Latest / Bad Decisions Podcast / AI Created an entire 3D Scene from an Image 👀
Transcript
- 0:00Higgs Field releases Supercomputer, a smart AI agent
- 0:04that can automatically plan and create photos and videos for
- 0:07your products and services. But is it actually any good
- 0:11we're going to find? Out you can now generate a fully
- 0:14interactive 3D environment using just a single reference image
- 0:19and codecs is now in chat GP TS mobile app which means it can
- 0:25control your computer. A-Team of humanoid robots have
- 0:29been live for the past 40-5 hours on YouTube doing a factory
- 0:34shift with no failures. This is a crazy story that we're
- 0:39going to cover in today's show, ladies and gentlemen.
- 0:41Welcome back. And as you can.
- 0:43See we have a different setup. We've got A2 Cam setup as we
- 0:47promised you guys in this season, we're going to be
- 0:49upgrading the show to bring you a higher quality of content
- 0:54every single week. As you know, we go live on
- 0:57Tuesdays and Fridays, bring you the latest and the greatest in
- 1:01the world of technology and AI. There is one downside to this
- 1:05setup. Although we do have a beautiful
- 1:06quality, I'm a little further away from my brother.
- 1:08Actually I'm feeling so much spacious here.
- 1:11It's much better now. I just.
- 1:12Said I miss you because there's so much space.
- 1:14I miss you. Too, but this is a much bigger
- 1:16space and you see. You see the difference between
- 1:19these brothers? I was expressing the sadness I
- 1:23felt being so far apart from him at the other edge of the table
- 1:26and he's like, I have so much more space.
- 1:28But it's the look to you in the eye.
- 1:29I express love. I express love too.
- 1:33That wasn't love. That was not love.
- 1:35Man, no. But this is a much better
- 1:36position for debates. I can look you straight in the
- 1:39eye and tell you that you're wrong.
- 1:41Fair enough, if I am ever. All the time.
- 1:44We just wanted to tell you guys that we're so excited to have
- 1:46this set up. We were planning this out with
- 1:49Pratt and Farhad for the past few weeks to ensure we can get
- 1:52all of the equipment right. I wasn't involved actually.
- 1:55Farhad didn't help at all, but we finally got it.
- 1:57Hopefully you guys get to enjoy this higher quality.
- 2:00Without further ado, shall we begin with our first topic of
- 2:03the day? Let's go.
- 2:05Higgsfield Supercomputer. What is it?
- 2:09Why should you care? Well, Higgsfield recently
- 2:13released Supercomputer, which is the promise of a smart AI agent
- 2:18that can automatically plan and create photos and videos for
- 2:23your products and services. This is the Holy Grail for
- 2:27businesses and entrepreneurs who do not want to spend a lot of
- 2:31time or don't have the resources to hire 50 influencers or an
- 2:36entire professional studio to create photos and videos for
- 2:39them. Well, these are some of the
- 2:41examples they've been talking about.
- 2:42You can create 50 UGC reviews of your product using AI
- 2:47influencers. You can create 20 market listing
- 2:50photos of your product for Amazon and Shopify.
- 2:53You can create a 2 minute long commercial for your product or
- 2:56services. You can create motion graphics
- 2:58for the new app that you're launching.
- 3:00You can create a cinematic scene of two actors having a fight for
- 3:04previous of your next short film.
- 3:06Whatever it is, this AI agent is supposedly smart enough to plan
- 3:10it out for you and do all of the work and come back to you with
- 3:14the results. But does it actually live up to
- 3:18the promise? We're going to find out on the
- 3:20show today because we put it to the test.
- 3:23Farhad, shall we look at the results?
- 3:25Let's go and see so farhaz, why don't you first tell us what was
- 3:29the process that you put this new feature into test?
- 3:32That's a great question. First, we're going to go to
- 3:35Higgs field. OK, arrow 4.
- 3:36O4 OK then this is finished by. As you enter Higgs field.
- 3:41For anybody who's unaware of Higgs field.
- 3:43By the way, Higgs Field is the most popular platform where you
- 3:47get to have access to all of the photo and video generation
- 3:51models. We're talking about Nano Banana
- 3:54by Google for Image GPT, Image 2 by Open AI, we're talking about
- 3:58Sea Dance and Cling for video generation and so much more.
- 4:02They're very well known because of the fact that they release
- 4:04features on a weekly basis and they're very up to date.
- 4:09Their user interface is also incredibly user friendly and the
- 4:13marketing is really strong. So you've probably heard of
- 4:16them, but if you haven't, I just wanted to give you guys a little
- 4:19beginner intro into what Higgs Field is Now.
- 4:21Supercomputer is one of those new features they just released
- 4:24and they're hyping this up on the Internet.
- 4:26We just wanted to test it out and show you guys the
- 4:28practicality of it as we always do on this show.
- 4:31Farhat and Pratt, what I actually did was I went into
- 4:34Supercomputer. So Pratt is going to click on it
- 4:36right now. It opens up a completely new tab
- 4:39within Higgs Field and you are going to be faced with a chat
- 4:43bot chat box. This is now the Holy Grail of
- 4:47any sort of chatbot on LLM. Claude looks like this.
- 4:50GPT looks like this. Gemini looks like this.
- 4:53Exactly. And because you're actually
- 4:55talking to these models, yeah, technically, if Pratt can do me
- 4:59a favor and click on the Claude Opus 4.7, instead of going to
- 5:04Claude and asking it to help you with the prompts for your ads or
- 5:08going to Gemini or going to ChatGPT, they've bundled all of
- 5:12that together in one place. Didn't we say that was the
- 5:16promise of Hicksfield? Anyways?
- 5:18They bundled cling and C dance in one place.
- 5:20You don't go to C Dance and use their model.
- 5:23You don't go to cling separately.
- 5:24You don't pay them credits to get credits.
- 5:27You're paying Hicksfield to get credits to use everything.
- 5:30Up until now, it's just been photo and video generation
- 5:33models. But now they know that the
- 5:35process almost always relies on the AI models, the LLMS, writing
- 5:40out the prompts. So they said, you know what, We
- 5:42know people are suffering in coming up with the ideas for the
- 5:46prompts, especially when it comes to business owners and
- 5:49entrepreneurs who have no creative background.
- 5:51We want to help those guys, creatives, and we're going to
- 5:54get to you guys later on. Creatives, artists, creators.
- 5:59We know how to use LLMS and we kind of have an idea of what we
- 6:02want to create. And so the answer that they're
- 6:06trying to solve here is not really for creatives and artists
- 6:10specifically, is for people who really don't have the time and
- 6:13the resources and the skills. So what are we seeing here?
- 6:18First, we're seeing that box where we get to tell the LLM
- 6:22what we want and we get to choose the LLM.
- 6:25So for instance, Fahad may prefer to use Claude.
- 6:28I may prefer to have ChatGPT do the job for me.
- 6:32But let's say we're going to go with Claude here because this is
- 6:35what they recommend. They use Clod in their own
- 6:37videos as well. We have 4 tabs of examples.
- 6:40Pride is going to help us. UGC content, marketing content,
- 6:44cinematic content and cartoon. Under UGC, you're going to see
- 6:48some of the examples I just talked about.
- 6:50You can make a use UGC, which is user generated content review of
- 6:54a product with a green haired creator.
- 6:56Imagine Farrad, you had towels that you were selling about 10
- 7:00years ago, right? You created these towels.
- 7:03You were selling them on Amazon. I was the UGC creator for the
- 7:05towels. You created the content yourself
- 7:08and you also paid influencers and send them the towels so that
- 7:13they would create content for you.
- 7:14That's called UGC content. Now imagine in 2026 if I were to
- 7:19recreate that towel again. The promise here is that you
- 7:22don't have to go ahead and look for all those influencers.
- 7:26You can just create your influencer and make the video
- 7:29and it will be as believable as the real deal.
- 7:32That's the claim. That's the example of the UGC
- 7:35content. Then there's the marketing.
- 7:36We're going to look at that really quickly.
- 7:37Pratt if you click it. Product photos So if you go to
- 7:42Amazon when you're buying your next shampoo.
- 7:45If you pay close attention, different photos front back in
- 7:48action, different type of. Photos they paid studios.
- 7:51To make that photos, yeah, to. Create those photos.
- 7:53We just spoke to Alex, one of our friends.
- 7:55So Alex has a studio. He was working with the skin
- 7:58care company, and he's taking photos in his studio.
- 8:01So he has to keep changing the backdrop and the lighting.
- 8:03And the whole promise here is that Alex can now just send the
- 8:07product photos here and create that product listing
- 8:09automatically. You get, you get the point,
- 8:12right? We wanted to put this to the
- 8:14test. So for anybody who has been
- 8:17watching some of the videos we've done in the past on this
- 8:19show, we told you guys we're working on an AI educational
- 8:22program in which we're teaching you how to use AI tools
- 8:25effectively in your career. For that AI educational program,
- 8:28we actually made a fictional productivity drink called Flow.
- 8:32It's a cognitive drink and we are manually creating all of the
- 8:37assets, the photos and the videos because we're going for
- 8:39that top tier quality content. And Pratt has done the job for
- 8:43us here by pulling up our Fig Jam link, which is the our
- 8:46entire research and production workflow, which has taken
- 8:52multiple weeks. And this is what we're teaching
- 8:54there, manually creating the content to really have the top
- 8:57tier results, stuff that you can put on social media and it will
- 9:00stand out. We wanted to put this to the
- 9:02test. So we said, OK, if we took this
- 9:04fictional can and didn't actually do any of this work and
- 9:08we said, OK, supercomputer Higgs field, you go and do the job for
- 9:11us. What can we actually get in the
- 9:13results? Pratt, if you go to the left
- 9:14side and pull up the most bottom chat, go all the way to the top
- 9:18for me, please. You're going to see my prompt.
- 9:21I gave it. Click on that image.
- 9:23For me. This is the can, the fictional
- 9:26can that we've. Created this is our input.
- 9:27Basically this is the input that we gave to Higgs field.
- 9:30Exactly. OK.
- 9:31And then we wrote a prompt and the prompt in summary says
- 9:35create 10 highly engaging and viral product photos for a
- 9:39marketplace listing for places like Amazon using this can.
- 9:43After that if you come down, you can see what's really important
- 9:46there is the skill. If you look at the skill, it
- 9:49says using skill product photo shoot.
- 9:52I'm showing you guys what's happening under the hood.
- 9:54This is the bit that you kind of have to do manually to achieve
- 9:57the great results that you want to achieve.
- 9:59But for an entrepreneur, a business owner who has no clue
- 10:02what Claude is or what skills are, doesn't want to spend the
- 10:05time. Heysfield has done that part for
- 10:07them. They've created a skill which is
- 10:09a set of pre built instructions inside Claude that tells Claude
- 10:15how to do a product photo shoot. Basically says OK, I want a
- 10:18studio, clean backdrop, all those things that you would know
- 10:21as a guy who does product photo shoots.
- 10:23They've written and out for Claude.
- 10:25As a prompt. As a prompt and it loads it up,
- 10:28then it starts asking you questions.
- 10:29So here it said what kind of photos you want?
- 10:32Do you want studio photos or lifestyle photos?
- 10:35And with multiple choice, I selected to the ones that I
- 10:37wanted the aspect ratio of my images.
- 10:40Depending on where you're using the images, it might be for
- 10:43Instagram. You want it to be vertical or
- 10:45horizontal. It's confirming those things
- 10:47with you. And after that it was done.
- 10:49If you come down, it created a set of photos for us.
- 10:54Now, Pratt, if you just click on the photos and we'll just go
- 10:56through some of them. Here.
- 11:03Yeah, these are the photos that are generated.
- 11:04You can zoom into one and we can take a look at it.
- 11:09This was done within 5 minutes. In my opinion.
- 11:15It's quite interesting. Yeah, quite good, usable.
- 11:18If you zoom even further to just check the quality of the text on
- 11:23the can, because I know what I've written on that can, it
- 11:26says early bird cognitive energy.
- 11:28Zoom in more prat. Let's see if the text occur.
- 11:30No zoom in, zoom in. I know he can't.
- 11:32Oh, OK, OK. It says cognitive energy, early
- 11:37bird go a little bit lower flow a bad decisions drink.
- 11:40Works. It's not that morphed.
- 11:44It looks pretty decent. Especially here when we have a
- 11:47close up of the can, it looks proper.
- 11:50It didn't make lots of mistakes, which means if again, I had no
- 11:55option to go out and hire someone out, I didn't have the
- 11:57money and the resources or the expertise to be able to achieve
- 12:02this within 5 minutes. That's good.
- 12:04It's pretty decent. This is where it's quite
- 12:06interesting. Let's take a look at some of the
- 12:08other photos on the right side. The only challenge I have with
- 12:12it is that it does look a little too AI generated.
- 12:19This photo specifically the first one, actually not that
- 12:22much. So if you zoom back, if I want
- 12:24to rate these photos, the 1st 2:00 on the left, they are quite
- 12:28good. You can actually put them up
- 12:30online. And by the way, remember, even
- 12:33if you're shooting real life, there are a lot of horrible
- 12:36product. Shoots 100% just like how you
- 12:38had your fair share of. Experience.
- 12:40So you can. You can even argue that if you
- 12:42give it to a graphic designer to use Photoshop to create that
- 12:45product photo shoot. Still, it can have flaws.
- 12:48These two are pretty good, I agree with you.
- 12:51The next two they give it off a little bit that they are AI
- 12:54generated and I'm not sure whether you want to use it on
- 12:57your listing or no. This is the challenge that I
- 13:00have and I wanted to share it with you because we want to be
- 13:02as honest as possible. That's why I was very careful
- 13:05with who I'm talking to when I'm presenting this new feature.
- 13:11If you're a creative, if you're the guy who's running the
- 13:14studio, if you're an artist, if you're a creator who wants full
- 13:18control over the quality and the output and care about standing
- 13:22out being the best. And also if you're serving
- 13:26people who have these businesses, then you want to
- 13:29have quality that really stands out.
- 13:31And this supercomputer is not for you.
- 13:34And I'll explain in a moment because we have proof of that.
- 13:37However, if you are a person with no experience, a
- 13:42programmer, A programmer with a new product or a new service, or
- 13:46you're selling dried fruits, or you're selling, I don't know,
- 13:50give me some other examples here.
- 13:51Anything. Clothes.
- 13:52Yeah. You're a business owner with 0
- 13:55creative skills, a 45 year old woman business owner who just
- 13:58opened up a new business. She's making jam at home, right?
- 14:04And you really don't have experience with Canva, with
- 14:07Photoshop, And there exists a lot of people like this.
- 14:10This is a fucking lifesaver in my opinion.
- 14:13Within 5 minutes to generate 10 product listings is insane and
- 14:17it's one prompt. The potential here is huge, but
- 14:21as someone who cares about quality, I want to show you what
- 14:25we've created when we've manually done this.
- 14:27So not going to supercomputer, but actually manually talking to
- 14:30an LLM like Claude and figuring out what the idea is beforehand.
- 14:34Using Pinterest for references, going very granular with the
- 14:38lighting and the art style in the photos, as Pratt is showing
- 14:43you right here. Spending a little bit more time,
- 14:46but getting the results that we precisely want.
- 14:49And in more a more complex scenario, right?
- 14:52So, so the the one that the the photo turned out really good.
- 14:55I would call that even if you want to shoot that in real life,
- 14:57those are simple scenarios. You can do it in one single day
- 15:00shooting. You don't need that many props.
- 15:02You don't need to change location.
- 15:04You can do those even in real life, shooting exactly the
- 15:07photos that you created. Pat, can you bring the one that
- 15:09Farah has created? Yes, these requires moving to a
- 15:12location if you want to do it in real life.
- 15:14These are difficult to shoot even in real life.
- 15:16You need a skilled photographer to shoot these photos.
- 15:19Exactly. The point that I want to make is
- 15:22exactly what Farah just said. When you are very precise about
- 15:27what you want, which is actually most of the time the case that
- 15:32happens to be true. As a creative and as an artist
- 15:35or a creator, you always want to have full control.
- 15:38That's one. But even if you're a business
- 15:41owner who cares about the detail, let's say someone like
- 15:43Steve Jobs, he was incredibly ruthless about how his ads were
- 15:47shot. And I imagine a lot of people in
- 15:49business are also like that. He may have not been the one
- 15:51using Photoshop or the editing software, but he made sure that
- 15:55everyone follows his vision. In that case, supercomputer is
- 15:59not the best choice because it's automating a lot of the
- 16:01workflow, which means it's doing 90% of the guessing that you
- 16:06don't have to do. And so when you're trying to
- 16:08create something very specific, and this is what we wanted to do
- 16:11with the cans inside the AI education program, because we're
- 16:13targeting people who care about the actual detail of every
- 16:17single thing they're creating, You come up with results like
- 16:20this. There is no morphing in the
- 16:21text. There's no random generations
- 16:24and everything is very specific. No, this looks.
- 16:29Like 5 digit or even 6 digits product shoot yes again, I I
- 16:34want to tell people something that even if you are using AI to
- 16:37generate your photos or video or you're doing a real life shoot
- 16:40or you're using 3D, there are tiers basic, premium and high
- 16:46end. You can have that in AI when you
- 16:48generate it and you can have it in real life and you can have it
- 16:50in 3D. Whether you are using AI3D or
- 16:54real life shoot, you want to be in that premium level because
- 16:57you can charge premium when you are in that premium level.
- 17:00That translates here as well. Or when it's your own product,
- 17:03then you're going to want to be. And that's what I want to get to
- 17:05here. If you want to stand out, you
- 17:08have to understand if supercomputer is widely
- 17:11available for everyone, go a little bit to the right.
- 17:14Pratt, please show this photo. Yes, this is the first one.
- 17:18Everyone's product on Amazon and on Instagram is going to look
- 17:21like this because everyone's going to use supercomputer.
- 17:23With minimum efforts. With with minimum effort, the
- 17:26person who's going to sell the most or be the most successful
- 17:30is going to be the one who actually spends time to learn
- 17:33the pipelines and not allow the AI to do the guesswork, but
- 17:37instead they do the guesswork themselves.
- 17:39They tell the AI to just do the job and work collaboratively
- 17:44with the AI. In my opinion, what we were
- 17:47showing you are simple use cases up until here.
- 17:51Again, as I want to be truly honest, I think supercomputer is
- 17:54incredibly helpful to a lot of people and massively large
- 18:00population. But the moment you start to take
- 18:04it to the complex zone of creating ads and imagery, it
- 18:09failed on us. At least we're going to show you
- 18:11our last chat. We had an idea that we're
- 18:13actually going to implement in the AI education program.
- 18:17We're creating full commercials. And for the full commercials,
- 18:20I've already created the storyboard with the guys.
- 18:22The ideas of what we wanted to do here in this specific ad, we
- 18:27wanted to have the flow cans as a POV experience of different
- 18:31elite performers in the world drinking the cans.
- 18:35So for instance, we wanted to have the president, we wanted to
- 18:38have a racer, we wanted to have a lawyer developer, and we're
- 18:42actually manually creating these images and they're working.
- 18:46But when we asked supercomputer to create them, you can see it
- 18:49understood the job. It created an E sports gamer
- 18:53image with the Flowcan. But the Flowcan is wrong.
- 18:56If you go to the right, it has the developer, the Flowcan is
- 18:59there, but the lighting is completely off and the photo
- 19:02again looks AI generated, which is something I would personally
- 19:05never use and I don't think anyone should by the way.
- 19:08Now can you fix this with multiple prompts?
- 19:10Perhaps yes. Is it easy to use?
- 19:12Yes, No, but something you the main concern cost.
- 19:16Thank you. The next concern that comes up
- 19:18is the cost. You have to understand now
- 19:21you're using Claude within Higgsfield, using your
- 19:24Higgsfield credits previously, or at least what we do at Bad
- 19:29Decisions. I spend my Hicksville credits
- 19:31purely for the image generation. I'll talk to Claude using my
- 19:34Claude subscription, which is going to cost a lot less because
- 19:37I'm directly dealing with Claude and I'm optimizing that for my
- 19:41job. And I'm not using a random skill
- 19:43that is not created by me. I'm using what I have created so
- 19:46I get exactly what I want at minimal cost and then I'll bring
- 19:49that when it's ready, when I've verified the prompts and my
- 19:52direction 2 weeks field and then through a few generations I get
- 19:56exactly what I want. But here, every single mistake
- 19:59that you make in this case, you're spending credits and it's
- 20:03expensive. Just for these two examples, so
- 20:06the ones that worked and the one that didn't, I spent 4% of my
- 20:09credit and we're on the more expensive plan and 4% of that is
- 20:12actually really large. So we were at 71.
- 20:15I think we're about 7067% now. Just for creating this. 2 chats.
- 20:202 chats because Clod is expensive.
- 20:22Using Clod Opus 4.7, it's incredibly costly if you use it
- 20:26through your Hicksfield credits. For me it doesn't make sense
- 20:29because I want to keep my Hicksville credits for the image
- 20:32and video generations. By the way, I haven't even gone
- 20:35to video yet. Oh yeah, this is just photo.
- 20:38See, Dance is fucking nuts. Expensive.
- 20:42I'm keeping all of my credits for my 60 seconds of footage,
- 20:46and I'll be using all of it. I think we're paying about $200
- 20:49a month, by the way, And I guarantee not everyone watching
- 20:52is on that plan. You can see how much.
- 20:55Yeah, See, that's if you go to, if you increase it to 15
- 20:58seconds, it's at 135. There you go.
- 21:02And by the way, when you generate a video, more often
- 21:05than not, the first video is not the one you're going to use.
- 21:08So imagine multiple of them. It's going to be a lot of money
- 21:12that you're going to spend. I'll be very careful using
- 21:15supercomputer. It is an interesting tool.
- 21:19I am a fan of Higgs field for one thing and one thing only.
- 21:24They're incredibly fast in this AI space and the marketing is
- 21:27great. Their UI is beautiful.
- 21:30I love their UI to be honest. The reason why I use them is
- 21:32because their UI is more beautiful than the other
- 21:36companies. But I'll be careful as you guys
- 21:40are watching this and I know the cost is very important to you
- 21:43guys. I'll be careful about spending
- 21:44money on the Claude Opus 4.7 here.
- 21:47If you already have a Claude subscription, if you already
- 21:50have a chat to be the subscription, you can just get
- 21:52your prompts there and understand your direction and
- 21:55manually do the work. It might take a little bit more
- 21:57time, and it will, but then you get the actual results that you
- 22:00can practically use. And ultimately, I wanted to talk
- 22:04about this. Practicality is the most
- 22:06important thing. Yes, this is a fancy tool.
- 22:09Yes, it's helpful, but will you actually upload this on your
- 22:14Instagram and actually sell your product?
- 22:16Will you upload this on X and actually get the views and the
- 22:20attention that you want? That's the most important
- 22:22question so far from our generations.
- 22:24I know the answer is no, but if you spend a good amount of money
- 22:29on it, I imagine you can get to it because again, you're using
- 22:32cloud outside. Eventually you can get to the
- 22:34same results, but here you'll be spending thousands of dollars.
- 22:37To get to To be fair, if you want to do a product shoot, it
- 22:39will cost you thousands of. Dollars it will, but I'm telling
- 22:42you there is a there is another way to not spend thousands.
- 22:45Of dollars if you learn how to basically use the AI because.
- 22:48There's two advantages to that. So the workflow we're teaching
- 22:51you in the AI educational program is the manual workflow.
- 22:55When I say manual, by the way, the guy is still doing most of
- 22:57the job for you. When I say manual as in you do
- 23:00the research by going to Pinterest and.
- 23:02Actually coming up with that orchestrating bro.
- 23:05Orchestrating Manual is a horrible name.
- 23:07Manual as in you decide what the output should look.
- 23:12Like you're a leader, you orchestrating.
- 23:14And I cannot imagine, again, like I said, I cannot imagine
- 23:18you not knowing what you want to create.
- 23:21Whether you're a business or a creative, ultimately I believe
- 23:25you should have control over what is being created.
- 23:27You can get recommendations. You should get recommendations,
- 23:31but you should ultimately be the one planning it out because you
- 23:34know what's best for your business and for your services.
- 23:37That's what we're teaching in this workflow.
- 23:39We're using very similar things. By the way, I'm not saying what
- 23:42Higsfield is doing is entirely different.
- 23:44Higgsfield essentially just connected Claude to Higgsfield,
- 23:47so that bit of copying and pasting prompts and talking to
- 23:50Claude is not automatically. But usually the Claude part
- 23:53rather than the GPT part, there is a lot of back and forth.
- 23:56It's not a single prompt you're getting.
- 23:58It maybe doesn't get it right in the first time.
- 24:00Maybe you need to get more ideas from it, maybe gives you a
- 24:02better idea and you want to expand on it.
- 24:04So I think if you keep your communication with the LLM
- 24:08separate, not only you will save cost, you will get a better
- 24:12outcome at the end of the day. Ultimately, before we move on,
- 24:16if you're a business or an entrepreneur with no creative
- 24:20skills and you don't care about learning them, I would recommend
- 24:23supercomputer. If you're a creative or a
- 24:26business or a creator who cares about the final outputs, quality
- 24:30and detail you like, you want to be granular with the detail,
- 24:33then I recommend that you learn how to do the entire
- 24:36orchestration yourself. That's what we're teaching at
- 24:40our a educational program. A lot of you have been asking
- 24:42where the link is. I'm going to say it again.
- 24:45We try to remind you guys on the show right here.
- 24:49If you go to AI dot bad decisions dot studio, there's a
- 24:52name field and an e-mail field. Put it in and then you'll join
- 24:55the wait list and get in as an early bird.
- 24:58You'll get advantages because of that.
- 25:00We'll be launching soon. I have promised soon.
- 25:03But guys, we're working on bringing you the highest quality
- 25:07education in the edits, the visuals, the script, bad
- 25:10decision style. Bad decision style.
- 25:12As you know it will take some time but will be launching soon.
- 25:16Yeah, it wouldn't be that far. June, July, we said May, but May
- 25:21is already 15th, but we're. Making sure the quality is top
- 25:25notch. Yes, for and we are adding some
- 25:27and by the way you guys saw the flow part.
- 25:29So that's gonna be, that's gonna be part of it.
- 25:31Are you ready for the next part? I'm always ready, man.
- 25:34I think before we get into this, I wanted to say thank you.
- 25:41To you for what? For being my brother.
- 25:44Oh, thank you so much. Before I start, I just want to
- 25:48say thank you to Pratt as well. Pratt, thank you.
- 25:52You're welcome. Thank you guys too.
- 25:55What a whole. Song and I just want to say
- 25:56thank you. Thank.
- 25:57You thank you guys for watching and commenting truly.
- 26:01Actually, if you read all the comments on YouTube, on Spotify,
- 26:04thank you so much for watching. Let us know what you guys think
- 26:07about Hazel's supercomputer. We're going to respond to the
- 26:10comments. I'm genuinely curious to see if
- 26:12you guys will be testing. It out or not.
- 26:13By the way, one of one of the audience commented that why we
- 26:16always say Claude and GPT and not Gemini.
- 26:18I actually reply to that comment.
- 26:20I want to say we use Gemini too. Gemini has its own benefits,
- 26:24especially when you want to do academic research.
- 26:26But when it comes to having the entire suite available,
- 26:30especially with coding these days, we only use cloud and GPT
- 26:34most of the time. But again, if you want to use
- 26:36Gemini, if you want to use Grok, they are great.
- 26:38By the way, I think Grok is also releasing Build, which is their
- 26:42coding platform. So that's something that we're
- 26:44going to cover probably in the next episode, so.
- 26:47By the way, may I interrupt for a moment?
- 26:49One of the new playbooks we're adding in the AI educational
- 26:52program is Understanding LLMS. We're breaking down every single
- 26:57LLM to the deepest details, not, oh, how to ask for a recipe, but
- 27:03how to use Clot properly. For instance, how do you upload
- 27:06a skill that Higgs Field has created for creating product
- 27:09photo shoots? Or alternatively, how do you
- 27:11create your own skills? What are skills?
- 27:14How do you train or fine tune your own model locally?
- 27:17All these things we're going to teach you inside the AI
- 27:20educational program. And the reason why I even
- 27:22brought this up is because, just like Farrad mentioned, Grok is
- 27:29also another LLM. We don't really talk about it,
- 27:32but we're using it all the time. Not for coding especially.
- 27:34We actually use crock for the show.
- 27:36Why? Because there's a specific
- 27:38function that is unique to grok when I'm looking at an X post
- 27:42and I want to make sure I Fact Check it before I tell you guys
- 27:46about it. There's a little grok icon
- 27:48beside every X post that actually helps you Fact Check it
- 27:52in real time. And it does the best job because
- 27:54it's checking it and cross referencing it across all the
- 27:58other posts. So depending on what your job
- 28:00is, you should be using different LLMS. For us, it
- 28:03happens. So that Grok is great for the
- 28:04show, but it's not great for coding.
- 28:07Clod is great for coding, but that's my opinion.
- 28:09Fad codecs. We're covering all of that in
- 28:12the AI educational program. But just telling you roughly
- 28:14here, it doesn't matter which one you use, just understand
- 28:18which is the right tool for the right job.
- 28:21Fados, it's called Ultimate AI Program.
- 28:24What did I say? AI Educational.
- 28:26Program Ultimate AI. Program Ultimate Always my bad.
- 28:29Ultimate Unreal Engine, Ultimate AI.
- 28:31My bad. So there's nothing wrong with
- 28:32Gemini. It's also like saying is Porsche
- 28:35better or BMW or Mercedes? Everyone has different opinions.
- 28:38They're all great. Porsche, they all make great
- 28:40cars. Porsche, But you're saying
- 28:41Porsche? I do believe that, but I'm sure
- 28:43some of our friends would say BMW.
- 28:45OK, you know what I'm saying. Fair, fair.
- 28:47People will have different opinions.
- 28:50Are we ready? Yes, I love the next topic.
- 28:52I love you. I love you too.
- 28:53Bro. Stop this.
- 28:56I think this new setup has made you miss me so much.
- 28:59Thank you. OK, let's go.
- 29:01OK, you can now turn. Let me think.
- 29:06You said I love you. I lost my train of.
- 29:08Thoughts. I'm gonna let Farah take his
- 29:10time, prepare for the next topic.
- 29:12I think just the sign of affection has interrupted.
- 29:16The distance? The distance?
- 29:17No. But let's go to it.
- 29:19You can now turn a single image into a fully interactive 3D
- 29:24environment, which you can move the objects within a few
- 29:28minutes. This is something that really
- 29:30blew my mind when I saw the demo.
- 29:33The way it works is that you will orchestrate a bunch of
- 29:36models together. The AI agents will look at the
- 29:40image, they identify different objects, then they will create
- 29:45an image from every single object, send it to a 3D
- 29:49generation model. So the 3D model with mesh will
- 29:52be generated because now we have the images of different objects,
- 29:56we can hide them from our environment.
- 29:59So we are left off with a static environment and the 3D objects
- 30:03within the scene. Then the AI agents will go and
- 30:05add Collider to the scene. And now you can import this
- 30:09interactive 3D scene into Blender, Unreal Engine, Unity,
- 30:13or any other 3D tools that you have.
- 30:15This is something that we've been waiting for for the longest
- 30:18time, but the reason that we are here now is because all of the
- 30:22different AI models, the 3D generation, the image generation
- 30:25are now so good, even the agentic AI that you can spawn
- 30:29multiple agents, do different tasks at the same time.
- 30:33Now we are here at this stage and if you look at the demo, the
- 30:35guy posts a photo of a street. Go back and just pause on that
- 30:39frame so we can describe a little bit back, a little bit
- 30:43more. Yes, this is perfect.
- 30:44Back, back, back a little bit before they lift the chair.
- 30:47That was good. That was good.
- 30:49OK, OK. I want to see a comparison to
- 30:51the. There's yes, this one.
- 30:55OK, so the image is at the bottom right corner of the
- 30:59screen, right? So it's image of an alley.
- 31:01You can see there's a table, there are chairs, there's one
- 31:03umbrella, there is a few other things on the left side.
- 31:06The first thing that the AI address will do, they will go
- 31:08and describe the scene. They will name the object.
- 31:11They will create a list. And it's done using clod.
- 31:13Yes, but you can change them. The best part about it is
- 31:15there's a GitHub repo you can go and use different generation
- 31:18models for this. Has proven to be the smartest
- 31:21for these kind of jobs. I'm noticing a lot of companies
- 31:26by default use CLOD for these things.
- 31:28Higgs field being one of them. Even by default it's on clod
- 31:31when you go into supercomputer. In their own videos they're
- 31:34using CLOD for their skills and here in this video as well I
- 31:38believe they were using and talking about CLOD.
- 31:41So by the way, shout out to Word Labs.
- 31:43I explained the entire thing without even telling that this
- 31:46is done by one of the team members in Word Labs.
- 31:49The entire orchestration of putting different parts
- 31:51together, identifying, turning into photo, removing the objects
- 31:55and then turning them into 3D. By the way, the environment will
- 31:57be 3D Gaussian splat at the end, very lightweight.
- 32:01You can go into it, you can design different things.
- 32:03I can think about a lot of applications because we tested
- 32:08with 3D Gaussian splat a lot. Imagine a scene from the movie.
- 32:12You can do it with that. Imagine taking this into VR from
- 32:15a photo. Take it to the VR.
- 32:17There are a lot of use cases for this, and I think this is super,
- 32:20super exciting. What do you think, Faras?
- 32:23I was. Fascinated by the way the
- 32:25process works? As Fara mentioned, so many
- 32:31different technologies had to improve simultaneously to make
- 32:37this possible. The original image being
- 32:40described perfectly. That means Claude and LMS had to
- 32:45have incredible vision models and then be able to translate
- 32:49that into text. Furthermore, once it is
- 32:53described, then we had to be able to have enough quality in
- 32:58our 3D generation tools such as Tripor Trellis to be able to
- 33:02create the 3D meshes as these objects have been discovered by
- 33:05the LLMS. And that has to be done relatively quickly as well,
- 33:10with the material generation and the textures and the UV
- 33:13unwrapping and so on. And there's so much going on,
- 33:16the fact that people look at it and it's not in minutes, but
- 33:19there's so much work in the back end.
- 33:21And exactly. And then for the LLM to
- 33:26recognize what the objects are against the background, against
- 33:30the environment, and then try to isolate those so that they can
- 33:35create a new prompt to edit that image with the objects now
- 33:39removed using probably Nano Banana Pro or GBT image 2 point
- 33:45O. Again, Nano Banana Pro and GBT
- 33:48image 2 point O are relatively new.
- 33:49Before these models, being able to edit an image really quickly
- 33:54and consistently and accurately was almost impossible.
- 33:57But because the image models have also improved, you get to
- 34:01have that environment without the chair now slapped onto the
- 34:05texture. You know the environment of the
- 34:06walls are not cleaned up. And you know, having the
- 34:09Collider, remember when we first started working with 3D Gaussian
- 34:12splat, was it 2 years ago? I think it's been 2 years.
- 34:15Viewing the colliders had problem, you would go through
- 34:18the wall. Now this is perfect, you can see
- 34:21that. Holes in the.
- 34:22Environment. You identify every single corner
- 34:25of the room and it's 360°. The photo is just one angle.
- 34:29Now the guy goes and turn around.
- 34:30It generates everything. The more I look at this, we are
- 34:34living in a simulation. If in 2020, what are we six?
- 34:38Yes. If 2026, we have this technology
- 34:40by 2036. This is so advanced that I can
- 34:44put you in an environment, a fictional environment.
- 34:47You won't even realize it's fictional.
- 34:49This is a simulation matrix. Wake up if you have Neo haptic.
- 34:53Sensors and so on and so forth to be able to sense the
- 34:56environment as well. Why haptic sensors?
- 34:59You're gonna have brain chips. Oh my God, back to the
- 35:02conversation of putting something in our brain.
- 35:06But this is very exciting. I'm actually.
- 35:07Thinking about virtual production.
- 35:10OK. Yeah, virtual production.
- 35:12Yeah, actually, yeah, because you won't think about it.
- 35:14You won't think about it, yes. I want to create an ad or a
- 35:19movie shot in an alley like this.
- 35:23There's a lot of objects. I want to be able to move the
- 35:25objects. The director is going to ask you
- 35:27that was the. Biggest problem in Gaussian
- 35:29split initially true. So they'd buy assets online, of
- 35:33course, or they'd have to manually create them.
- 35:36And the biggest problem is when there is virtual production
- 35:40happening. This is some extra knowledge for
- 35:41you if you don't know, during the shoots, there's always CG
- 35:45artists on standby to make changes and modify the scenes
- 35:49depending on the director's vision, because he might have
- 35:52planned something in the storyboard but wants to change
- 35:54it last minute. Let's say when they're doing The
- 35:56Mandalorian, there's a massive rock, and for whatever reason
- 36:00that rock has to be moved. If the assets and the
- 36:04environment were created like this, they technically have
- 36:07control to move things around. Now they can't even add.
- 36:09They can. Add they can they can test.
- 36:12Oh, actually testing. Testing magic.
- 36:14I want to add a Blade Runner car right in that alley and I want
- 36:17to see. And how fast they can iterate on
- 36:19different scenes. It's insane.
- 36:21I think for virtual production this is going to be really
- 36:23interesting. Actually, we should follow us.
- 36:25Give it a try. It's a GitHub report, right?
- 36:26They haven't brought it into Word Labs yet.
- 36:29It's open. Source I believe.
- 36:30Yeah. So you can, you can go on
- 36:31GitHub. Yes, it's there.
- 36:33We should definitely give it a try.
- 36:36It's. Created by Nielsen.
- 36:38Yeah, shout out. We really need to give him a
- 36:41shout out. This is amazing.
- 36:43This is amazing. Let us know what you guys think
- 36:45about this. Are you gonna try it out or not?
- 36:46Since it's open source, you can go in and try it out today and
- 36:50let us know what you guys think about it.
- 36:52But for the next part, as you know, codecs desktop app can now
- 36:58do almost everything, coding using your browser, using your
- 37:02computer, making videos. Now they brought codecs into
- 37:06Chat GBT's mobile app, so you can do everything from anywhere.
- 37:10That's the promise. When they launched Codecs in
- 37:13ChatGPT mobile app, which is something similar to what Claude
- 37:17did a few months ago, they called it.
- 37:18I don't. Understand.
- 37:19So what can you do that you couldn't do before?
- 37:22OK, let me tell you this. Now you're running codecs on
- 37:25your computer and you're leaving the studio.
- 37:27You can leave the studio, leave your computer on, and then when
- 37:31the task is finished, you can keep the conversation and keep
- 37:34talking. And you can say, hey, for
- 37:37example, you made this UI for me, I don't like the color,
- 37:39change the color. There is also better use cases
- 37:42which they use. For example, you have a video, a
- 37:44presentation and you're going to the office.
- 37:47You will say that hey, go to that presentation, change the
- 37:49title or change the budget. You just realize client has more
- 37:52budget. Instead of charging them 100K,
- 37:54you want to charge them 150K. Ask codecs.
- 37:57Hey, change the PDF because it's not only about coding, it has
- 38:01already have. This.
- 38:02Called dispatch. So this is the ChatGPT version I
- 38:05see. And the funniest thing that they
- 38:06did you know open AI has a developer account on X.
- 38:10So they showed all this half turned on laptop in different
- 38:14environment that shows that now you can have agents everything
- 38:17in every single place. Pat, can you bring open AI
- 38:20developer X accounts? So yeah, these are all open
- 38:22laptops that are running so you are controlling them using your
- 38:26phone. The last one is on a toilet
- 38:30seat. You know, for me, as someone
- 38:34who's been watching you for the past few months, working with
- 38:39codecs and cloud code every day, one of the biggest challenges is
- 38:43getting you off the fucking computer and asking you to leave
- 38:46to a restaurant. Because you're always waiting to
- 38:50confirm something before we go. And because when it's
- 38:53processing, you cannot really talk to it and you need to wait
- 38:57for the results. So we always either 1020 minutes
- 39:00late and I get frustrated. And when we're outside, you're
- 39:03always thinking about the things that you could have done before
- 39:05we left. What this solves is that you can
- 39:08leave and continue working on the tasks even if you're away.
- 39:12Yes, that's correct. And I think even for my UK,
- 39:15which means you have to do more. You get to do.
- 39:17More, I mean, Sam Altman posted something which is a little bit
- 39:20of course, marketing. He said, oh, I was outside
- 39:23running under sunshine and Codex was doing my work and I was
- 39:26playing with my kids. Of course he has to say it for
- 39:28him. What's?
- 39:29Team making with. Codex Yeah, what is what is no
- 39:33codex guys, guys, remember it's not only coding if you wanna
- 39:37make presentation, is he? Making a presentation.
- 39:40I don't know, making reports, he's doing something.
- 39:43Brainstorming the reports. I don't know what he's doing,
- 39:45but he said he's using it. I would say honestly I haven't
- 39:50used dispatch because usually when I work with Claude and GPT
- 39:55the responses that I get are like 3 paragraphs.
- 39:58Reading them on the phone is difficult but for people who are
- 40:00doing smaller tasks I think this is really really useful.
- 40:03Maybe finally I will download this.
- 40:05I mean I have the app. I will give it a try and see how
- 40:09it works. I mean, for me, I used to use
- 40:12open claw before actually with you guys, right?
- 40:15If you're going outside, we are having something lunch or
- 40:17something and I will be texting on the phone and like can you
- 40:21open this file or can you render it or is it done?
- 40:23So for me it's really helpful. Actually, maybe Pratt, when we
- 40:27are leaving to Malaysia, we can leave the computers on and Ali
- 40:32comes here quite frequently, so he just checks the computers are
- 40:35on, we just leave them. On, you don't need to check if
- 40:38they're on. No, I mean you'll figure out if
- 40:40they're not on. Yeah, yeah, yeah.
- 40:41When your codecs app doesn't work.
- 40:43But what like I said, what I find truly helpful is I know how
- 40:49frustrating it is when we leave and we're halfway through doing
- 40:53something. And technically, since you're
- 40:55not completing the task yourself, the AI is doing the
- 40:58job, you get to leave and still continue the tasks.
- 41:02But what a horrible way to deal with socializing.
- 41:06Because even now, Oh. Yes, somebody comes.
- 41:09Outside, you'll constantly be looking at your phone, talking
- 41:13to your AI agent, codecs and cloud dispatch because you're
- 41:16getting shit done even when you're.
- 41:18Outside like you baby, it's the biggest problem.
- 41:20I have with it. We are in such a productivity
- 41:23health conscious boom right now that everyone has to be 100%
- 41:27locked. In no bro, it's no no, it's
- 41:30someone said we are going to be slaves to our devices now,
- 41:33slaves to AII Pray go to the demo.
- 41:36Please go to the demo demo video.
- 41:39Demo video. Yeah, go to the comments.
- 41:41Somebody made a beautiful meme about what you just said.
- 41:45Was it there? Was it here or in the No, I
- 41:50think it's in the open AI developer one.
- 41:52Go to the open AI developer one. Exactly how it was just said
- 41:55that we are going to be slave to Claude and codecs just to do the
- 41:59test. But seriously, think about that
- 42:05guys. That's true.
- 42:06Think about it. That's true.
- 42:08I mean, it's going to happen. I just wait.
- 42:10Let me just show you something funny.
- 42:11While you pull that up, I'm going to be very honest about
- 42:14where I stand with all of this. You're a slave to AI by the way,
- 42:16before you. No, seriously.
- 42:19Not joking. As you may already be aware, we
- 42:22are in an age where everyone has started becoming health
- 42:26conscious. Many of my friends wear whoops.
- 42:31Oh my God. Come on.
- 42:33Many of my friends wear whoops to track their daily activities
- 42:39and their sleep. I'm not a huge fan of Whoop and
- 42:42I'll explain why in a moment. Many people have started having
- 42:46peptides in Silicon Valley. It's becoming the new drug.
- 42:48They stop drinking alcohol and going out and socializing.
- 42:53They're taking peptides and working even more.
- 42:57And now with this, being able to work anywhere and all the time,
- 43:02because technically you can have unlimited ideas and implement
- 43:06all of those ideas with your AI agents.
- 43:08And now that the AI agents are with you and your phone, you can
- 43:11work all the time. So you probably will feel guilty
- 43:14if you're not working when you're outside.
- 43:18On a surface level, everything I just mentioned is great for
- 43:22humanity. The fact that we can have Whoop
- 43:24track our fitness and our food and our diet and our sleep so
- 43:29that we can optimize our life is great.
- 43:32The fact that we can take peptides and be able to
- 43:35regenerate muscle and our cells is great.
- 43:39The fact that we can have a agents do our tasks for us even
- 43:43remotely is great on a surface level.
- 43:45But the deepest problem I find with all of them is that if you
- 43:50don't know how to control yourself and don't have
- 43:53discipline as a human being, you're going to eventually
- 43:55become a slave to your Whoop. You're going to become a slave
- 43:58to productivity. But this TV froze.
- 44:01Brad, be careful taking. Are you taking screenshots?
- 44:03Yeah. Yeah.
- 44:04I just wanted to make sure. I thought our recording went
- 44:07went off. OK.
- 44:09Why did I say that? You're going to become slaves to
- 44:12productivity. I'm a fan of working hard.
- 44:16You guys know we produce content at a rate that is almost
- 44:20unbelievable. We are working pretty much all
- 44:22the time. I am fan of working and being
- 44:26productive, but I have a feeling that a lot of people will have
- 44:30problems balancing their lifestyles when we're going to
- 44:36have AI agents connected to us all the time everywhere.
- 44:40What I've seen, for instance, with WOOP and with peptides and
- 44:43what I've heard from research. Put them in the same category
- 44:45bro. I'm comparing, comparing for a
- 44:47moment, whoop. We agree for a moment with WOOP.
- 44:50What I've seen is I've seen the placebo effect on my friends
- 44:53when they wake up after a night of having lack of sleep.
- 44:58We've literally slept in the same house and we wake up the
- 45:02same exact amount of sleep. We feel great.
- 45:06They feel great. The moment they look at the
- 45:08whoop and the whoop says 55% recovery, you didn't have a good
- 45:11sleep. They suddenly feel tired because
- 45:13of the placebo effect of what they just saw.
- 45:16The peptide example is not specifically peptide, but
- 45:19generally the Brian Johnson style of living.
- 45:23I use peptide as an example. It's not actually peptides.
- 45:25What I'm talking about is being super health conscious.
- 45:28Like let's follow the perfect diet and follow the perfect
- 45:31sleep. What I saw is there's a lot of
- 45:36cases of humans becoming depressed because of lack of
- 45:42socializing, of course, because of the fact that everyone's
- 45:45become super health conscious and.
- 45:47I'm not going to drink, I'm not going to smoke.
- 45:49I'm not going to go out. I'm not.
- 45:50Going to go out, I'm going to be perfectly healthy.
- 45:53I'm a fan of being perfectly healthy.
- 45:54I'm on a diet. I take care of my health.
- 45:57But when you go, when you go to the extreme end.
- 46:00The Paul Saladino end. Yeah, don't, don't eat this.
- 46:03Don't eat that. Everything.
- 46:05Becomes harmful. You cannot enjoy a donut.
- 46:07You cannot have an ice cream. You cannot enjoy anything.
- 46:10Thank you, Brad, for that. What?
- 46:12What is that? There's a laptop in the
- 46:15microwave. Why?
- 46:17Cooking. Kodak.
- 46:19Cooking. That's the problem.
- 46:21I have when people go to the. Extreme end the same thing is
- 46:25happening also with AI. We we see the same problem.
- 46:27There are people who are like lock in if you're not using
- 46:30cloud code and you don't have 100 Mac minis running your.
- 46:33Business on. I was against buying Mac mini
- 46:36from day one. Didn't buy it, luckily didn't
- 46:39buy it. No point of buying it.
- 46:41I don't know why you should. What?
- 46:42What I'm trying to say is I understand what you mentioned,
- 46:46what Sam Altman said. Again, I understand they have to
- 46:49do marketing for the products everyone does.
- 46:52I totally understand. But on a practical level, as we
- 46:55always like to discuss it on the show, I understand being in the
- 46:59sunshine and having codecs do things for you is great.
- 47:03But I actually don't think he made a successful company of
- 47:06open AI by running things remotely on a phone.
- 47:09The fact that he's successfully in Open AI is because he put in
- 47:12a lot of hard work I think, making it sound like everyone
- 47:15can just run AI agents to do all of their businesses for them, I
- 47:19don't think. That's the case, I think.
- 47:20You need skills, you need hard work, and you need a break.
- 47:24And the last thing I want is having lunch with my brother and
- 47:28he's on his phone talking to codecs.
- 47:29That's why actually I don't download it on my phone.
- 47:31You know what I'm saying, It just could you for that one
- 47:34specific use case. You're running late for a
- 47:36meeting. You need to fix your
- 47:37presentation, yes, but if you're at lunch and you're talking to
- 47:41codecs the entire time. I think no.
- 47:43It connects with our next topic. Bring the robot, put codecs on
- 47:46robot, let's socialize with the robot and see what's up.
- 47:49Then you keep socializing with robots.
- 47:51Yeah, why not? Actually, I was thinking on this
- 47:53show, the moment we will have a good enough technology, the
- 47:58robot would be sitting there. I mean, you guys cannot see it
- 48:00in my space. No, no, not your space.
- 48:02Between us. Right Between us.
- 48:04No. You just need the full screen.
- 48:05Give me the full screen. Give me the full screen.
- 48:07I'll put the robot here. Yeah, right beside me.
- 48:10Are you? Hugging the robot is the robot
- 48:12sitting on your lap. No, no, no.
- 48:14Just beside me. He's standing up.
- 48:15He won't get tired. Why not she?
- 48:17It's. Are you gay?
- 48:19No, why would I be? You just said.
- 48:21Even if I was, why would I need a robot?
- 48:23The. Robot is sitting.
- 48:24On your lap it's a robot standing up.
- 48:27But next topic is also about robots, so all.
- 48:30Right. My God that's a huge robot
- 48:35though. Thank you.
- 48:38Shall we move on to the next topic?
- 48:41Sure. Today marks one of the most
- 48:44important days in robotics. As we speak, a team of humanoid
- 48:50robots have been live on YouTube for the past 40-5 hours doing a
- 48:55factory shift fully autonomous work with no robot failures for
- 49:0145 hours. This is the longest unedited
- 49:06footage and proof of humanoid robots working, where we get to
- 49:14judge their performance and measure the practicality of
- 49:18having robots doing labor. The task is simple package
- 49:23sorting at a factory line. They're standing at a conveyor
- 49:26belt, and they have to ensure that the barcode of the packages
- 49:30sits face down so that it can be scanned as it moves across the
- 49:35line. They've done 56.
- 49:38Well, actually, I'll rephrase the numbers because what we're
- 49:41watching right now. Is the actual.
- 49:43Live stream, 46 hours, 16 minutes.
- 49:47Over 57,000 packages have been managed.
- 49:53It got stuck. Wait, you're watching live,
- 49:55right? Actually, it got stuck.
- 49:57It got stuck. Don't fall down.
- 50:00Don't fall down. Can I tell you guys, Can I tell
- 50:03you guys what is happening right now?
- 50:05OK. No, it's back.
- 50:06It's back. I want to talk about this.
- 50:08His name is Rose. This is very important to talk
- 50:10about. This company is called Figure.
- 50:13The robot we're looking at is Figure 3, and it's running a
- 50:17neural network, a proprietary model created by Figure called
- 50:22Helix 2. Every task that you see right
- 50:25now is being done fully autonomously.
- 50:28Unlike some of the other robots that are teleoperated, there's a
- 50:31human operating operated. This is fully autonomous.
- 50:35The robots are reasoning directly from the camera pixels
- 50:39that they have, and they're acting autonomously.
- 50:43What's really cool is that they were supposed to do only an 8
- 50:47hour long shift, but because the robots had no failure, they
- 50:51continued it until now, which is 46 hours.
- 50:55When I say robot failure, I'm talking about them stopping
- 50:58working and not being able to continue the job.
- 51:01Like out of battery or non-technical failure?
- 51:03No, no, no technical. Failure they have.
- 51:05So we actually have had multiple robots do the job.
- 51:08They switch places and how it works is quite fascinating.
- 51:12Like taking shifts. Exactly.
- 51:14The robots switch off to charge when their batteries get low.
- 51:18So right before the batteries become super low, they message
- 51:21each other and an autonomous swap happens between the robots.
- 51:26So another robot comes and immediately takes over for the
- 51:29package sorting. That way there's no downtime.
- 51:34So imagine if you are running this factory and you want to
- 51:37ensure enough products are being made for humans.
- 51:40The robots will continue to shift their spots.
- 51:45So it's 24/7 on. These are the same robot that we
- 51:47covered in the previous episode. They cleaned the bedroom.
- 51:50These are Helix 2.0, right? And they do the bed.
- 51:52Same. They do the bed.
- 51:53Yeah, they're exactly the same. 24/7 365 the robots, I think we
- 51:58just saw it right Pad. That was the one you just showed
- 52:00us. Now why did I say no robot
- 52:04failure? I was checking the founders
- 52:06account on X. He said that.
- 52:09Brett. No, I I know it's Brett, but I
- 52:13forgot the last. Name Brett or Brett, Brett T or
- 52:15DT. If one of them ever has a
- 52:20software or hardware issue, it will autonomously leave for
- 52:24maintenance and another one steps in.
- 52:27Brett Adcock Now I wanted to talk about the average time that
- 52:36it takes a human to do package sorting.
- 52:40Per package, a human takes an average of 3 seconds.
- 52:43The robots are now at human parity.
- 52:46That means the robots also take about 3 seconds, just like
- 52:50humans. That's.
- 52:51Very interesting to do this job and honestly I think this is so
- 52:55good because I have respect for, I have respect for people who
- 53:00will do these kind of jobs, but I think those humans can
- 53:04actually find a bigger purpose and can do things that actually
- 53:08worth way more for the society. We're not born into this world
- 53:12to do this job. That's what you're saying.
- 53:13Exactly. So now when robots can do it.
- 53:16But what if the robot? The fact that there's another
- 53:18robot just walking behind me. Is what if the robots get so
- 53:21intelligent at 1.5 years from now and tells me that hey, I can
- 53:25do something much better? Why would I do this?
- 53:28Someone, you human. You human, I can do.
- 53:32No, No. What would it do?
- 53:33No. You but right now.
- 53:35No, no, five years. Again.
- 53:36Imagine this five years time, the robot will say.
- 53:40I can do so much more. I can build the factory.
- 53:43No, because the robots can make robots that are much faster than
- 53:49humans. Sure.
- 53:51So they will make robots that can do the job, if what you're
- 53:54saying is. True, they will become redundant
- 53:57and irrelevant, No. For this job, yes.
- 54:01For this job, for this job, what we're showing here, the reason
- 54:05why I said it's for any job, the reason why I said this is one of
- 54:09the most important days for robotics is because up until
- 54:13now, all we see is launch videos and demo videos that are edited.
- 54:17This is the first time we get to see a live stream. 46 hours of
- 54:21robots doing human labor, something that humans have to do
- 54:26at a factory, a robot can now do with no downtime because they
- 54:29just shift places. They don't need to sleep, they
- 54:31just recharge. And eventually the batteries
- 54:33will become so strong, they'll probably have fast charging.
- 54:36And within 30 minutes or 10 minutes, they can come back to
- 54:39their shift. And we'll eventually have so
- 54:40many robots that they can do all of these jobs.
- 54:43Now, in the beginning, there's going to be massive displacement
- 54:48of human labor. Human labour will be redefined.
- 54:52There is no denying that. And there's no way you can stop
- 54:55it, because even up until this moment, who gets to do these
- 54:59jobs? Usually the human who is faster
- 55:03will be rewarded more. The human who is better at this
- 55:05job will probably be rewarded more.
- 55:07In that company, in that factory line, the robots are being made
- 55:11so that they can take over because they'll be eventually
- 55:13more efficient than humans. Now, do I sympathize with the
- 55:20fact that we're going to have massive job losses?
- 55:23Yes, of course. But the only thing that you can
- 55:27think about is where we born, where we made as humans to do a
- 55:34redundant job of looking at packages and turning them upside
- 55:38down for our entire lifetimes. I don't think so.
- 55:41I feel like we have greater purposes.
- 55:44Unfortunately though, for so many people who have to do jobs
- 55:48like these, who may be unhappy with their jobs because it's not
- 55:51fulfilling, they never had another option.
- 55:55Now, will there be an option available to them immediately if
- 55:59they're let go because a robot will come and take over their
- 56:01job? Probably not.
- 56:02I don't think there will. Be I think in the future, in the
- 56:05next decade, the shift that you mentioned will happen the.
- 56:09Redefinition. Not a redefinition.
- 56:11That's not a word. Redefining of Labor.
- 56:14Definition is a. Word redefinition, I don't think
- 56:17is a word redefining of human labor.
- 56:19We have to figure out what new jobs it's a word, the act or.
- 56:25Process of defining something again and differently.
- 56:28Redefinition. So it's a word.
- 56:30There you go. Thank you, thank you.
- 56:32What I'm saying is this, we are going to have a displacement of
- 56:36jobs of human labor, especially human labor that is going to be
- 56:42easily replicable by robot, in this case, literally turning a
- 56:47package upside down. Up until now, this has been
- 56:49complex for robots to do because they need to understand how to
- 56:52handle the packages. You have to be very careful.
- 56:55Different kind of packages you have boxes.
- 56:57The machines that we've had up until now are not as deliberate
- 57:01as these machines to be able to hold the packages and move them
- 57:04without breaking the content within it.
- 57:07These packages could be eventually the packages we
- 57:10received from FedEx and DHL and you have to handle these
- 57:14packages with care and only humans could do it because of
- 57:17the fact that we have these fingers and senses.
- 57:19Is there an instruction for the label to be face down?
- 57:23Just pay attention. There are sometimes that the you
- 57:26see right, You can just put that that package on the conveyor
- 57:30belt, but every single time it makes sure that the label is
- 57:34facing down. I just said that's what the role
- 57:36is. Oh, really?
- 57:37I just said that in the beginning.
- 57:39Thank you for paying attention. Oh wow, The whole point is there
- 57:43is a scanner that scans the bar code when the packages.
- 57:48That's what I'm saying. When the packages are sent in
- 57:51through the conveyor belt, a human has to stand the entire
- 57:55day and just change the orientation so that the packages
- 57:59can be scanned properly. Makes sense.
- 58:01Imagine DHL. Yeah, it all.
- 58:03Makes sense. Imagine FedEx international
- 58:05delivery. Is this a thank you Pratt?
- 58:09Thank you. There are actual humans who have
- 58:11to do the job, waiting at a conveyor belt, switching the
- 58:15orientation of the packages so that they're facing down and the
- 58:18scanners can properly scan them. This is just one tiny example,
- 58:23so many in the car, thousands of jobs like this, which is why I
- 58:28think it's important to discuss this because when you look at
- 58:33one example, you have to understand and try to connect
- 58:35the dots and say these robots will eventually be trained on
- 58:38many other tasks. And what I want to make really
- 58:45clear here is I'm not saying I don't have emotions for this
- 58:49event that I'm seeing before me. As exciting as it is to see a
- 58:53robot handle packages like a human and do it autonomously and
- 58:57figure things out, I think it's fucking cool.
- 58:59As someone who loves sci-fi movies, this is something that
- 59:02we used to watch in films and now it's a reality.
- 59:06But it is scary because how many people are doing these jobs?
- 59:10But bro it's happening again, sorry about your emotion.
- 59:14Wouldn't even take into consideration your job, which as
- 59:19a 3D artist is now also being done by AI, is the exact same
- 59:23thing. It's the exact same thing.
- 59:25You mean the Super computer? Yes.
- 59:26Of Higgs field. I'm not super computer.
- 59:28The AI generation could that generates A-frame when you were
- 59:32sitting down behind the computer and generate A-frame using a 3D
- 59:35software, that's the exact same thing.
- 59:37Fair enough. Keep your emotion for yourself,
- 59:39Saruman. No, what I'm saying is.
- 59:41No, but you're right what I'm saying, you're right what you
- 59:45feel. If, let's say I'm not, you were
- 59:47working in that factory and you just lost your job.
- 59:51Yes, because figure robots came around and took everyone's jobs
- 59:55in your factory. I would feel sad.
- 59:58Yeah. Because you just lost your job.
- 59:59Yes. And I don't have an immediate
- 1:00:01replacement for you. Sure.
- 1:00:03The point I'm trying to make is it's an exciting technology and
- 1:00:07undeniably, whether we like it or not, it's happening.
- 1:00:11Yeah, and it's going to be a rapid shift.
- 1:00:14What I'm saying is the exciting outcome that we will get
- 1:00:19eventually. I don't know when.
- 1:00:22Five years, 10 years, 15 years is that we will have to redefine
- 1:00:27human. Labor.
- 1:00:29No, that's for sure. That's the reason.
- 1:00:31Why this person, when he was a person, was getting paid is
- 1:00:35because he provided a specific value by doing this job.
- 1:00:39That value is no longer required of him.
- 1:00:43He need no longer needs to provide that value because a
- 1:00:46robot can do it and can do it more efficiently.
- 1:00:48So that human needs to create value for society in a different
- 1:00:51way and get compensated for that.
- 1:00:55But the only challenge is when robots come and become
- 1:00:59mainstream, a massive amount of different jobs will suddenly
- 1:01:03disappear. Yeah, that's why you need.
- 1:01:06To suddenly figure out a way to. That's why people really start
- 1:01:09now, That's why you have to start now to upskill yourself,
- 1:01:13learn about all these new tools and technology to see what
- 1:01:16position you will you will have in the future.
- 1:01:19And if you're watching this show, you're ahead because you
- 1:01:22know about all these different technology, whether it's coding,
- 1:01:24whether it is art or even robotics.
- 1:01:28I think if you're not paying attention to any of these
- 1:01:31technologies, again, this is robotics is a specific job, but
- 1:01:34I give you the example of what happened to our own industry
- 1:01:37recently. Yes, it has happened to us too.
- 1:01:39It happened. It happens to every single
- 1:01:41industry. The only way is to up skill
- 1:01:44yourself and get updated with the latest.
- 1:01:47I'm talking more on the what you said on an individual level
- 1:01:50makes perfect sense and what everyone should do is what you
- 1:01:52just said. I'm thinking more as a Mac
- 1:01:57micro, no macro macro effect, more theoretical overview of
- 1:02:05what can happen. We have to somehow, as humans,
- 1:02:12as the government, as. Individuals.
- 1:02:15But you don't excite universal basic income.
- 1:02:18You said you don't like it. No, that's the only way.
- 1:02:20Yeah, but it's gonna happen. Like for example, I think Tesla,
- 1:02:23their factories, it's automated, right?
- 1:02:25It's already automated. Yeah, a lot of the work.
- 1:02:28And Xiaomi, like their phones and their cars, are fully
- 1:02:30automated. They're robots.
- 1:02:31So we went all We are already there.
- 1:02:34Again, these are great. No, but I'm talking about a
- 1:02:37massive shift. No, but it's not gonna be
- 1:02:40suddenly the massive shift. It's not gonna be suddenly.
- 1:02:43It's gonna be factory by factory, worker by worker,
- 1:02:45country by country. Usually it's first states in
- 1:02:48China. But wouldn't you say in the next
- 1:02:4910 years? Sure, yeah.
- 1:02:51But in 10 years? Suddenly much faster, right?
- 1:02:53Exactly. Like but the amount of people
- 1:02:56who will be let go. That's why universal basic
- 1:02:59income. But every government has to come
- 1:03:02up with universal basic income but.
- 1:03:03Not every country will have the massive ship.
- 1:03:06China is going to send robots to every.
- 1:03:08Country, I think China. I am actually not familiar with
- 1:03:10Chinese politics, but I think they kind of have something
- 1:03:13similar already. Yes.
- 1:03:16But what I'm saying is far the reason why this figure news is
- 1:03:20exciting is because it's the first time you're seeing robots
- 1:03:23to 46 hours of autonomous work. Humanoid robots.
- 1:03:26I'm not talking about machinery with arms, humanoid robots.
- 1:03:29This is very exciting. These humanoid robots will
- 1:03:32eventually be trained on many different tasks.
- 1:03:36Lawn mowing Lawn mowing How many people make money by lawn?
- 1:03:40Mowing how many college kids make.
- 1:03:43Money by lawn mowing. Do you understand what I'm
- 1:03:46saying? Car washing.
- 1:03:48Gone, gone, No, honestly, but all of.
- 1:03:52These jobs, but at the same time, may I say something that I
- 1:03:55the positive effect for me is all of those jobs.
- 1:03:59You can agree that 90% of people who did them were not truly
- 1:04:02happy doing them. I don't think washing a car is
- 1:04:05the most fulfilling purpose of a human.
- 1:04:08I don't think lawn mowing is the most fulfilling purpose of a
- 1:04:11human. I think humans can have greater
- 1:04:13purposes solve graded problems. But it proves we are talking
- 1:04:17about billions of people. That's what I'm saying.
- 1:04:19Right. What if somebody?
- 1:04:21Likes yeah, what if someone likes to give service I.
- 1:04:23Just said no. I'm not saying no service is the
- 1:04:27whole term. Is a service.
- 1:04:29I never said service. I never said service.
- 1:04:32To do it. But no humans should be in
- 1:04:34service of others. That's how you get paid.
- 1:04:37Whether you're making a film, doing theater, comedy, products,
- 1:04:42apps. You're serving other people, so
- 1:04:44you're serving other people. That remains.
- 1:04:47What I'm saying is a huge amount of different services will now
- 1:04:51suddenly be done by robots in the next 10 years.
- 1:04:55All of those humans need to find a different way to create value
- 1:05:00for society. On a very basic level, the way
- 1:05:03we find purpose in life as human beings and also make money is by
- 1:05:08providing value. The higher.
- 1:05:10The higher the amount of value, the more the society is willing
- 1:05:14to pay you for it. The reason why Mark Zuckerberg
- 1:05:17is so rich is because he's provided great value.
- 1:05:21I use his softwares every day. For the past 15 years, he's
- 1:05:25provided great value. My business runs on his social
- 1:05:28media. No, but that would be redefined
- 1:05:30too. What I'm saying is all of the
- 1:05:33people short. Let's say out of 100 people, 10
- 1:05:37people, 100 people, general number, arbitrary number.
- 1:05:40If 100 people in the world do lawn mowing, 10 of them might
- 1:05:43truly enjoy it. Those 10 people can still
- 1:05:46continue to do lawn mowing with the robots.
- 1:05:49That will not change. But the 90 people who just did
- 1:05:52it as a way to make money to support themselves, they now can
- 1:05:56have a different job that may be more fulfilling.
- 1:06:02What I'm saying is we have to find a way to figure that out.
- 1:06:05That's the redefinition of Labor.
- 1:06:06I think Elon and all thought about it.
- 1:06:08There is no way to figure it out.
- 1:06:09By just, I think. I think humans don't know.
- 1:06:12Yeah, don't know. Again, could you know you pay
- 1:06:14them? Like I said, could Elon, could
- 1:06:17Elon or anybody as a matter of fact, predict social media
- 1:06:21influencers 30 years ago? No, no way you couldn't.
- 1:06:24It was impossible. But this is, this is.
- 1:06:26Could you imagine? That bigger again?
- 1:06:29I agree that we don't. No, I never said it's not much
- 1:06:31bigger. I agree.
- 1:06:32I said it's bigger. That's what I'm saying.
- 1:06:34For governments and for big companies, in order to plan, you
- 1:06:38need to plan based on the known facts, right?
- 1:06:41Not not you don't know what is going to happen.
- 1:06:43And I think. You don't understand my point.
- 1:06:45I agree with you. I never what I'm saying is how
- 1:06:48the fuck can we figure it out so quickly or how can we figure it
- 1:06:53out. That's what I'm saying cuz this
- 1:06:54is the first time in history. No, but you said that universal
- 1:06:57basic income or universal high income, it's the idea.
- 1:07:00No, that's why I said that's the only idea.
- 1:07:02That's the only idea, actually. It's well, that's the only idea
- 1:07:06that we have, Yes. Who else not I have.
- 1:07:09I have to think about this. Think about it, think about it.
- 1:07:11By the next episode, Pharos will have a better idea than the
- 1:07:14universal high income. Maybe I'll be the guy who comes
- 1:07:16up with the idea. I seriously don't what you're
- 1:07:19offering to me what you're offering.
- 1:07:21To I'm not offering the universal high income to you,
- 1:07:23remember, I'm offering it to the entire world.
- 1:07:26If you want to take it, I'm I can actually, I'm not giving it.
- 1:07:28I don't have that much money, guys, relax.
- 1:07:31What I'm saying is in the event of displacement, it is great to
- 1:07:36have universal high income so people feel safe, they have food
- 1:07:40and they have shelter. Anything above that we can
- 1:07:44figure it out. But that's the base that
- 1:07:45everybody should have. If that starts from big
- 1:07:48countries like China, US, Canada, if they start changing
- 1:07:53their structure of their labor, then they need to have
- 1:07:56immediately a plan on how to compensate this labor because
- 1:08:00ultimately. It's a huge disadvantage for the
- 1:08:02country's economy themselves if suddenly all of the humans or
- 1:08:06half of them 25% don't have jobs.
- 1:08:09Who will go to Costco? Who will go to?
- 1:08:11Who is going to buy? Objects or items or clothes,
- 1:08:15food. If if the cycle of economy
- 1:08:18doesn't run, If that cycle doesn't run, it hurts the
- 1:08:21country as a whole. So if the country's government
- 1:08:26replaces the humans with robots for efficiency, they have to
- 1:08:29figure out a way to compensate the humans for sure.
- 1:08:32Because the humans are the ones buying products for sure.
- 1:08:35Unless you say the robots start making money and they start
- 1:08:37buying the products. They want to start living like
- 1:08:40that. That way you're cooked.
- 1:08:42But what do you guys think? Actually putting in the comments
- 1:08:45because I think this conversation has came up
- 1:08:48multiple times in this show and I would love to use other
- 1:08:52people's other than me, Pratt and Faraz's, you know, point of
- 1:08:54view on this. So if you guys have an idea of
- 1:08:57how the future would look like with robots coming in and doing
- 1:09:01a certain, I would say a great degree of our jobs, what will
- 1:09:05happen to the job market, to the labor market?
- 1:09:08Let us know in the comments. As always, it's been a pleasure
- 1:09:12doing the show with you. We hope you guys enjoyed this
- 1:09:14episode. If you did, please do not forget
- 1:09:16to subscribe to the YouTube channel if you haven't already
- 1:09:20and. Follow us on Spotify.
- 1:09:22And Apple Podcasts, as it does us a great favour.
- 1:09:25It helps other people find this show.
- 1:09:28So if you find it valuable, I'm sure other people will too.
- 1:09:31So the way you guys can make it discoverable for others is by
- 1:09:35helping us with the subscription and the follow.
- 1:09:38As always, it's been a pleasure. As I said, we will see you guys
- 1:09:41on Friday on. Tuesday, Friday, oh, today's
- 1:09:45Friday. Have a great weekend guys.
- 1:09:47Have a great weekend. Ciao.
- 1:09:48See you guys on Tuesday. Ciao.