Latest / Bad Decisions Podcast / Seedance 2.0, AI Agents getting Faces & Voice, Google releasing their most intelligent open source model & OpenAI Bought TBPN
Transcript
- 0:00Sea Dance 2 Point O, the world's strongest video generation AI
- 0:05model is finally available. Your AI agent can now have a
- 0:10face and a voice. Google just released their
- 0:14strongest, most intelligent open source AI model that is free for
- 0:19everyone to use. Open AI acquired TBPN.
- 0:23Let's talk about it. What's up, beautiful people?
- 0:26Welcome back back everybody to another episode of Bad Decision
- 0:29Podcast. And here is where we bring you
- 0:32the hottest and the latest and the greatest in the world of
- 0:36science, technology, AI and business.
- 0:39And we've got 5 hot topics today.
- 0:42It's Friday and. There are a lot of things that
- 0:45happened in the past two days since we last spoke.
- 0:48So this is the weird thing, we both went for a massage at the
- 0:52end of the night and when we went for the massage you're
- 0:55supposed to go in and come out with your body feeling better.
- 0:58But for some reason I feel more pain after having done the
- 1:02massage and I'm in agony right now.
- 1:05By the way, just want to let you know my mind is perfectly
- 1:07excited for the show, but my body is in excruciating pain.
- 1:12Maybe I should take you to another massage.
- 1:14No, again, the whole point of a massage, and I love massages, is
- 1:18to get rid of pain, right? Not to be in pain.
- 1:22Pain may be equals to growth or not in massages, not in massage
- 1:25workout. Again, I don't know about you
- 1:28guys I love to get massages but this last time I just came out
- 1:31of it feeling more in pain and I'm I don't know what to do
- 1:34about it. I.
- 1:35Don't want to get another massage?
- 1:36Go to another one. To get rid of them and I think I
- 1:38just have to do it this through. But yeah it's Friday, I'm
- 1:42feeling great mind wise getting ready for the show.
- 1:44We got 5 topics so let's start with the first one go If you
- 1:49guys remember that Tom Cruise and Brad Pitt fighting video
- 1:53that took over the entire Internet, it scared everyone
- 1:56because the video was so realistic and it was made by AI.
- 2:01This is the video that we're talking about, Pratt, if you
- 2:03just go to the beginning of this, this fighting scene right
- 2:06here, it was so realistic that Hollywood had to send cease and
- 2:10desist letters to Bytedance, the owner of C Dance 2 Point O, this
- 2:16video generation model. Now, why are we talking about
- 2:19this? Or because as of today, C Dance
- 2:222 Point O, the world's strongest AI video generation model is
- 2:28available for people like us It's available on almost every
- 2:32big platform that you know of Higgs sealed AI freepic hey Gen.
- 2:37in video AI cap cut you name it. They have it.
- 2:41And this model is known for being incredibly photorealistic
- 2:46with strong character consistency as you can see and
- 2:50minimal artifacts, especially in heavy motion shots.
- 2:54It has multimodal inputs which means it can read text prompts,
- 2:59audio and photos for references and it comes out and exports for
- 3:04you, a native audio director level camera control, realistic
- 3:10physics and multi shot cinematic output.
- 3:14Every single C dance video that we've seen on the Internet has
- 3:17blown us away, which is why so many people were scared because
- 3:21it is a huge jump from what we had before.
- 3:26For example, I'm comparing it to something like Clink 3.0 and it
- 3:31does one thing really well that most models fail at in my
- 3:34opinion. Action scenes and visual
- 3:37effects. If you pay close attention,
- 3:40every part, every shot has some sort of visual effect in it.
- 3:44And physics itself is so realistic, which is what most
- 3:48models fail at. Like the fighting, the punch
- 3:51just doesn't seem to land when you use other models.
- 3:54But here you got entire fighting scenes that I'm sure people
- 3:57didn't even prompt for. You know, you cannot describe
- 4:00every single punch, every single gunshot, but the model just has
- 4:04strong understanding of how action scenes work.
- 4:08OK, that guy just walked into the as I said that, that guy
- 4:12literally just walked into. No, but it's still, you know,
- 4:15from a text prompt that understands every single element
- 4:19in the same camera, movement, lighting, composition, colors.
- 4:22It's crazy good. And what is really interesting
- 4:25now is that as you mentioned, it's available on almost all the
- 4:29platforms. Yeah.
- 4:30We are not depending on a feature or a platform anymore.
- 4:34Once a model is out, all the big platforms will have it
- 4:38immediately. Yes, now even 11 labs, which of
- 4:41course recently they have video, they also have C * 2.0.
- 4:44Listen, every platform has been waiting for the moment of this
- 4:49release because they know how many users will be demanding it.
- 4:53But you know, and so the. Enterprise people can take it,
- 4:55right? Yes, and I'm going to talk to
- 4:56you about that and the reasoning behind that because it has
- 4:59pissed so many people off. But the reason why every
- 5:01platform has it is because the platforms who market this and
- 5:06win the SEO optimization. If you go to Google and search
- 5:09for C dance Pratt, can you go to Google and do that search?
- 5:12Actually C dance 2 point O. Let's see who comes up first.
- 5:14I think cap cut should come. First, whoever wins that rank
- 5:16thinking is gonna have the most users because people are gonna
- 5:19want to generate with C dance 2.0 it being the strongest video
- 5:24generation model right now. 11 Labs.
- 5:27C Dance. Sponsor arts list.
- 5:30So it's the second one. What is the what is the third?
- 5:33I don't even know them. What is that?
- 5:35Divine or design? Was it?
- 5:37What is it? I haven't heard of them.
- 5:41Then it's the OK, these are the sponsored 1 so we have seed
- 5:44dance official ones and Higgs field.
- 5:45So Higgs field is at the top. There is a catch with C Dan 2
- 5:52point O. As Farhad mentioned, it hasn't
- 5:54rolled out in every country yet, so you may not be able to access
- 5:57it. And even it is, even if it is
- 6:00rolled out in your country, if you've been trying to get your
- 6:03hands on C Dan 2 point O, you may have realized that it is
- 6:06exclusively available on team or enterprise plans with business
- 6:12e-mail verification. And I saw so many people under
- 6:16Hicksfield's Instagram angry, furious, saying we subscribe to
- 6:21the creator plan and you promise that we're going to get early
- 6:25access to all the upcoming AI models.
- 6:27How come you suddenly hide that behind the paywall of the
- 6:31enterprise plan? And initially I was pissed off
- 6:34too. Then I went to Freepic and I was
- 6:35like, wait a second, Freepic is also offering it under a
- 6:39specific plan and is the same case with most of these other
- 6:42companies. So I looked this up on Grok and
- 6:46try to find out exactly why this is happening.
- 6:48And this is a deliberate strategy driven by legal risks
- 6:52and costs. Let's cover the legal risks
- 6:56first. By Dance had to pause their
- 6:58global rollout of C Dance 2.0 back in March after Hollywood
- 7:02studios sent the cease and desist letters.
- 7:05Platforms can protect themselves by requiring business
- 7:09verification, which creates paper trails and accountability.
- 7:14Think about it for a moment. If Bad Decisions logs in with
- 7:18their business e-mail, we are more likely to use this for
- 7:21professional use, ads, marketing and client work rather than
- 7:26viral deep fakes and personal misuse.
- 7:29Imagine how many people would go in there and randomly use it for
- 7:32the wrong reasons because the quality is so goddamn high.
- 7:36So this way they can protect themselves knowing that they're
- 7:39already in trouble with Hollywood and many other
- 7:42companies for copyright. But I I read somewhere, was it
- 7:46Hicksfield or Free Pick that they said they're going to roll
- 7:48out to everybody? Soon they will.
- 7:50Not going to last. That long they will eventually.
- 7:52Because then because VO3 and all will get so good, this wouldn't
- 7:55make sense anymore. I think this is just for I think
- 7:57the period. Just the beginning, but there's
- 7:59also the reason of cost. And the reason why we have to
- 8:02discuss that is because video generation is one of the most
- 8:06expensive tasks to run at scale, especially when it's multimodal.
- 8:11You're inputting audio, you're inputting video, You're
- 8:14inputting, excuse me, you're inputting photos, text prompts
- 8:19and audio and all of that. Perhaps, and I haven't used it,
- 8:24so I don't know, but all of that has to be sent to the cloud,
- 8:28processed and then every frame of the video with this level of
- 8:33quality has to be sent back to you.
- 8:35And imagine every tier of every one of these companies going in
- 8:39and using this. So that is impossible to run at
- 8:42scale in the beginning. Therefore, why I believe one of
- 8:45the other reasons why they're doing this aside from legal
- 8:48risks is cost. And what's really funny is
- 8:52there's one place these companies like Hicksfield and
- 8:54Freepik can get access, and that's Byte Plus their official
- 8:57API. Apparently the commitment they
- 9:00need to make is an annual amount of 1 to $2,000,000.
- 9:06That means if you're Hagen or Hicksfield, you have to say I'll
- 9:09be paying you one to $2,000,000 for the upcoming year.
- 9:13Give me your API. I want to have access to C Dan 2
- 9:16point O for my users. That is an incredibly high
- 9:20amount of cost. For the for the big platforms
- 9:23like Higgs field and Freepik with which they have hundreds,
- 9:261000 millions of users, I think they can make that API in a few
- 9:30months. Yes, they will have that API,
- 9:31but small. The reason why they don't have
- 9:33it on a free tier or a lower tier is because they need to be
- 9:37able to keep making money. Yeah.
- 9:40And the team and enterprise plan allows them to do that,
- 9:43especially in the beginning to recover costs.
- 9:45But these are two reasons. But this is what I found.
- 9:48Do you believe there's other reasons or this is perhaps?
- 9:51No, I think this is the probably the season.
- 9:53This is the third that you mentioned in the beginning.
- 9:55That's one of the big reasons and cost is the second one.
- 9:58But I don't think it's going to stay behind the pay ball for
- 10:01more than a month. No, it is.
- 10:02Going to be a no. The pay wall will still be.
- 10:04No, I mean a huge pay wall that you have to get the enterprise.
- 10:07Enterprise. Yeah, how?
- 10:08Many creators have team and enterprise.
- 10:11Not many, not many files. Pratt, go to the first video
- 10:17generated by Alex. I wanna show you files the
- 10:19prompt that he used if you go to the comments.
- 10:21What is the video Let's? Watch Let's Watch a video.
- 10:23Just backflips on the. Yeah, so there's a guy who's
- 10:26doing a backflip on a jet. Looks pretty good because I was
- 10:30looking for someone who's actually accessing it himself
- 10:34rather than videos from big platform right?
- 10:37Because again, those guys will have less limits.
- 10:40Those guys will try unlimited. Let's go down and look at the
- 10:43prompt. The funny thing is he prompt the
- 10:47timeline like how you what happened.
- 10:51At which second? Exactly.
- 10:52So it's not if a guy does the backflip on the plane, it's like
- 10:570 to 3. The guy is standing at this feet
- 11:00on F16, three to seven. He performs.
- 11:04He gives a thumbs up, 7 to 12. He does a completely casual,
- 11:09perfectly clean, full backflip, 12 to 15.
- 11:13He brushes a tiny bit of dust off his short.
- 11:16So imagine instead of writing the whole text prompt that a guy
- 11:20on an F16 do a backflip, we give the timeline time codes.
- 11:26And what I would envision in the future, you can actually have a
- 11:30timeline instead of a video. There in your timeline you can
- 11:34have text prompts. So how you have timeline in
- 11:37Davinci and Premiere? Pro Yeah, the tracks.
- 11:39And exactly. And then there was a plus.
- 11:41So at 4 second plus add, I don't know add cinematic lighting
- 11:44here. At 8 add another character here
- 11:47or this character start jumping or this character starts.
- 11:50Shooting editing in. Real time exactly.
- 11:52Editing the video in real time. I forgot the name of the
- 11:54director who said if you want to be a good director, you have to
- 11:58be a good editor. I, I can't remember one of our
- 12:00podcast just mentioned it, but this is exactly that, right?
- 12:03You take everything and you start editing it.
- 12:06Of course not by text prompt, but in the future you can do it
- 12:09by photo references, video references.
- 12:11Sure, I can imagine that this is an insanely good video.
- 12:14By the way. Of course, we all know that this
- 12:18is humanly impossible, which makes a video instantaneously
- 12:21fake. But the quality of it, it's
- 12:25really good, is insanely high. I really appreciate you showing
- 12:29the prompt because I didn't see the prompt.
- 12:31This is really cool the fact that you can time code the
- 12:34creation of the AI video. I'm generally curious how C
- 12:40Dance 2.0 is going to perform. I am almost 100% certain that we
- 12:46are not getting the raw quality of what they showed us back in
- 12:51March. The Brad Pitt one, right?
- 12:52Yes, knowing what has happened with the copyrights, the legal
- 12:57risks and the costs, knowing that many users are going to be
- 13:01flooding, especially after seeing this show or us posting
- 13:04on Instagram and other creators talking about this.
- 13:06Everyone's going to want to try CDN 2.0 and we've discussed this
- 13:10in detail in the past. Compute is incredibly expensive,
- 13:15so for these guys to be able to produce the quality of videos
- 13:18for us, I have a feeling they're going to have to turn down a
- 13:21knob just a little bit, of course.
- 13:23But over time, as there's going to be more competitors coming in
- 13:25and trying to beat C Dance 2 point O.
- 13:27This is the kind of quality we can expect from all of the
- 13:30models. I imagine six months down the
- 13:32line, for sure Clink will have a similar quality, will probably
- 13:35have the similar quality. Luma runway.
- 13:37I think they all need to. Elon is going really hard on
- 13:40Grok with Grok Imagine. I'm pretty sure they're going to
- 13:43have Grok Imagine. Physics is actually incredible.
- 13:48Their quality is only 720P, which I'm sure they're working
- 13:51on to improve. Once they do that, I imagine
- 13:54they can catch up pretty soon, but you can always come back to
- 13:57the show if you guys want updates on all of these video
- 13:59generation models. Let us know what you guys think
- 14:01about C Dance 2.0 being out. Are you guys going to get access
- 14:04to it? Do you have a business
- 14:07enterprise plan or are you going to wait for this to roll out to
- 14:10every other plan out there? Far as I'm ready for the next
- 14:13Let's go. Your AI agents can now have a
- 14:16face and a voice. This is the first video chat
- 14:20skill for any AI agents by Piccolabs.
- 14:23Imagine going to a Google meet but instead of talking to actual
- 14:26someone, it's your AI agents. It's your open claw agent that
- 14:30talks to you and perform tasks during the call.
- 14:34You can say that hey check my calendar, book an appointment
- 14:37for me for next Wednesday. I want to have this meeting.
- 14:39Boom, done. And imagine you can have other
- 14:42people also come into the same Google Meet.
- 14:45Let's look at the demo that Pika Labs created and then we're
- 14:48going to analyze it. This is really cool.
- 14:52If it works, it's really cool. Wow.
- 15:01Oh my God. Hi, Cheryl.
- 15:02Hey, Letty. Finally meet face to face.
- 15:06Is the animation really like that?
- 15:08Looks like a high quality animation.
- 15:11Stylized, right? Yeah.
- 15:12Can you book me a meeting with Demi?
- 15:16I found an opening in your calendars at 3 PMI think we
- 15:20should go all in on Sprockets. Remember the competitive
- 15:24research indicated otherwise. touché OK debate time.
- 15:30Is hot dog a sandwich? What?
- 15:33OK, the quality of this guy is pretty low speaking.
- 15:36Sandwich even the voice. I'm not really a fan of the
- 15:38entire voice of these agents, but for me, this is the first
- 15:42sign of having an AI agent. Whether you're using cloud GPT
- 15:48pickup, whatever LLM that you're using, this is the first sign of
- 15:53them talking to you and interacting to you in real time.
- 15:56So instead of typing to open claw or typing to claw, you
- 16:00cannot talk to them and they will talk back to you.
- 16:03And now they you can put a face to them.
- 16:05You can give them a name and then you can have multiple
- 16:07agents. So you can have Alfred who will
- 16:10take care of your daily tasks. You can have another AI agent
- 16:14that can do a lot of things. You can bring them to meetings
- 16:17with you. So I think this is the first
- 16:19step. And as we've seen with Cdents
- 16:21and other features, the moment somebody does it, all the big
- 16:25players will do it. So I think we are going to get
- 16:28close to having voices, faces for every single agent.
- 16:33So Picolabs is an agentic AI company, right?
- 16:36They allow you to create your own agents, yes.
- 16:38And now they're putting faces to these agents with voices,
- 16:42correct, That run in real time. Correct in the meeting.
- 16:45OK, so Hagen allows us to create and clone ourselves.
- 16:51Yes, for business purposes or personal content, but that is
- 16:56not real time. And also that doesn't do tasks
- 16:59for you, right? So you create your own clone and
- 17:02then that clone can make content by your prompt, right?
- 17:05But this is agentic, right? This is, I'm going to tell them,
- 17:07hey, go and check my calendar book something for me.
- 17:10So imagine this is your open claw, but in the face.
- 17:12And you know, when you go to any meeting in today's age on Zoom,
- 17:16there's a no taker silently sitting in the corner, Otter AI
- 17:21or some other company. And you always look at it and it
- 17:24always joins the meeting before other people.
- 17:27And it feels like there's someone watching you and
- 17:29listening to you. So I always mute it before
- 17:31anybody joins the call. But the moment the call begins,
- 17:34they're listening in on the conversation.
- 17:36Now they can interact in between the conversation.
- 17:40So imagine if you told it about all of the requirements of your
- 17:44client and what they want, what they want to avoid predictions
- 17:48about the company. And you may forget yourself what
- 17:52you need to present in that moment.
- 17:54You can quickly ask the AI agent and say, hey, by the way, when
- 17:56we worked on the company's proposal, I was talking about
- 17:59this specific topic. Can you remind me what it was?
- 18:01And the AI agent will talk to you and tell everybody what it
- 18:05was. Yeah, yeah, and and.
- 18:07Just like, just like another fucking.
- 18:08Person like a business analyst you can ask at the end of the
- 18:11call hey can you summarize the action items and tell everybody
- 18:14what they need to do and then OK Pratt you you need to do this
- 18:17far as you need to do this far as you need to do this basically
- 18:19having. Another person in the call.
- 18:22Yeah, exactly, exactly. I imagine meetings are not gonna
- 18:27be 10 people and one agent anymore.
- 18:31It's gonna be 10 people and 10 agents.
- 18:33It's gonna be 10 agents. So it's gonna be no you, for
- 18:37whatever reason, you need to be present and have a discussion.
- 18:41You need to make critical decisions in the meeting.
- 18:43Let's talk about a high level. Meeting.
- 18:44Yeah, high level meeting. You all have your agents right
- 18:47now. There's one, there's the note
- 18:49taker in every meeting. What is going to happen?
- 18:52Is there going to be 1 agent per person in the call?
- 18:55So there will be a total of 20 people, but it's 20 people,
- 18:59humans and 20 participants in the call, 10 humans and 10
- 19:07agents, but an agent for every human, all adding into the
- 19:12meeting. Generally, I think with agentic
- 19:15AII think the number of people should in the meetings should
- 19:19dramatically decrease. But for now, how can you trust?
- 19:23How can you trust that? What's happening in the meeting
- 19:26is going according to your needs.
- 19:29Every single discussion you need to.
- 19:31Are you going to go and listen back?
- 19:33You're going to, you're going to have one or two people.
- 19:35What I'm saying is in the corporate world anyways, when
- 19:38you have more than four people in a call, usually two or three
- 19:41people are not even participating at all.
- 19:44They're just there. So instead of them being there
- 19:47listening, they can send their agents.
- 19:49When you say 10 people, it rarely happens that 10 people
- 19:53attend a Zoom call and all of all ten of them are
- 19:57contributing. It's very rare.
- 19:58That's true. So I would love to have shorter
- 20:02meetings with less amount of people if you want to be
- 20:05involved. And if you have nothing to add
- 20:07and you just want to listen, just send your agent.
- 20:10It will give you a summary. Would you feel like that is a
- 20:13rude thing to do? If you go into a meeting and all
- 20:17that you see is people's agents showing up, so you go to the
- 20:21meeting, considering that it is an important meeting and
- 20:24everybody else just sent their agent, would you find find that
- 20:26to be? Rude, yes, but if I know
- 20:28everybody is sending their agent, I will send my agent too.
- 20:31No, but you wouldn't know. You're going into a meeting and
- 20:34you show up. Suddenly it's five agents
- 20:38talking back to you. OK, now, yes, it's going to be
- 20:41rude. One year from now, you're not
- 20:43gonna realize if they're agents or not.
- 20:46No, but OK, you say you're not going to realize, but I'm sure
- 20:49100% certain all of these different companies like Google
- 20:53Meet and Zoom are going to have an AI label.
- 20:56Have you seen that video? The zoom guy and he keeps
- 20:59insisting that put three in front of your face.
- 21:01Put 3 here. No, but that's him with a deep
- 21:03fake. That's a different story.
- 21:04No, I'm talking about an AI agent popping up just like now
- 21:08you know there's a note taker. It's a label plus a notification
- 21:13that tells you this meeting is being recorded, Yes.
- 21:15So when you show up, five agents show up at the same time and
- 21:20you're not meeting other humans. Would you find that to be rude?
- 21:23OK to to scenario am I paying or am I getting paid?
- 21:29If you're getting paid, you'll. Shut up if you're paying.
- 21:32If I'm getting paid at that meeting, send your agents, send
- 21:35your cousin. I don't care if I'm paying.
- 21:38You better show up. I think the ultimate form is to
- 21:43have high level meetings with humans and agents and low level
- 21:47meetings for catching up. Simply catching up.
- 21:50As in let's all report on what has been done and what has been
- 21:54achieved that can be done purely with agents.
- 21:57That can save so much time. And this goes back to our
- 22:00discussion 2 days ago for Jack Dorsey's company Block, where we
- 22:07discussed their new architecture of company structure.
- 22:11They're getting rid of the middle management layers and
- 22:14trying to build a Central Intelligence unit.
- 22:16This is an example of how it can happen.
- 22:18Instead of having managers report contacts up and down the
- 22:22chain, you can just have the AI models and agents come and do
- 22:28the work. They will send the messages up
- 22:30and down. It's honestly an interesting
- 22:33technology. I am for this.
- 22:35I do like to see every human coming in with their own agent
- 22:38and the agent is there to add to the meeting.
- 22:40Not take away, but add to the meeting.
- 22:42Right now we have the note taker in every fucking meeting that I
- 22:45go to. I don't have mine, but everybody
- 22:47else does and it sends everyone an e-mail of what has happened.
- 22:51No, but if our our meeting is that we go to to be very honest
- 22:55it is no, it's all very productive and very small and
- 22:58very lean big corporates they have every Monday morning they
- 23:02have all hands meetings, yes. Then you have every day at 6:00
- 23:04PM daily updates. Then you have for every clients
- 23:07you have user requirements you have.
- 23:08Question do you think these big corporations are going to allow
- 23:12the low level employees send their agents to the meeting?
- 23:16But the. Low level employees won't be
- 23:17there anymore, just an agent, the low level employee.
- 23:20So then the agent is doing the work.
- 23:21Exactly. So that's a.
- 23:22Different story here. I'm talking about humans and
- 23:25agents. It's about humans using agents.
- 23:28It's not about agents using. Agents I'm thinking about
- 23:30company using agents like a becoming like bad.
- 23:33So you want the company, but then.
- 23:35So B, no, but what you're saying doesn't make sense.
- 23:37I'll tell you why then this is unnecessary for what you're
- 23:40saying. Why?
- 23:41The agents don't need to get on a fucking call to discuss
- 23:44things. The agents can do it without a
- 23:46call. We need A.
- 23:47Call. We need to see someone face, so
- 23:49imagine why you have your. Agent and my agent need to meet
- 23:52face. Very simple.
- 23:53Why do they need a face? It's much better.
- 23:56If there is no. Human.
- 23:57Voice no for agents. They Oh yeah, of course.
- 24:00Oh, no, Agents. I'm not saying agents.
- 24:02You're. Saying the agents will do the
- 24:03work, and they will. Talk.
- 24:04Let me give you a scenario. Let me give you a scenario.
- 24:06There is this company that is delivering services to another
- 24:09company. Usually when there is a service
- 24:11delivered in big companies, there are multiple teams
- 24:13involved, right? You have an IT team, you have a
- 24:15legal team, you have a marketing team, you have a member from
- 24:18each team to attend all these meetings.
- 24:19Yeah. So usually they don't.
- 24:20No, usually they don't contribute anything.
- 24:22They just sit down and listen. Yes.
- 24:25All this marketing, legal and all can send their agents and
- 24:28then we and the guy who's receiving the services we go in
- 24:31as humans. Makes sense?
- 24:33I said it humans and agents and you.
- 24:36You won't have a video call. Of just agents, of course.
- 24:40They don't need faces. They can just talk without
- 24:43faces. They can just talk through code.
- 24:45Maybe they don't even need to. Talk.
- 24:47They just interfere with Yes, no.
- 24:48They interface. They interface without needing
- 24:51the video call. This is for us.
- 24:53That's what I told you. This entire feature, this entire
- 24:58feature is for us now, one more cool thing that I was thinking
- 25:02about. I want you guys to just think
- 25:03about this for a moment. Pride you too.
- 25:05You as well. Imagine gaming, people tend to
- 25:10prefer multiplayer games. You like to play games with
- 25:14somebody else, and that has been the trend at least for the past
- 25:2115 years, especially because of online gaming.
- 25:25Previously, you'd play games alone at home or you'd invite
- 25:28your friends, but nowadays everyone plays online and you
- 25:31want to play all of these role-playing games like GTARP or
- 25:34Fortnite because you're having fun with friends.
- 25:37A lot of people lose their friends over time in their video
- 25:41games. One person gets married, another
- 25:44person gets a job, and then you're there seeing your friend
- 25:47last online 396 days ago and you're crying because they're
- 25:51not playing the game anymore. Everyone's gone through that.
- 25:55But now imagine you can build your AI agent to be good at a
- 26:00game and somehow video games will allow that in the future
- 26:04where we have as many agents as we do humans and you go Co-op
- 26:09with your agent and your agent also has a face and a voice.
- 26:14So instead of me just jumping into a game with Pratt as I
- 26:17always did, if Pratt is no longer online, he got married
- 26:20for whatever reason by the way, not yet Pratt.
- 26:23We, I, I call my agent comes up with a face and a voice and
- 26:30we're on a call and I'm talking as if it's a real person and we
- 26:33play League of Legends or we play Dota together.
- 26:38It's crazy. That's.
- 26:39You guys, you guys. Simulation, right?
- 26:41That's simulation. Sure, but you guys see that's
- 26:44actually really cool and that will happen of course. 100%.
- 26:48Do you know how many times I wanted to play and friends run
- 26:50online and I'm like fuck, I wish I'd played multiplayer.
- 26:53The the joy of playing video games is when you play with
- 26:56friends. Now the argument can be made,
- 27:00can you be friends with an AI agent?
- 27:02Bro sis, people are already doing it right.
- 27:05People are already connecting on an emotional level with their
- 27:08ChatGPT, which is not really that human yet.
- 27:12What I'm saying is if you put a face and a voice and if it has
- 27:15contextual memory of all of your conversations, and then you can
- 27:19train it on a specific video game, and eventually video game
- 27:23developers will allow AI agents to also have accounts, again
- 27:28labeled as AI. You will go into the game
- 27:30knowing that this is an AI you're playing with right now.
- 27:33You're doing it when we, you know.
- 27:35Who is it? No, no.
- 27:37Right now we're already fucking doing it bro.
- 27:40Co-op versus AI. Oh yeah true.
- 27:42Of course most of these AIS are hard coded rule based.
- 27:46But when you play League of Legends, the training before you
- 27:49start playing with real people is against the machine.
- 27:52You play against the machine to train because other humans would
- 27:56get angry at a low ranked player.
- 27:57That's the whole game. And people train their champions
- 28:01by playing against machines. Now these machines have gone
- 28:05incredibly good at the game to the point that they can beat you
- 28:08because now they're adding AI. Think about actually open AI
- 28:11started like that. By the way, if you guys don't
- 28:12fucking know this, go look up open AI.
- 28:14When Elon Musk was still there, they actually went ahead to TI,
- 28:18The Internationals off Dota 2. They beat the fucking champions
- 28:22and AI team beat the Champions. That was one of the big moments,
- 28:26right? You remember that moment, right?
- 28:28You remembered that. So think about it.
- 28:30All of these AIS didn't have a voice and didn't have a face,
- 28:35and they belonged to no one. But now with open source models,
- 28:39the fact that you can run a model locally on your computer,
- 28:42which is the next topic we're going to discuss, the fact that
- 28:45you can own the intelligence, the fact that you can train your
- 28:48own agents, and now you can put a face and a voice to them.
- 28:53There is going to be a fucking world like Ready Player 1 where
- 28:56you go in to a game with your AI agent that has been trained on
- 29:01your behaviors, your characteristics, is your friend
- 29:05and you go and play video games with.
- 29:07Them. Yeah, it's perfectly possible.
- 29:09It's. Bro it's it's crazy like the
- 29:11fact like even right now there are applications where you can
- 29:15have like AAI girlfriend or something and then you can pay
- 29:18and you can talk to her and you can have like a relationship and
- 29:21people are actually putting money in that.
- 29:24Can you can you do a quick grok search of what you just said and
- 29:28say how many people are doing? I'm genuinely curious.
- 29:31Like AI girlfriends, what is the status on that?
- 29:34How do you get the number? I don't know, but you rock as
- 29:37trends. No, you will locally.
- 29:38Train your girlfriend right? No, no, these are cloud servers
- 29:41people go and why you no one has AI girlfriends locally.
- 29:45I mean, some people do. What then?
- 29:47All the nerds will have it locally.
- 29:48No, but I I'm genuinely curious. Pratt, can you pull it up here?
- 29:52Yeah, the search. Let me just quickly.
- 29:53Pratt is very following the search before he pulls.
- 29:56Off, yeah. If they put.
- 29:58No, because I'm genuinely curious as to how many people
- 30:01are already dating an AI girlfriend or talking to an AI
- 30:05girlfriend or boyfriend. We shouldn't be discriminating
- 30:08against men. No, but men are the one who. 10s
- 30:12of Can we see the question How many people actually invest in
- 30:15AI girlfriends now? 10s of millions of people
- 30:18worldwide actively engaged with AI girlfriends or romantic AI
- 30:22companions as of 2025, early 2026.
- 30:26Though exact investment paying users is lower, likely in the
- 30:29low millions for premium features.
- 30:31So 10s of millions of people are investing in AI girlfriends.
- 30:37AI companion apps including heavy girlfriend focused ones
- 30:41reach 220 million globally by mid 2025.
- 30:46It's 60. $1,000,000 the first. Time.
- 30:48Imagine you have this in a game like you're you're wearing VR
- 30:51and you have AI girlfriend that you can talk to every day and
- 30:53she's really nice. That's ready, Player 1.
- 30:56Right. Yeah, it's insane.
- 30:57But Ready Player 1, the characters were the most of the
- 31:00characters that we were looking at were humans playing as
- 31:03characters in a simulated world. Blade Runner is the best.
- 31:06Example, right? Yeah.
- 31:08Blade Runner Joy. Yeah, Joy.
- 31:10Joy ran out of battery or server once.
- 31:12No, no. He went out of no, he went out
- 31:15of the zone. Yes, he went out of the zone.
- 31:17But it's actually even grok has a companion mode if people don't
- 31:20know about it. Brad, are you?
- 31:23Are you? If people don't know about it,
- 31:24they should try out the Grok companion.
- 31:26The companion mode. It's the same thing, really.
- 31:30Anime girl, right? The anime girl, The anime waifu
- 31:32do we have? Access to it.
- 31:34I think it's on the phone right? Or it's on the PC as well?
- 31:37I'm not. Sure about that.
- 31:38Come on, don't lie to us. We know you're.
- 31:41We know what you're doing after this show ends recording.
- 31:44We know what you're up to, but I'll let you guys know.
- 31:47I'm gonna check with Pride if he's working with his AI
- 31:50girlfriend or not, and I'll let you guys know in the next
- 31:53episode. Make sure you remind me in the
- 31:54comments if I forget. A piece of advice for you, if
- 31:56you want AI girlfriend or boyfriend, train them locally so
- 32:00you can have access to all the data is and your conversation
- 32:03stays private far as. But can we no, for a moment, can
- 32:07you guys really for just just a tiny moment, imagine that
- 32:12scenario. No, it's gonna.
- 32:13It's gonna happen. And that can potentially boost
- 32:18the gaming market back into it's the highest moments when we had
- 32:23COVID. Right now, based on what I'm
- 32:27seeing in the market, people are incredibly distracted with the
- 32:31flooding of content. That is happening.
- 32:33Of course, due to AI as well and more people creating content,
- 32:37less and less people are investing time in console
- 32:40gaming. Console sales are down and what
- 32:44I think can happen in the future if we can make game exciting, of
- 32:49course, by developing great games and AI can help that
- 32:51because anybody can create games.
- 32:53And so if people, Oh yeah, this is this is the AI companion, by
- 32:58the way. Oh, so it's the anime version?
- 33:00Yeah, by the way, don't. Not down on the anime, not my
- 33:03not your cup of tea. How many people watch anime in
- 33:06the world? Put in the comment guys if you
- 33:08guys watch anime. IA lot of people watch anime?
- 33:11Yeah, I mean, I watched one. Yeah.
- 33:13And you loved it? I loved.
- 33:14It so good. Change some, man, right?
- 33:16Yeah. But what I'm saying is, if
- 33:20everyone can create their own version of the games that they
- 33:23want, eventually you're going to have some really good games born
- 33:30out of that, of course. And once we all can go in there
- 33:34with our AI companions. Not this version, not the
- 33:37girlfriend version, but a really cool AI companion.
- 33:40Imagine you can build your best buddy just like you did, but you
- 33:44don't. Build your best buddy.
- 33:45No, let me explain to you you. Make friend with your best
- 33:48buddy. Sure, sure.
- 33:49I'm not denying that. OK, but go and look at the data
- 33:53of how many people feel lonely in the world.
- 33:56Sure, go and look at the data of how many people feel isolated
- 34:00and left out. The data is insane across the
- 34:04world. The.
- 34:05Fix is give them an AI companion.
- 34:07I'm not saying this is the fix and the best fix.
- 34:10What I'm saying is that this is going to happen for sure, OK,
- 34:15and form a lot of people. It's going to happen in a good
- 34:17way, and for a lot of people in a bad way.
- 34:19Roughly one in six people worldwide experienced notable
- 34:22loneliness. One in six people, That's a lot,
- 34:25dude. One in six people.
- 34:28It can lead to an incredibly dangerous path to have people
- 34:33rely on artificial intelligence for friendship and
- 34:38companionship. Not saying that is what we
- 34:41should do. I'm simply talking about video
- 34:43games. I'm not talking about the real
- 34:45world. I'm not talking about you
- 34:46shutting your doors and talking to an AI model for the rest of
- 34:49eternity. I'm simply saying if you
- 34:53remember need for speed pat search need for speed.
- 34:57Was it most wanted or Nitro? Need for speed not Nitro.
- 35:01Nitro is the one of the need for speeds starts by this cinematic
- 35:05intro with a guy, a relatively obese guy, who is going to race
- 35:11you. That's the beginning of the
- 35:12game. Everyone loved that character,
- 35:15Everyone loved the character. The same is true with GTA.
- 35:19Some of the openings of the earlier versions of GTA, the
- 35:21moment you start the characters, the stories that they build.
- 35:24Imagine that person talking to you being real and you can Co-op
- 35:29in a video game with them. But it's not rule based.
- 35:31It's not like you finish the mission and everyone goes
- 35:35through the same exact mission. They are exclusively friend with
- 35:39you and they play with you. Right now, right?
- 35:42Not to the level I'm. Speaking about.
- 35:44The fact that they are outside of the game.
- 35:46You go on a Google meet, you I text Pratt and I'm like hey
- 35:51let's play Discord in 5 minutes, let's chat and he comes on
- 35:54Discord. We'll chat for an hour, then
- 35:56only we'll get into a game and then we'll switch games.
- 35:58That's a great business idea. You should sell it.
- 36:01Sell game companions. I know.
- 36:05For a fact, having been in the E sports world, I can guarantee
- 36:09you what will happen initially is a huge push back by players
- 36:13because the AI models are going to be incredibly good at the
- 36:15game. At some point they're going to
- 36:17be better than most human players.
- 36:19So somehow we'll have to either have a balance.
- 36:23So what we do right now in ranked games is you will have a
- 36:27balance of ranks. So if I'm in gold tier, Pratt is
- 36:30in bronze this year, we cannot play together, so you can only
- 36:34have gold and silver. Play against gold and silver.
- 36:37What will happen in the future is you're going to have two
- 36:39humans and three AIS in a team. You cannot have two humans and
- 36:44three AIS in one team and five humans in the other team.
- 36:46It may be unfair. So what will happen is we'll
- 36:49have to balance this out. But I'm telling you, this is
- 36:52going to happen. And for gaming specifically, I'm
- 36:55not going outside of gaming. It is dangerous outside of
- 36:59gaming, Gaming specifically, I think this is going to be really
- 37:02fun. I think so too, actually, You
- 37:05know, definitely it's going to happen 100%.
- 37:09All right, are we ready for the next topic?
- 37:11Let's go. Google just released their most
- 37:14intelligent open source AI model, free for everyone.
- 37:19You can download it on your computer and run it on your
- 37:21devices for free without the Internet.
- 37:25No cloud dependency, no rate limits.
- 37:29All the data is going to remain private to you on your devices.
- 37:34This is a goldmine for anyone who is trying to use AI models
- 37:39to set up businesses or even for personal use.
- 37:43Now this is different than the Gemini models.
- 37:46You've experienced Gemini models when you went on Google and
- 37:51asked questions from their LLM. This is different than Nano
- 37:54Banana that runs on Gemini models in the background.
- 37:58We're talking about Gemma models.
- 37:59Gemma models are their open source series, which means they
- 38:04have no restrictions on commercial use.
- 38:08Now why this is even more interesting than the previous
- 38:11Gemma models is because it is released under the Apache 2.0
- 38:16license. This has given every person full
- 38:23permission to modify, distribute and build derivative products.
- 38:29Gemma 3, the previous version, and Llama which is metas open
- 38:34source and other open source models sometimes have certain
- 38:38limitations. They are open source, they're
- 38:40free to use, you can download them locally.
- 38:43But the moment you build products and want to scale and
- 38:46sell to lots of people, there are certain limitations.
- 38:51But under the Apache 2 Point O license, they completely change
- 38:55the game compared to their previous Jemma 3.
- 38:58You can build any product that you want, train it without
- 39:02restrictions. Now this is called Jemma 4 and
- 39:07it is their state-of-the-art open source large language
- 39:10model. They claim that this is the best
- 39:13open model in the world for their respective sizes.
- 39:18It is available in 4 sizes right now that can be fine-tuned for
- 39:22specific tasks. 31 billion parameters dense for great raw
- 39:27performance, 26 billion parameter mix of experts which
- 39:31is faster for low latency, and 2 billion and 4 billion parameters
- 39:36for edge device use. Like phone?
- 39:39Exactly. Edge device being the devices on
- 39:42the edge of the network. These are your phones and
- 39:44Raspberry Pi's and laptops. Smaller devices that have
- 39:47limited hardware. They can now use the 2 billion
- 39:50and 4 billion parameter. If you are new to the show and
- 39:54do not know what parameter means, think about the
- 39:57capability and intelligence packed into these models.
- 40:00The higher the parameters, usually the smarter they are.
- 40:03But what's interesting is that the further we go down the line
- 40:08of AI intelligence, these new generations with lower
- 40:12parameters have more knowledge packed.
- 40:15Than the older generations with higher parameters.
- 40:18So you can compare let's say a 31 billion parameter Gemma 4
- 40:22with a Gemma 302 hundred 300 billion parameter.
- 40:26And that's because of their new architectural design.
- 40:29They constantly make more efficient architecture.
- 40:33Before I continue far out, what are your thoughts on Gemma 4?
- 40:35I think this is it's a conspiracy theory, but it's a
- 40:39push by the government for Google to release an open source
- 40:43model. You think so or 100% conspiracy
- 40:46theory because it's not mentioned anywhere.
- 40:49The reason Google comes out and give one of the most
- 40:52intelligence open source model could be that all the best open
- 40:56source models are currently from China.
- 40:59We have Alibaba Quan. We have Deep Sick and people who
- 41:02want to build in India, in China, probably in Russia, on
- 41:06Android devices and on Android devices.
- 41:08They all will go to Alibaba Quinn or go to DeepSeek.
- 41:12When they go and build with this product, they will stay in their
- 41:15ecosystem because it's open source.
- 41:17They can even contribute to the code and make these better and
- 41:21make them more optimized. All the US companies Grok, which
- 41:25is supposed to be open source, is now closed source.
- 41:27Meta supposed to be open source. The latest one is closed source.
- 41:30Open AI, Anthropic, they're all closed source.
- 41:33There is no one actually working on US based open source models
- 41:38and I think we need someone that is was.
- 41:41Competing with the Chinese models.
- 41:43Apparently. I mean, I haven't seen the
- 41:44benchmarks. This is no the benchmarks are
- 41:46complete. Better than Quan?
- 41:48Quan and DeepSeek and all of them, according to Demis, the
- 41:55founder of Deepmine at Google and the CEO who is running the
- 41:59AI at Google, it is comparable as you can see at the bottom
- 42:03over there at a 31 billion and 26 billion parameter model
- 42:08compared to the others, Kimmy Quinn at 397 billion parameter,
- 42:14at 1100 billion parameter. That is insane.
- 42:17Think of it this way, a much smaller brain that can run
- 42:21efficiently is far more intelligent than a giant brain.
- 42:25That is what this graph is representing this chart.
- 42:29But I'm looking forward to use it and especially the A posture
- 42:31license give us flexibility to even use it for clients.
- 42:35Use it for anything, use it for services.
- 42:36We can use it for our own apps. Definitely gonna give it a try.
- 42:40The architecture that you should always think about when you're
- 42:43building AI apps or AI enable app is that you have to have the
- 42:48flexibility of switching the models, right?
- 42:50So you can always have a drop down, Hey, do you want to use
- 42:52Quinn? Do you want to use deep sync or
- 42:54you want to use I mean Gemma? Do you want to give that?
- 42:57Sometimes companies don't even publicly announce to their
- 43:01clients or users what models they're using in the background.
- 43:05You don't have to if the options, you can do the
- 43:08switching in the background and the back end.
- 43:10If you are a software as a service, sure, but if you're
- 43:13delivering to a client, you should tell.
- 43:15The client not 100%. You cannot hide it.
- 43:18I'm not telling you what I'm fucking using for you.
- 43:22And by the way, one more sorry, one more thing please.
- 43:26A lot of companies who are selling software probably were
- 43:30getting banned from using Chinese model as their open
- 43:33source model. And they couldn't also use Jema
- 43:363 and Metalama because it had restrictions even though it was
- 43:39open source. This will help with that.
- 43:41This is going to help with that. And if you're a user of LLMS and
- 43:45have never tried building anything, you're just an average
- 43:48user. You go online, use the cloud
- 43:50services such as ChatGPT and cloud.
- 43:53We've always urged you guys to go ahead and make your own
- 43:56applications because you can. Now you can solve any problem
- 44:01that can be digitally solved. You can do it on your own and it
- 44:05is almost free. Almost free.
- 44:07Why I say that? Because we are using clawed
- 44:09code. It is the smartest model for
- 44:11coding. But when you actually build an
- 44:13application that is powered by AI, the moment you build the
- 44:16structure and architecture, you no longer have to use clawed
- 44:19API, which is expensive to run this system.
- 44:23You build it once, you pay for that, But then the moment you
- 44:26build it, you can deploy it with an AI model like this, which is
- 44:30incredibly smart. And you can always upgrade it
- 44:32when they release new ones. Now, do I agree with Farhad?
- 44:36Partially, yes. I agree that this is a huge push
- 44:40by Google to win the market share.
- 44:44This is textbook ecosystem capture.
- 44:46They're trying to win the trust of enterprise and developers.
- 44:50Think about it this way, every developer's instinct after this
- 44:53is let's begin the next project with Google's architecture
- 44:57because it is smarter and there are no restrictions.
- 45:00And once you build a reliable business model and want to
- 45:03scale, I'm talking about the developers and businesses who
- 45:06started using JAMA Four. And there's already millions of
- 45:09downloads, I believe since this was released.
- 45:13Google Cloud is the upsell. So yes, they're offering this
- 45:18for free, but imagine you start out this application in your
- 45:22home and then you have 10 users, 100 users, 1000 users, and now
- 45:26you're selling globally. You have more clients.
- 45:29How can you make a reliable network and a system?
- 45:32Well, you can no longer run it on your computer at home.
- 45:35You need to go to a cloud server because you're already making a
- 45:38lot of money. Guess who's ready to take that
- 45:40money? Google Cloud, because it is so
- 45:43easy to use. Since you're using JAMA, it is a
- 45:45one click deploy and that's why, if you've been thinking why
- 45:49would Google release this for free if they can make money?
- 45:54Well, because they're playing the long game and it makes a lot
- 45:57of business sense. The reason why I said partially
- 46:00is because I actually kind of also agree with Farrad saying
- 46:04this is pushed by the government too.
- 46:05But I'm unsure of that because, you know, you can't really trace
- 46:08that. We know that US and China are in
- 46:11a technology war. It makes perfect sense for Trump
- 46:15and his team to tell some of these companies to win the open
- 46:19source game. I remember clearly must have
- 46:23been Jensen or Elon, who said China.
- 46:26Is Jensen Huang, Nvidia's founder, who said China is
- 46:30winning the open source game? China is way ahead.
- 46:33This must be one of the ways to fight that.
- 46:36I 100% agree and I cannot wait actually to try it to see how
- 46:42much improvement we have because we have tried Quan and Olama.
- 46:45Let's see how much of a difference it make if we use.
- 46:48Gemma for I do want to say one thing, a lot of people sent us
- 46:51questions after we talked about the AI educational program that
- 46:55we're working on, on ai.bat decisions dot Studio, asking us
- 46:59what sort of things we'll be teaching because we haven't
- 47:01really released modules or categories of content.
- 47:05I'm telling you right now, one of the things that we're
- 47:07teaching people is how to build applications and software and
- 47:11websites for their own personal use and for clients.
- 47:15The thing is, the reason why you need the application is up to
- 47:19you. We don't really need to know the
- 47:20reason. All we do is teach you guys what
- 47:23are the best tools and how you can actually go ahead and build
- 47:27a usable program using codecs of GPT using cloud code and then
- 47:32bring in AI models that are open source like this to power your
- 47:37program to be smart. This is one of the things that
- 47:39we're covering. Just a sneak peek as to what you
- 47:41can get. If you guys haven't signed up
- 47:43yet, the link is down here below, and I'm gonna so wow.
- 47:48There's a really long band ai.bat decisions dot studio.
- 47:51There's a first name and an e-mail input that you can have
- 47:55right here. Once you do, you get in the list
- 47:57and then we'll send you a confirmation e-mail so that you
- 48:00will be notified when we launch Far Out.
- 48:03Yes. Before we move on to the next
- 48:05main topic of discussion, there's a video I wanted to show
- 48:08you. OK, that has blown my mind and I
- 48:13hope you haven't seen it yet because I wanted it to be a
- 48:15surprise for you. Pratt, can you pull it up?
- 48:18This video that you see before you is not real.
- 48:22It's not AI generated. This simulation was made in
- 48:26Houdini and rendered in Blender which is a free 3D software.
- 48:32Now look at how insane the caustics of the water, the foam
- 48:37particles. The wave. 100 million particles
- 48:41made with hydro FX in Houdini. All the bubbles, all the sprays,
- 48:45and you can see multiple simulations working together.
- 48:48You're seeing the water simulation and if you pay close
- 48:50attention, the hippo's feet in the sand is pushing the sand
- 48:56into the water. How the hell can someone make
- 48:59this? This is an insanely talented
- 49:01artist and these softwares are incredibly powerful.
- 49:05I don't know what to say about this.
- 49:07I thought this video was real by the way when I saw it.
- 49:09When I see videos like this I would tell myself in a few years
- 49:14how we distinguish reality. Bit simulation.
- 49:17This is pixel perfect. How long it took to render this
- 49:22you? Think I don't want to guess.
- 49:24I don't want to guess. Probably with few machines I
- 49:26would say. Is Jeremy the guy who built it
- 49:29or he's reporting it only? I think he's the one who made
- 49:32it. It has 1.4 million views.
- 49:33Wow, this must be the guy. Jeremy, shout out to you.
- 49:36This is insane. The technical skills required to
- 49:40make such a video is something I cannot comprehend.
- 49:45And bro good luck for AI look at your generations to create stuff
- 49:49like this. This is pixel perfect.
- 49:51This is exactly like reality. Wow, I love that.
- 49:56I'm blown away. I'm blown away.
- 49:57What I was more fascinated by, I know Houdini is capable of
- 50:01incredibly technically complex simulations.
- 50:08What I'm more fascinated by is Blender's ability to render such
- 50:12a photorealistic shot, especially when we're talking
- 50:15about water and particles. I was surprised that he had to
- 50:18take it to Blender to render this, but yeah, just fascinated
- 50:21by it. This is the post.
- 50:22Go and show Jeremy some love and yeah, let's move on to the.
- 50:27Next, wait a minute. Yes.
- 50:28Did he have the inspiration, the idea for this video from real
- 50:33life? Because to sit down behind the
- 50:36computer and think about Hippo's movement, either you have to
- 50:39watch hours of National Geography because I'm looking at
- 50:43the movement. This is not something that you
- 50:44see everyday. This is not something that you
- 50:46dream about or you wake up that oh, I want to create.
- 50:48This How do you know you're not in Jeremy's dreams?
- 50:51But the dreams are not even this real.
- 50:54So you either have to, you know, watch it live in nature or maybe
- 50:59watch a documentary or something and get fascinated by it and go
- 51:02behind the computer, spend weeks, months.
- 51:05Maybe you see the trees at the back by the way.
- 51:08Like the idea of the first thing is OK let's put it this way, AI
- 51:11can do this. AI video generated state that
- 51:143.0 VO 5 can do this. In the future, who will have the
- 51:18idea of doing this? How can someone prompt doing
- 51:23this? This is not OK.
- 51:25Let's put Jeremy's technical skill aside that he is great at
- 51:28Houdini and simulation, and he rendered in Blender what went
- 51:33through his mind when he was creating this piece.
- 51:36I'm also thinking about that because to put that into
- 51:39perspective, of course, with Houdini you have a lot of
- 51:42control. But again, how high the hippo
- 51:46will go up, how the water will move, what is the effect?
- 51:49To me it's it's really mind blowing.
- 51:51One more thing, I'm fascinated by Pratt go down.
- 51:54I just saw the comment that Pratt was showing is current
- 51:58movie CGI is painful to watch. I just read that first line and
- 52:01if you go back up, I'm just looking at it and he's so right.
- 52:05Some of the CGI we see in movies with budgets of millions of
- 52:09dollars and then we have this made by an individual.
- 52:12Can you Scroll down why did Jeremy replied to that comment?
- 52:16To be honest, the new Jurassic Board movie has some of the best
- 52:18fluid simulation I've ever seen and also amazing animation.
- 52:21OK, he. Gave I'm not referring to
- 52:22Jurassic's new. No, I haven't.
- 52:24I haven't. Watched it.
- 52:25I haven't watched it. I'm talking about other films
- 52:27that I have watched that I was disappointed by.
- 52:29This is just incredible. Shout out to you Jeremy.
- 52:32Far out. Let's move on to the next.
- 52:34Let's go. Open AI acquired TBPN, the
- 52:38technology business programming network.
- 52:41This is the first ever media acquisition by Open AI.
- 52:46The full team of TBPN, the 11 people who work on the show,
- 52:50they will join Open AI, including the Co founder John
- 52:53and Jordy. Hosts will keep the full
- 52:56editorial control, so the programming, the guest booking,
- 53:00the editorial calls will stay within the team.
- 53:03Open AI is there to accelerate global conversation around AI
- 53:08and of course fund them. It was real little bit strange
- 53:13for people on X seeing an AI company, an AI giant coming in
- 53:18and acquiring a media company and a media company that is not
- 53:21even on TV network. They are on YouTube X and Twitch
- 53:25and Spotify. Their show goes around every day
- 53:28for three hours from 11:00 AM. It's live stream.
- 53:31They bring guests ranging from Jensen Huang, Mark Zuckerberg,
- 53:35Alex Carr from Palantir, all the big names in the tech space, all
- 53:38the big names in the AI space and they get an average views of
- 53:4270,000 across all platform. So their audience is not huge.
- 53:47They are not Mr. Beast, but they have a very niche audience,
- 53:50people in the tech and business space.
- 53:53And I was reading the comments, one of the reason Open AI
- 53:58acquired TBPN was they want to have acceleration about
- 54:03conversation around AI and also create a space for real
- 54:08constructive conversation. This made me think a little bit
- 54:11because they were already actually doing it.
- 54:13They would bring CEOs from competing companies and they
- 54:17would talk to them. But now that, you know, Open AI
- 54:20acquired them, what will happen to those conversations?
- 54:23Would it be skewed toward Open AI or would it be from other
- 54:27companies as well? What do you think?
- 54:31Let's start with Open AI's decision to acquire TBPN.
- 54:39First of all, I like the show. We know the show from X, We've
- 54:44seen it on X over and over again.
- 54:46It's popped up and being shared by many other people.
- 54:49And I do think they're doing a really good job, especially
- 54:51because a lot of traditional media has lost trust.
- 54:55More and more people are listening to individual creators
- 54:59and small teams rather than traditional massive media
- 55:03machines. So first of all, I want to
- 55:07congratulate the guys because I know they were incredibly happy
- 55:09about the entire acquisition. An incredibly genius move by
- 55:16Open AI. Why?
- 55:18Because right now everyone knows, especially after what
- 55:22happened with the elections, that podcasts are an incredibly
- 55:27valuable media format. People trust long form content
- 55:33much more than they do short form content.
- 55:36Open AI and all these other companies had to go ahead and
- 55:39sponsor creators on Instagram and on Threads, even on X for
- 55:45posts, for single posts. But how much attention do people
- 55:49give to single posts? How much trust do people have to
- 55:53one single post versus a podcast or a show that is recurring
- 55:59every day or every other day? There's a huge difference.
- 56:03And so why? People know Joe Rogan is the
- 56:06strongest and most reliable source of news, and therefore
- 56:11the prices of sponsors for Joe Rogan have gone through the
- 56:14roof. The $100 million sponsorship
- 56:17deal, not sponsorship deal with Spotify and all the sponsors
- 56:21that came through the show. So people are aware of that,
- 56:26even at a political level. And now when we're seeing tech
- 56:29companies go in and instead of just paying creators one by one,
- 56:33they acquire a show, I believe it's a genius move.
- 56:36I haven't seen this done at all. I think this is a very fresh and
- 56:40new thing. That they acquired the entire
- 56:42channel. Yes.
- 56:43So it is a genius move and Sam Altman is doing this.
- 56:46He acquired Open Clock and then went for TVPN.
- 56:50He is taking every piece of the puzzle and trying to bring it
- 56:53into one ecosystem. What is the play in the long
- 56:57term? I'm afraid of that because
- 57:03ultimately speaking, Open AI has investors.
- 57:08Open AI has people to respond to.
- 57:11Sam Altman is leading, but he has to respond to the board, to
- 57:15the board for his actions. They just got their raise,
- 57:19billion dollar raise. How many billions, 122 billion
- 57:23and about 100 to 200 million is now spent on.
- 57:25This is one of the first things they're spending on.
- 57:27Right, 100 million from 122 billion.
- 57:29Nothing. So he's not just spending on
- 57:32compute. He's not just spending money on
- 57:35resources when it comes to building better AI.
- 57:38He's spending money on distribution and not the
- 57:40traditional way, not by making more ads or making more videos
- 57:46or posts. You yourself set Anthropics
- 57:49posts are getting millions of views.
- 57:50Yeah. So what can open AI do?
- 57:52I can try to make better posts, or I'll just buy TBP and TBP and
- 57:55come over here. You're not going to talk about
- 57:58me. And that's the scary.
- 58:00Part they have the editorial control.
- 58:02That's the scary part. We know that Elon hates Sam
- 58:06Altman. Yeah, Elon is one of the most
- 58:08significant players in the AI space.
- 58:12Can TBPN bring Elon on the show? If they bring If TBPN brings
- 58:16Elon to the show, are they going to be entirely unbiased when
- 58:20having a discussion with him? I don't think so.
- 58:24I don't think so, actually. I don't think he has, no.
- 58:28My question is, can they be entirely unbiased?
- 58:31You cannot be right at the end of the.
- 58:34That's the thing, And Sam Altman did make a post where he said
- 58:38we'll try to keep ourselves out of the editorial as much as
- 58:41possible. We're trying to be unbiased even
- 58:44though we acquired them, but I don't see how that even works.
- 58:48But. That's the thing.
- 58:48I don't. I don't.
- 58:49Understand, right. So, So what, what is the
- 58:52different? I, I think there's a different
- 58:53play here. Open AI could come in as a
- 58:56sponsor like Figma, like 11 Labs who are sponsors of the show.
- 59:01They could come in as a sponsor, pay few millions and done when
- 59:05they bought over the channel. There's a bigger play, there's a
- 59:08bigger move. So we shall see in the future
- 59:11what they have planned for TBPN. I'm Elon Musk, I'm sending you
- 59:16Bad Decisions Adm one year from now and I say, hey, I want XAI
- 59:21to come and acquire the Bad Decisions podcast.
- 59:25How much big numbers? Very big numbers.
- 59:30Yeah, but then you cannot talk to Sam Altman.
- 59:32Sure. That was fast.
- 59:36That was fast. OK.
- 59:36All right. What about your answer, Pratt?
- 59:41Your answer. If he gives, let me see the
- 59:44money first. Let me see the no.
- 59:46Let me see the cash first. Let me see the cash.
- 59:48Big money, Big money put the cash on the table.
- 59:50You have it. What are you thinking?
- 59:51I. Have never played GTA and all
- 59:52those things. You got to see the cash first.
- 59:54I want to go to space man. I'm I'm.
- 59:56Yeah, yeah, yeah. Listen, I love I'm.
- 59:58I'm actually teaming on anyways, yes, you know.
- 1:00:01It also matters what is the terms of the acquisition like.
- 1:00:05I'm sure now that they're acquired, these hosts have to
- 1:00:07stay. They cannot replace the hosts.
- 1:00:09But, and then the biggest what if they tomorrow they decide
- 1:00:11they don't want to do the show? And you know, thinking about
- 1:00:14acquisitions, if something happens to the host, what
- 1:00:17happens to the 100 to $100 million?
- 1:00:19Gone. But that's that's the same thing
- 1:00:22what happened, but it's the pounds of the company, right?
- 1:00:24Yeah. It's, it's just a strategic
- 1:00:26move, I would say. They're making a lot of
- 1:00:28strategic. Moves, yeah, but no.
- 1:00:29But an interesting, I would say anthropic move was, oh, let's do
- 1:00:33some fake leaks of my mythos and my open source code and then get
- 1:00:3870 million views. And they are.
- 1:00:39Going to you think their AI models are telling them to do
- 1:00:42this? Kind of you.
- 1:00:44Think Sam Altman is going to a hidden room inside the Jupiter
- 1:00:48AI compound. It's a concrete room that has 10
- 1:00:51different locks. And he goes in.
- 1:00:53He sits down in a meditative pose, and then there's a voice,
- 1:00:56a blue light. Goes across.
- 1:00:58You watch too many movies bro. Yeah, and then there's a voice.
- 1:01:00Sam, it's time to acquire. You're going to acquire TBPN,
- 1:01:05and then we're going to slowly tell them what to talk about
- 1:01:07Open AI. I don't know.
- 1:01:08Again, by the way, I'm not saying they're going to do this.
- 1:01:11And knowing TBPN, they've brought as many CEOs and
- 1:01:15founders as possible in the tech space.
- 1:01:17Talk to every angle except Elon. I probably.
- 1:01:21I don't even ask Elon. He didn't come.
- 1:01:23Yeah, I'm sure they've asked Elon.
- 1:01:24They are an X. I'm sure they've.
- 1:01:26Asked Elon is very I think he's very no weird in selecting the
- 1:01:30podcast he goes to he goes to the most famous one and he goes
- 1:01:33to some very random ones yes. So I don't know how he how he
- 1:01:36chooses which podcast to go. This is an.
- 1:01:38Interesting play. I'm very curious how this will
- 1:01:41turn out. Are we going to see Anthropic go
- 1:01:43and acquire another show and then Gemini acquire another one?
- 1:01:47But this job would be one of them.
- 1:01:50No, we are not for sale. You just.
- 1:01:52Said you're going to. You said Elon, you just named
- 1:01:55every single company. I'm not going to say anthropic
- 1:01:58and anthropic, but these are the big ones actually at Google.
- 1:02:03I didn't say I. Actually didn't start this to
- 1:02:05sell, but I'm sure they haven't started this to sell.
- 1:02:08I don't think they started this. They didn't even know.
- 1:02:10They didn't even know. And I'm sure the deal and the
- 1:02:14terms were so good that they said yes.
- 1:02:16As in, when you're told that you get full editorial access and
- 1:02:20you get 100 to $200 million budget, sure.
- 1:02:23It makes a lot of sense to say yes.
- 1:02:25Imagine what you can do on the show.
- 1:02:27Yeah. So hats off to TBPN.
- 1:02:30Genius move by open AI. Let's see where this goes.
- 1:02:33We're going to keep you guys updated on the show as usual.
- 1:02:38Do we have anything else? Nope.
- 1:02:39We are going to be bad before. Friday, yes, it's weekend time,
- 1:02:42which means you're going to have some time to play around with
- 1:02:45some AI tools, which means you're going to tell us in the
- 1:02:48comments what you are going to play around.
- 1:02:50Because listen, I'm always interested in reading your
- 1:02:53comments. And in the last episode, we
- 1:02:55didn't have as many comments on the last episode than we did in
- 1:02:57the previous one. So I'm urging you guys to go
- 1:03:00ahead and comment. I am genuinely curious what you
- 1:03:02guys are up to this weekend. I'm going to be reading some
- 1:03:05books and I'm going to be working on an application and
- 1:03:10the AI program that we're working on.
- 1:03:12So three things. Yes, AI program.
- 1:03:14AI program for me. And I'm also, am I going to get
- 1:03:18a massage again? I don't know.
- 1:03:19Gosh, I don't stop with that. I don't know.
- 1:03:21I definitely need to get rid of the body pain.
- 1:03:23I'm definitely not going to the gym that's cancelled.
- 1:03:25So that's my plan for the weekend.
- 1:03:27Pratt, anything. I'm just gonna keep prompting.
- 1:03:29Grok that. Elon, look at us.
- 1:03:31Elon, look at us. How do you?
- 1:03:33Prompt to grok How can we make Elon watching shows?
- 1:03:37Yeah, maybe it recommends it to him.
- 1:03:39Grok knows Elon better than anybody else, I assume.
- 1:03:42Right. OK, Father, you didn't tell us
- 1:03:43your plans. Before I told.
- 1:03:44You AI education platform. OK, the whole weekend.
- 1:03:46Yes, the whole week. All right, all right.
- 1:03:49Thank you, watch everybody, for watching and listening.
- 1:03:51Thank you, watch everybody. No, thank you very much.
- 1:03:54Everybody OK, Alright. I thought you.
- 1:03:55Said we see you guys on Monday. Ciao.