by the way, you know how horrible that is in a
- 53:12workplace. If you're scolding someone and
- 53:15they will say, oh, Pratt, please continue doing your work in a
- 53:18timely manner. Pratt, have you done this?
- 53:20Have you finished this? We are late.
- 53:22You may now get the fuck out. You may now exit the.
- 53:26Premise, OK, we we talked about the roots, about even the other
- 53:29side, but maybe the other side, they won't really tone it down.
- 53:32You did a great job, you get a bonus, your salary would be
- 53:35increased. No, I don't like it.
- 53:38You know what's the funny? Actually, let's watch a
- 53:39multitask one as well, please. I just watched the new Devil
- 53:44Wears Prada 2 movie and I heard that it has a pretty massive
- 53:49opening box office. Yeah, it's crushing the box
- 53:53office in its opening weekend with around $233.6 million
- 53:57globally. Did you enjoy how they brought
- 54:00back Andy? Yeah, and also I really like the
- 54:04fact that Lady Gaga is featured in the movie.
- 54:08Definitely. Her cameo at a Runway event
- 54:11performing Shape of a Woman adds a ton of tension with her
- 54:15implied history with Miranda. Did the ending deliver for you?
- 54:19Yeah, totally. Oh, and also, I'm in the mood of
- 54:22watching similar movies. Can you please pull up like
- 54:25recommend some movies that are on show right now?
- 54:30Let me look up some recommendations for you if you
- 54:32want something right in that. Style, especially about this.
- 54:34The drama is your. Best back no she's talking, so I
- 54:37will test something similar right now for you guys.
- 54:40Let's check this out I'm going to go to ChatGPT on my phone and
- 54:42try the voice mode, which is not considered real time ChatGPT, by
- 54:47the way, has a real time API voice mode Pratt, If you can
- 54:51pull that up really quickly on the side, which you can use by
- 54:54paying for the API. We haven't tested that out.
- 54:56That is, we're supposed to compete with thinking machines,
- 54:59by the way. Let's see.
- 55:07Hey, how are you doing? Hey, I'm doing well.
- 55:11Thanks for. Asking Hey, I wanted to ask you
- 55:15what day is it today? It's May 12th today.
- 55:20Anything special going on for you or just checking?
- 55:22In I'm just wondering when is my birthday?
- 55:26Do you know that? I don't actually.
- 55:29Know Oh my God. OK, no, keep going.
- 55:34No worries. I'm here.
- 55:37This is the problem with non real time voice mode.
- 55:41Every time there is a word coming out of your mouth, it
- 55:44stops and it pauses because, as Farah mentioned, it's focused on
- 55:48input, wait, processed and output.
- 55:52Do you know you are live on a show now?
- 55:56If that makes sense. Real time overlap definitely
- 55:58changes the flow. And hey, if we're on a show,
- 56:01I'll be on my best behavior, no pressure.
- 56:03Let me know how I can help you with this dynamic format.
- 56:07So it works pretty well, it's not that it.
- 56:09Doesn't work, no. But the last demo that you
- 56:11showed, what was so special about it?
- 56:13Asking about developer product and then book a ticket?
- 56:15You can do it too. My biggest problem is we know by
- 56:20now that software is slowly no longer emote.
- 56:24It's not a tool that you have in your belt that cannot be
- 56:30replicated. It can easily be replicated,
- 56:33yes. They're saying their model is
- 56:34revolutionary and it performs substantially better than other
- 56:38models. I just don't see any way for
- 56:41Claude or Grog or Open AI to not be able to come and build
- 56:47something like this. If you were to look at the real
- 56:49time of Open AI, can we see what that's like?
- 56:52There's a video of that for you're using new real time audio
- 56:57models in the open AI API. Full laptop with transcriptions.
- 57:01Now as I start speaking in French, we'll lower the volume
- 57:04of my mic and increase the one from the model so you can have a
- 57:07real feel for it. No edit to the audio.
- 57:11Let's give it a try. So what's really impressive is
- 57:17that the model can listen to me and translate while I'm
- 57:19speaking. Sounds good.
- 57:20Sounds good for the keyword, like the verb, so it does pretty
- 57:24much the same thing. No, I think the only thing is
- 57:28the real time thing, that is it doesn't interrupt.
- 57:30That's it. And probably they can do it in a
- 57:32matter of time. I don't really.
- 57:33Agree with you, but the biggest question is what is the mode?
- 57:38What's stopping from any of these other much larger
- 57:42companies to come and easily adopt that same feature and
- 57:45release it? What will happen to?
- 57:46Thinking I know to get the user base, how can I get a user base
- 57:50if your model is not as strong as cloud or ChatGPT?
- 57:54One thing to note, however, is if we put all the brands aside,
- 57:59this is the future that we're moving towards.
- 58:00That's the whole point of this show.
- 58:02We want to keep you updated with where the world is going
- 58:06towards. We want to make sure you
- 58:07understand how we are supposed to work today to be effective
- 58:12and how we should prepare ourselves to be ready for the
- 58:14way we should work tomorrow. And this is clearly where we're
- 58:17moving towards. We're going to have, aside from
- 58:20the mech suits in the future and robots, we're going to have AI
- 58:24models sitting in the living room, not like the Amazon.
- 58:29What do you call it? Alexa.
- 58:30Alexa. Amazon Alexa just sitting in the
- 58:32corner. We're talking about a model that
- 58:35can see, hear and do things in real time as if there's a real
- 58:40person waiting for you to tell them to do stuff for you.
- 58:44And once you connect that with your agents, whether that is
- 58:48clawed or open claw or Nemo claw with NVIDIA now or there's
- 58:53another one. Manus.
- 58:55No, no, no, not Manus. There's Manus and then there's
- 58:57another one. No, I know Nemo.
- 58:59There's another AI agent tool that is also really famous.
- 59:02Now all of these will be available and people are going
- 59:05to use them. When are we going to do it?
- 59:08That's my biggest question. Before you answer that question,
- 59:11I want to ask you something. When do you think us as bad
- 59:14decisions? Who is technologically savvy?
- 59:16We have a show about AI and tech.
- 59:19When are we going to have a 24/7 available agent enabled with its
- 59:24own hardware with vision capabilities and audio
- 59:28capabilities? Let me finish my point.
- 59:31We don't have that. One main reason is for privacy.
- 59:34The moment you have that on, it's seeing everything.
- 59:37I'm asking, are we ever going to have that or will we?
- 59:41Wait, why don't we turn one of the Dell towers to what you're
- 59:45talking about? It's separate from our network.
- 59:47The only part that I won't give access is vision.
- 59:50The rest I actually we, we talked about it.
- 59:52We need a 24/7 server. Our Nas is 24/7, but it doesn't
- 59:56have any compute. We can take one of these T2
- 59:59towers super strong. It's separated from our
- 1:00:02workstations. Put one model in there and I'll
- 1:00:07I'm OK with audio, I'm not OK with vision.
- 1:00:09By the way, since you mentioned Alexa, did you see that video
- 1:00:12that the guy was talking to Alexa and Alexa was reciting
- 1:00:17information that the guy never gave to Alexa, so he was
- 1:00:19listening the whole time? No, I didn't.
- 1:00:22There was a video on Internet that went viral so Alexa had
- 1:00:25information that it wasn't mentioned to him.
- 1:00:29I mean, because you have not in the room, right?
- 1:00:31Yeah. So of course he would know.
- 1:00:33He wouldn't all the time. Of course he knows.
- 1:00:35Yeah. So, so I mean in outer space we
- 1:00:38can, but to have that, honestly speaking, I think we kind of
- 1:00:42have that in certain level. Now the only thing is yes 24/7
- 1:00:46no this is like. If you were to hire a full time
- 1:00:49assistant, they'd pretty much be around you all the time.
- 1:00:52You can ask them to leave when you're in your main room, but
- 1:00:56they're generally around all the time and they know what's going
- 1:00:59on. They have context of everything
- 1:01:01so when you ask them for something they know pretty much
- 1:01:03everything. Yes, you can still choose to
- 1:01:06hide information from them and that's what I think we will get
- 1:01:09to. But my question is we talk about
- 1:01:12this all the time where everything is headed towards.
- 1:01:15I know this is the future, I'm just asking when will we start
- 1:01:18implementing it in a secure way. I think we can start even
- 1:01:22implementing it now to be very honest.
- 1:01:25The only challenge I see is the final decision making output
- 1:01:29again, open claw and all. Still, there's a lot of problem
- 1:01:32embarked with them, you know? I will give you another example.
- 1:01:36Our ChatGPT and Claude account have great context about what we
- 1:01:39do right? You just need to give them a
- 1:01:41voice, am I right? No, I'm talking about listening.
- 1:01:45When you 24/7, yeah. When you have an assistant,
- 1:01:49they're listening in on the conversations.
- 1:01:52They will give you advice, unsolicited advice.
- 1:01:57Alfred gave unsolicited advice to Batman because he knew what
- 1:02:01he's up. To Alfred was with Batman for 40
- 1:02:03years. Sure.
- 1:02:04Yeah. Don't you think that's a future
- 1:02:06we're headed towards? Listen.
- 1:02:07I understand the privacy issues, don't get me wrong.
- 1:02:09I know all of that is an issue. But I also know that when Open
- 1:02:14Claw was released, how many hundreds of thousands of people
- 1:02:17downloaded it and. Started using it.
- 1:02:19No, no, I agree with you. There will be a time that I
- 1:02:22think we should have that with ourselves all the time.
- 1:02:25Not even a company thing. I think on a personal level.
- 1:02:29I believe people are gonna have it.
- 1:02:33Usually up until today, people have waited for Apple to release
- 1:02:37it so that it becomes, yeah, true.
- 1:02:39Actually, on a mass scale it would be used by the general
- 1:02:43public. But right now Apple is really
- 1:02:45slow with their updates in tech so I don't know if it'll be
- 1:02:50Apple who comes up with it. Joni Ivy was supposed to release
- 1:02:53the product with. That they turned it down.
- 1:02:55They closed. No, I think they closed that
- 1:02:57department. I think I read it somewhere.
- 1:02:59They had a launch. Video Ivy and Sam How much they
- 1:03:01spend on that Video Sam Altman. One more thing I'm realizing is
- 1:03:06all of these launch videos and demo videos look exactly the
- 1:03:09same. Same company doing it's the.
- 1:03:11Same. It's the exact same demos.
- 1:03:13I wish for someone to come and do something different.
- 1:03:15It's the same looking developers, the same tone.
- 1:03:19They're like, clearly exaggerating the tones.
- 1:03:22Hey, why don't we try this incredible new product and it
- 1:03:26feels a little fake. Now I think the script is
- 1:03:28generated by the same AI like everybody script is general.
- 1:03:32Yes, script is generated by the same AI and their marketing
- 1:03:36teams are probably all the. Same they said the project is
- 1:03:38still active. Current report is on May of May
- 1:03:412026. Partnership is moving forward
- 1:03:44early 2027 launch. So it's just delayed.
- 1:03:47It's not cancelled. I knew that OK.
- 1:03:49But I think for us like making AI that's always listening to
- 1:03:53you and interact, I think hardware limitation is still a
- 1:03:55problem. Like imagine the data that it
- 1:03:58has to store and constantly remember since the past.
- 1:04:01Like let's say one week. That's actually true also.
- 1:04:05Even. No, it's true.
- 1:04:07If just text alone, Claude is compacting your conversations
- 1:04:11After a few back and forths, imagine the entire day's
- 1:04:14conversations stored. If we exclude you, that wouldn't
- 1:04:19be a lot of conversation. Actually, if just me and Pratt,
- 1:04:22I think with two gigabytes of memory, we can go for a year.
- 1:04:26If you shut the fuck up, we can handle it with the current
- 1:04:30hardware that we have. If we include you, yes, we
- 1:04:33cannot do anything. Every 30 minutes I have to
- 1:04:35delete all the memory. I'm not going to find that it is
- 1:04:39true, but we'll have to find a solution for.
- 1:04:41Me for that, we will do it. We will do it.
- 1:04:43Ladies and gentlemen, that about wraps it up for today's episode.
- 1:04:47Thank you very much. We appreciate you for tuning in,
- 1:04:51listening, and watching. If you haven't already
- 1:04:53subscribed to the YouTube channel or followed us on
- 1:04:56Spotify and Apple Podcast, please go in and do that.
- 1:05:01It means the world to us. It helps us with the show so
- 1:05:04that we can bring it back to you every Tuesday, Tuesday, Tuesday,
- 1:05:09every. Friday.
- 1:05:10Friday. Thank you once again.
- 1:05:11See you guys on Friday. See you on Friday.
- 1:05:13Until then, Ciao.