Duolingo Goes AI-First, Runway's New References Feature, Model Context Protocol Explained

30 Apr 2025 · 33 min

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Denoised Podcast Episode Summary

Episode Title

Duolingo Goes AI-First, Runway's New References Feature, Model Context Protocol Explained

Hosts: Addy Ghani (Media Industry Analyst) and Joey Daoud (Media Producer and Founder of VP Land) Release Schedule: New episodes every Tuesday and Friday

---

Episode Overview

In this episode, the hosts discuss significant developments in the integration of AI into the creative industries, particularly highlighting Duolingo's shift to an AI-first model, Runway's new generative features, and the implications of Anthropic's Model Context Protocol.

Key Topics Discussed

  1. Duolingo's AI-First Transition
  2. Announcement Details:
  3. Duolingo is transitioning to an AI-first company.
  4. Aim is to rethink workflows; minor tweaks to existing systems won't suffice.
  5. Gradual reduction of contractors as AI takes over tasks previously assigned to them.
  • Implications for Creative Professionals:
  • Concerns over job security and the displacement of contractors.
  • Discussion on how AI could be leveraged across marketing and language teaching.
  • Emphasis on the need for employees to adapt and learn to use AI tools effectively.
  • Company Culture:
  • Duolingo expresses commitment to employee support through training and mentorship.
  1. Runway's New References Feature
  2. Introduction of a New Feature:
  3. Runway launched a new feature during its Gen-4 competition, enabling users to provide reference images for generating video content.
  4. The tool combines up to three reference images with text prompts to create consistent outputs.
  • Performance Insights:
  • Runway's Gen-4 model is noted for its impressive performance in image generation and understanding of real-world physics.
  • High consistency in generated outputs is highlighted as a key strength.
  1. Anthropic's Model Context Protocol (MCP)
  2. Overview of MCP:
  3. MCP acts as a two-way API standard connecting AI tools with user data sources.
  4. It allows for seamless integration of AI tools with existing software like Blender and Resolve.
  • Practical Applications:
  • Discussion on using Claude as an interface for Blender to automate tasks like model building.
  • Potential for AI to replace tasks usually handled by technical directors in software like Maya.
  • Market Implications:
  • Speculation about how integration of AI will affect pricing models for software and services in the creative industry.
  • The hosts debate whether users will pay for both the software and the AI integration or if it will be bundled.

Key Takeaways

  • Adoption of AI in Creative Workflows:
  • Professionals must adapt and leverage AI to stay competitive.
  • Companies are prioritizing efficiency and cost-saving through AI integration.
  • Future of Creative Tools:
  • AI tools are increasingly positioned as integral to creative processes, streamlining workflows.
  • The expectation is for seamless integration of AI capabilities in existing software products.
  • Market Dynamics:
  • The competition among AI developers will revolve around ease of integration and user experience.
  • There is a growing trend towards subscription models that include AI tools as standard features.

Conclusion The episode encapsulates a pivotal moment in the intersection of AI and the creative industry, emphasizing the urgent need for professionals to embrace AI tools and adapt to rapidly changing workflows. The discussions on Duolingo, Runway, and Anthropic reveal a landscape where efficiency and integration dictate the future of creative endeavors.

Links and Resources: For more information and detailed show notes, visit [Denoised Podcast Website](https://denoisedpodcast.com). Feedback and reviews are encouraged on Spotify and Apple Podcasts.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:00The one thing Duolingo is known for is their amazing TikTok feed and characters. The owl, the sort of recent story Duolingo owl died. They had a funeral for the owl, and then they had footage of the owl dying when it was hit by a Tesla Cybertruck. It was a whole week-long story that took over the internet of the owl dying and then was resurrected again and came back. I don't know if any AI can generate that kind of story. That's so awesome. In this episode of Denoised, Duolingo is switching to an AI-first company. The latest on runway. And what Claude's model context protocol means. Let's get into it.

0:35all right welcome back everyone addy good to see you as always my pleasure how was your weekend it was fantastic i still didn't get to see sinners and uh i think what's going to end up happening is i'm going to end up streaming it which is such a shame i did manage to see it in 70 millimeter in imax i'm so jealous went down to irvine to go watch it because it was sold out here in la there was no screen i went to a noon screening sunday and it was packed full house so good awesome see yeah excellent film worth there's a single shot in it that is worth alone just seeing an imax okay and it pains me because i know the majority of people did see it i mean it's a great shot anyways but the majority of people that did see it it was cropped from the full imax for three to whatever the uh you got a panavision taller yeah you get the full uncropped immersive image it was uh it was amazing to watch um nice yeah i did see that the blu-ray they said would have both ratios and release streaming i don't know probably not unless do i have to buy a new tv like a four three aspect this is the thing where i think like the apple vision pro really shines right that i can't think of any other way where you could try to match the full immersive experience yeah aside from a headset yeah you know imax and vision pro or the quest would be such a good pairing yeah like premium filmmaking and delivery through those devices.

1:57Yeah, I mean, that's probably the best way if you wanted to watch a 3D film like Avatar or something. Yeah. There should be like an official IMAX streaming or app or some protocol, which there probably is. And IMAX has a streaming protocol. So you should also release some exclusive content that is stereoscopic, maybe from the set that will just kind of incentivize AVP users to consume it on their device. Yeah, who knows when we start seeing more of that, if more people shoot behind the scenes stuff with like the Blackmagic Cincinnati Immersive. Yeah, exactly. Oh, I think you have struck on a new thing.

2:28I think we just developed a new idea, a new behind-the-scenes format right here. Yeah, awesome. All right, first story. This just came out today. Duolingo posted an all-hands email that they sent out to all of employees saying that Duolingo is going to be an AI-first company. What does that mean? What does that mean? All right, being AI-first means we will need to rethink much of how we work. Making minor tweaks to systems designed for humans won't get us there. In many cases, we'll need to start from scratch. We're not going to rebuild everything overnight. And some things like getting AI to understand our code base will take time.

3:00However, we can't wait until the technology is 100 % perfect. We'd rather move with urgency and take occasional small hits on quality than move slowly and miss the moment. The big bullet point of how they're going to be switching is we will gradually stop using contractors to do work that AI can handle. So that's been the headline a lot of people run with is that they are going to stop reduce hiring freelancers or contractors and replace it with AI. Yeah, it's happening, isn't it? Yeah. ai is taking our jobs look it's not as bad as like full-time employees getting let go like contractors are sort of iffy and for most contractors because you're looking to fill in a stop gap for something where you are not hiring a full-time role yeah as it how contractors should be used i know for like a large use of cases contractors are hired when uh it probably should be an employee yeah just to take advantage of yeah uh these more ease of hiring contractors Yeah, I mean, they said at the bottom of the letter, Duolingo, we're a company that cares deeply about its employees and they're gonna support everyone with more training, mentorship, tooling for AI and also kind of, you know, pushing how AI is used in across the board if everyone should, you know, use AI in their day-to-day workflow.

4:08Yeah, I don't understand the nuance of what the quote was said, it will take some time for AI to understand their code database. Like, what does that actually mean? And do you actually trust an artificial intelligence system with essentially your business. Because if the code infrastructure is altered or compromised in any way, that's the end for Duolingo, right? Yeah, I'm sure that's probably the thing of like getting it to understand the code base, but also having enough barriers in place where it doesn't destroy. Delete all. Yeah, it doesn't, you don't vibe code your code base and whoopsies it.

4:44Yeah, I feel like this is, we're not gonna see the end of this. This is perhaps the beginning. I think Zuckerberg has teased some of this at Meta, where, you know, just paraphrasing here, he was something like, yeah, a lot of our coding can be done with AI in the near future. Yeah. So AI is an obvious use of the code base, obvious use, and just for the user experience and how they teach the languages. And I think they're teaching chess now too. But also like, I'm curious, what are their aspects? Like, are they going to expand this to like marketing, generating assets? Like, what are the roles that they would hire contractors for that they're looking at replacing with AI besides coding?

5:20Because coding is like, I feel like a low hanging fruit of you, maybe they had a bigger project, they had to bring contractors on and now a fewer people can do it and leverage AI tools for that. But what other things are they doing, which is what I'm curious about? I don't think at an executive level, there is like a line that they won't cross as far as not using AI for that one particular function of the company. I think if it makes sense, And if it's ready enough, they'll go for it. And I'm just saying that across the board at all the tech companies. Like right now we're at a point where we're in a downturn economy.

5:55The economy is weak. A lot of the companies are not seeing that quarter to quarter or year to year growth. Labor is increasingly more expensive. So if there is an efficiency to be had with any technology, AI or not, they're going to go for it. Yeah, I mean, yeah, I had to be like the, you know, the core as a corporation, but it's like, yeah, you're kind of obligated to try to find the, like, you're obligated to the shareholders to the investors to find the most efficient route to operate that business. How do you shield up against this? I feel like it's on you too. You got to dig into all the AI tools.

6:31You got to figure it out where either you harness these tools and you launch your own thing and you become the solopreneur who is able to create whatever the first billion dollar company that's like solo run or make yourself a more valuable contractor by being the AI contractor who can come in and harness these tools. Yeah, I think - Or 10X yourself. For sure. It's a clever cat and mouse survival game for the staff versus the decision makers and the leadership. As a contractor, what I would do, and I'm currently a contractor, I contract for a lot of stuff. What I would do is be at the forefront of the thing that I'm good at and try to use the best tools to increase my gains and my efficiency.

7:21So if you're a software programmer, there are programming platforms that, you know, Vibe coders love to use. Yeah, you should be leaning in a cursor and at least understanding or figuring out how to use these other tools. Yep. And how do you actually combine that with traditional, classical, you know, software engineering architecture? And how can you scale like a team of 10 with just yourself? and if you can present that case sufficiently i think you're a contractor who's going to see a lot of work uh maybe not a duolingo because they're pretty bent on right duolingo is a singular company and they're taking a you know position in a lean but there are tons of others so many other companies too where it's like they're not going to jump into this headfirst or they're like they might not even be aware of all everything out there you know you can come in and if you know how to use these tools and be like oh hey i could do the same project that you were going to hire a small company to do with people I could do it you know solo or with like two of us because we know how to harness these tools I'll give you my example a couple of things that I've thought of doing is using a lot of AI powered marketing tools a lot of automation tools to get better SEO on my name my website automating a lot more LinkedIn cold messages creating more content and having a regular cadence like same thing that you do with YouTube maybe do that with LinkedIn like These are things that anybody can do.

8:40The tools are available to them. And the person that does them will have a far better chance of getting that next big gig versus the person that doesn't, right? Unfortunately. If the quality is there. Because I've gotten so many awful, clearly AI-ridden cold emails. Hey, Jim. Hey, Joey. You'll love your latest podcast on denoised on whatever the random topic was that we talked about. Oh, okay. Is it able to extract? yeah because i can always tell them like where's your data source because also sometimes they'll bring in some random stuff i'm like oh you got this from some old website i used to post on but then i forgot i got some other ways there's just like some uh you know like oh it was so inspiring this thing and it's like have you thought of some completely unrelated thing that they're trying to pitch you know i take it back actually my clientele is too smart for this they'll never fall for this yeah they're just like thinking like these are studio executives Yeah, using AI to handle marketing or help boost your SEO, yes.

9:34Using it to start sending out cold messages to people also in tech. Yeah, that's probably a little cringy, isn't it? But I do see a lot of it done. And on the B2C, like on a consumer-facing side, it's absolutely atrocious to be on the receiving end of it. Yeah. Like we get a lot of cold calls and things like that. It is an effective sales tool for salespeople, unfortunately. Is it just a numbers game? Is it just like, yeah, I was just going to send all this crap out. and it's like, well, we know a couple of percent are going to reply. Yeah. I was talking to a friend of mine who lives in London and is a gold trader, like physical gold trader.

10:09And their success rate on cold leads is 1%. So they go out to 100 people. One of them will want to buy gold from them or trade gold. And he's like, yeah, my team is trying to figure out how to up that number to like 300 and 400 till we get three or four leads a day or a month or whatever. but I'm trying to figure out how to increase the conversion. So I get three or four out of the hundred. Right. Yeah. Which is like the human thing to try to figure out. But the automation thing is like, no, just send more. Right. The old automation thing. But I think that's where AI will come in too, because it's like, well, you could potentially increase that conversion percent.

10:46Yeah. By using AI to make outreach that's more personal. Yeah. Cause yeah, I mean, a lot of times the more personal custom outreach will work better. It just takes a lot of time and effort to do that. For sure. No one does that. yeah sales is like one of those necessarily evil things that we live with and have to understand um it is an essential part of business and commerce and economy it's also not ideal for people that are not built for it yeah yeah but it is essential like if you want to sponsor this podcast reach out we could use some new cameras yeah you know i was thinking of the marketing thing too because also one thing duolingo is known for is uh their amazing tiktok feed and marketing and characters you know i know about the l and the sort of recent story i don't know if you saw this one with the owl where the duolingo owl uh died was hit by they created an entire story they had a funeral for the owl and then they had footage of the owl dying when it was hit by a tesla cyber truck and it was a whole like a week-long story that took over the internet of the owl dying and then was resurrected again and came back.

11:53I don't know if any AI can generate that kind of story. That's so awesome. Can we turn that into a movie? I think the studios are waiting for new IP. I mean, I feel like if anything, maybe they will just lean into making the owl absolutely crazy with AI. But yeah, run by their internal team. What a crime to humanity to like take such a clever marketing team and then just unleash AI on it all. As a tool for them, I want to see what just kind of crazy off the wall stuff they come up with. Well, hopefully they're allowed to do that and not just some bots. I would hope so. I mean, if they're AI first company, they should be here.

12:27All you please use VO2, use Runway Gen 4, use all of the AI tools. And then the Duolingo marketing team can leave Duolingo and become the Duolingo mafia and do some amazing marketing stuff at other companies. Yeah, they're going to start the next like liquid death and that new next gen of crazy marketing stuff. For sure. All right. speaking of gen 4 and runway got some more runway updates so they had their uh gen 48 competition over the weekend which is their like 48 hour filmmaking competition and in usual fashion they released or they teased a new feature that's coming out uh it was only for the participants of gen 48 so it's not out in wide release yet but then it will be probably in a week or two but it is references and so this kind of feature we've seen in other apps you can kind of give it reference images.

13:16We've seen this in like Kling, Pika, where you can give it a picture of a person, picture of an object or a picture of a location, and then it can combine them and make a video or an output. The interesting thing with Runway is it is built on their Gen 4 model and their Gen 4 model is really, really, really good. I've used it for turning image to video. Really good, really good real world understanding, physics, all that looks great. The references feature is with their the frame section which is their image generation section which they do have an image generator which i don't know if anyone not as many people are aware of that okay like like a mid-journey or like yeah i think most ai companies have different multimodal products yeah runway's known for video but they have a separate image model sure so it is give it up to three images person object place and then give it some text direction and then it'll create the image outputs but much better geography understanding and world understanding.

14:12Okay. So you can be like, you know, go remove the camera upwards. And it's sort of like people in the space with the image that you give it or the multiple characters that you give it and it generates the image. But then you can have a chat similar to what we've seen with chat GPT. Yeah. But just the kind of consistency is really unmatched from what I've seen people post online. So it looks more like a 3D software. Like a 3D software. But it's like the, you know, you give it a picture of you and it's like looks good here. new output looks like the same person looks like the same person so like one of the best consistencies that i've seen in the in the stuff that i've seen posted online because it's not open to the public yet but one of the obviously we've talked about consistency and consistent geography and scenes and you know that being a big challenge with ai generated work yeah it looks extremely impressive and and and promising that's so awesome so you follow cristobal on x and every once in a while you'll send me uh one of his latest tweets yeah he posts some uh fun commentary about stuff.

15:07He's a, he's a very, uh, visionary. Yes. Uh, very ahead of our times now once in a while, Cristobal will exaggerate, but Hey, he delivered here. I mean, the runway team delivered on something it feels like is very much next leveling up from traditional video generation. What about their other, uh, product that is speculative audio driven interface? Yeah, this was something Cristobal posted. And so it is a, I'll describe the video. The video is him narrating in real time like a video generation and basically holodeck voice commanding you know let's pan down from the sky oh we're in a field okay let's see the field all right let's add some let's add a city block here and then just like city blocks drop from the sky and starts coming let's modify the graffiti and the graffiti kind of comes in and in real time he is communicating and giving direction and the world is updating in real time to his voice so there's a couple things going on here of i mean this looks like the future and like yeah what we'll get to but a couple kind of elements that we need like one is just the real time feedback and generation so obviously now whether it's images or video when you're doing ai generation you give it the image you give it something you hit generate yeah and then even the fast ones you still you got to wait around yeah fast ones maybe take a few seconds more intensive ones it could take 30 to 40 seconds yeah you get your output back and then you're like no then you like change again yeah and you wait so there's a lot of waiting involved yes for sure uh it's funny i don't think cristobal ever sat in like a dailies room but this is exactly what directors do to vfx artists they just sit there it's like bring in the building move that tree bring in a bicycle but same thing because you're saying do this but then they go back with the notes and then they have to do it and then you wait and they come back yeah oftentimes i mean i feel like now with unreal and some of the real time tools it's pretty instant yeah but you can like just drag a asset from the marketplace or whatever library that the company already has.

16:59So it's taking like a 3d iterative creative workflow and bringing it into AI, which is not real time yet. Yeah. It's pretty revolutionary. So if we get there with real, with real time, cool. Also in this demo, and it's like a two minute long video and it's like, things are coming and going and animating and happening in real time, which is also not really the way things work right now. You give it an input and you wait and it makes a video and that's the thing. You don't like edit the existing video, you make a new video. So if this is what he's demoing, so things are happening in real time, but they're also just sort of like coming in and out.

17:32The world is changing. His generative world is changing in real time. And also he's talking to it. There's no text. So, I mean, that's not such a leap. You could do a voice to text. So that's pretty straightforward. But he's like having a conversation as if he is directing or telling someone, but the AI is doing it in real time. I don't think the future, at least for M &E professional users, is going to be talking to a system or prompting. Unless, I mean, I would say in the early stages, in the brainstorming, pre-production, storyboarding kind of phase. Like any artist will tell you their best work is done on a keyboard and mouse.

18:12And let's say you command the runway system to bring in this bicycle. And, you know, like we see in the video, the bike is just kind of floating into the interface. but an artist would actually want to place it exactly where it needs to be rotated scale it yeah you know so like that precision level of control is going to be needed i think this is great for perhaps introducing the public to what is possible with real-time ai systems but as it hits the professional market like we're going to be back to keyboard mouse or some kind of like tactical yeah i mean but you know i mean i think both will exist because you know you're going to have like you know we have avid and premiere and resolve and then we also have cap cut and uh iMovie and uh whatever other online v.io v.io uh so yeah you know like there's a we use those tools and it's like how do i could you know how do i get my fine-tuned control where do i get my control out of this and it's like it's supposed to be easy to use yeah or quick to use you know so you know both worlds both pipelines will exist but for sure yeah crazy demo what he was showing here and i'm also just as i yeah i'm curious what multiple elements going on here what's what they're going to roll out with first speed obviously would be the biggest improvement the only thing i've seen close with like really fast generation is prea which has like a pretty real time yeah generator in mid journey and like you can almost input a video and it'll start to do the frame yeah very low res and mid journey added a new like turbo feature where you could make it lower res but like kind of it'll generate in real time pretty close to real time right prea also has a 3D space.

19:48And one of their new features is a 3D space where you can kind of generate 3D elements and map them in a 3D space and then use that as your foundation to get more precise control over the camera and your composition. Yeah, so Chris Ball's video, something he teased, but no more information about that. Just off topic here, Chris Ball reminds me of a character from the show Silicon Valley. All of these. Everyone behind. Yeah, I'm really bummed that Silicon Valley is over because it would just be so good now with all of Sam Altman and all of the AI characters. Yeah, for sure. That show is great. Okay.

20:23All right. We agree on something else. I love that show. I don't want to live in a world where someone else is making the world a better place than we are. Rest and vest. All right. And our last story, model context protocol. MCP. MCP. So this model context protocol was something that Anthropic posted a paper about at the end of last year. And it's basically sort of like a API for AI. Yeah, is how I took away with it from their paper or from their blog post model context protocol is an open standard that enables developers to build secure two way connections between their data sources and AI powered tools.

21:02Okay. I mean, it's no different than a traditional API from like an Adobe or Stability or a big player like that where a user can build a full sort of set of features and then just leverage their backend for the AI. So the thing that this popped up recently or why I dug into it more was there was a Blender demo. So Cloud and Blender, MCP, there's a GitHub with a plugin that is basically use Claude as the interface. You're talking to Claude and be like, hey, build me this model in Blender. And then it controls Blender and builds the model for you. Yeah, I think I've seen that video. It essentially is using Blender's API because through Blender's API, you can build things, right?

21:49You can build me a box, this length by this length by this height. Now subdivide the geometry, now put a dent there. How do you build that Blender donut? right like it'll probably do that for you yeah and i've seen someone else demo clod connection with uh resolve yeah um similar idea the demo they showed was like doing timestamp okay tags and stuff so not the most useful but being able to start a chat interface yeah through clod and control directly through the api yeah yeah yeah that's pretty cool so yeah this more got me thinking about because we talked about the future of ai and tools and stuff what is gonna what what do we think is going to rise more is it going to be more stuff like this where claude or chat or whatever chat interface integrates with the existing tools and then you talk to it or like the claude computer control where it's just like going on top and using the computer slowly but maybe it'll get faster similar to how we talked about the atlas robot and the atlas robot is we're like oh it seems kind of silly but then we're like actually not a bad idea because it's kind of humanoid and so it already can just adapt to things that already exist in the real world Exactly.

22:55Same with like Cloud can adapt to the software that already exists. So is it like, is that going to be the way that happens? Or is it going to be the AI is integrated inside the tools like Descript is doing with their video, their Vibe video editor tool? So it's like integrated or just like we're just going to layer on Cloud or something on top of existing tools? Yeah, this feels like two API shaking hands, right? We have the Blender API and the Cloud API. They're shaking hands and magic is happening. like every command that should be uh software driven or human driven is now coming from the ai and all of the stuff that's being done in blender is being done by the blender api right it's like the perfect combination of traditional technology and new technology right when blender didn't have to build an ai integration into yeah their software right exactly so maybe blender is not the best example because blender is free but let's take a application like maya which is kind of a expensive i think it's like over a few thousand dollars a year okay so if you're a studio and you're using maya and now you have a hundred artists who are using maya and everybody needs to be so let's take a very simple task of you're making a movie and you have five primary characters and those characters need to come in get imported into your scene the rig needs to be loaded and all the automated stuff that needs to happen for an artist to start working at this point you can hire a TD, a technical director to write that script out and turn it into an executable that each artist can run on their Maya instance.

24:32And then all that stuff can prep them to work essentially. Or you can turn to the Anthropic API and just tell Anthropic like, hey, can you make me a quick tool, little applet that'll do this task? Okay. So Anthropic has just done what the TD would have taken a couple of days to do. So now you've taken two days of human labor, turned it into Anthropic output. Do you pay Anthropic for that? Or because you're paying that incredible Maya license, do you want that bundled with Maya? Where is the revenue ultimately going to come from? Is it going to be from the application layer where Maya is? Or do you actually care about the magic sauce of the model where anthropic is like if you did this today could like does maya have uh api or like it has a really robust api uh it's probably uh much more powerful than blender you would still i mean at this point you still need someone with technical knowledge to connect the two one time one time right and then you could tell claude do whatever you want and then the artist can probably talk directly to Claude and have it do what Maya needs to do.

25:46Yeah. Right? And so it goes back to my question of, is that studio paying Autodesk for Maya, or is that studio paying Anthropic, or is it paying both? I think it's paying both, but I think you're paying Anthropic whatever pennies. Yeah, Anthropic is not making the money here. Right. I mean, they're making a little money for the couple tokens you're using to run a command, but this is a pretty i mean probably a pretty simple command for claude to generate yeah so the money is not really in the api tokens i mean as lucrative as it is at a macro scale right like if you're talking millions of artists right i mean we're talking about one example on a on a very specific uh piece of hardware but something more widely user-based um like a a script or like a veed or cap cut if they start building in ai in the back end and it needs to do edits or analyze b-roll or help run a cut and if it's using chat to bt or clod as an api on the back end yeah i i think the the move the move is for uh anthropic to go to autodesk and sell a bulk license of clod to every maya license so uh the next version of maya that you download on your desktop will have built-in cloud in it yeah but it should be built in and it should be included it like i think everyone got tired of like when the apis were kind of pricey and then all these companies added ai features but then sold it as an add-on i think everyone got pretty tired of that where it's just like i didn't ask for these ai features i think a lot of companies were pretty early in adding them too so they like i think now everything's a lot more useful and practical but people's first taste of ai integration was like a year or two ago and it was still like pretty useless kind of annoying yeah a lot of pop-ups okay a lot of like you know five bucks an extra month or so like i we use click up for all of our project management stuff and like i love click up but then they added another ai thing and then they wanted to like upsell by like five or ten bucks for every seat yeah and it was like i think maybe now it might be better but at the time it was like yeah it's kind of not that great okay i don't really want to pay extra for this and it's like i'm really paying for chat gpt like and it does it better than you do so i think it's improved but what i'm saying is i think i don't you know i think ai is going to be expected to be part of everything and that like people are going to be pretty tired of having to pay extra it should be like no this is standard yeah and i think and that's also what google did i mean i think partly it was because uh they couldn't sell the add-on for people to pay up to get gemini so then they just kind of raised the prices a couple bucks for all of the google workspace seats but now all the workspace seats come with gemini which i'm like i'm not that against actually so i get access to all these tools and it's kind of useful but yeah i think i think that's going to be the new way forward which is like this should be it's going to be expected to be standard like yeah i think nothing will get dethroned if you will like for example i don't see maya being utilized less by film and tv community regardless of how much ai is adopted Like we're still going to need a 3D environment with the robust set of tools where environments are built, characters are animated.

29:00Yeah, that level of control that you need. Yeah. Yeah, sure. Is AI going to generate some of the models? Sure. Is AI going to do some of the animation? Sure. But the existing underlying platform is Maya. Question would be less about, you know, something like Claude coming into something like Maya, but more like, will it integrate with one of these generative 3D object tools or, you know, an image generator to generate textures? Like, where is that coming? And is that in the pipeline? Yeah, like a good example would be Marvelous Designer, which is a standard tool for creating clothing for characters.

29:38so could marvelous designer be ai powered so you don't have to like actually cut the fabric and sew the fabric and all the things that uh a cloth designer has to do and can that just plug right into maya like the way it does now marvelous designer plugs into maya so that yeah i think there's efficiencies there but for sure the money is in the things that we use every day in our lives in your case google workspace is not going anywhere right no no like for millions of people with small businesses like google workspace is the heart of that small business that's where the emails happen that's where the invoices are generated and so on right can that be augmented by gemini and can gemini get a little piece of that action financially yes so that's probably what's end up going to happen yeah they raised the price by like a dollar too and it's like whatever and else you didn't really have a training so there was no option for like yeah yeah there wasn't really an opt-out of that but yeah i've been seeing it more and more and i think we talked about this before where between google and apple they're probably the best toys where it's like they have so much of our personal data yeah you know say we went about that but they could make it very useful the most useful ai in our lives by actually you know scheduling things for us or putting it on the calendar or like understanding the context of who we talk to and uh i've been seeing that more and more with uh gemini and it pops up more and more now and when i'm in google drive okay in my docs it's like you want to ask it hasn't really been useful but it's too helpful i see it i see the potential with uh uh you know of it getting there yeah um going back to sort of my prediction for a lot of these foundational model companies you know like your Anthropic, your OpenAI, your Stability, your Adobe.

31:23It's all about being as frictionless as possible in the adoption of AI. What I mean by that, like have the most robust, easiest to use API. Have the best cloud platform so you're not running anything on a local computer. Have the safest, most secure thing for government contracts. Like if you can check those boxes, if you can just make it as easy as possible for somebody to spend money on your AI product. That's the win. It's not ultimately about the best quality or the coolest feature. It's just going to come down to very practical things like that. Yeah. How much does it help you speed up to do the things that you've been doing?

32:04And how easy is it to actually do? Yeah. And built-in, seamless. You don't have to figure out a clever prompt. You don't have to get it to do what you want it to do. It just does it. So once someone figures out how to get it there and integrated and... Set it and forget it. Eventually the point where you don't even realize that, like, you know, a quad or whatever is happening in the background is just doing it. Yes. That's the way. That's when Anthropic is making money as a sleek, right? Because it's just fully integrated into this workflow. Yeah. Yeah. Well, links and everything we talk about, as usual, at denoistpodcast.com.

32:37And thank you for your support. I hope you were able to leave us a review or a comment on Spotify or Apple Podcasts. that's where a lot of our growth is happening. And we appreciate it if you can do that for us. Yeah. All right. Thanks everyone. See you in the next episode.

From the publisher

Duolingo's memo announcing an 'AI-first' approach signals a shift away from contractors. In this episode, Addy and Joey analyze what this means for creative professionals, alongside Runway's impressive Gen-4 updates with its new References feature. Plus, we explore how Anthropic's Model Context Protocol is enabling Claude to control professional software like Blender and Resolve, raising questions about the future of creative tools.

#############

The views and opinions expressed in this podcast are the personal views of the hosts and do not necessarily reflect the views or positions of their respective employers or organizations. This show is independently produced by VP Land without the use of any outside company resources, confidential information, or affiliations.

More from Denoised

All 101 episodes
Duolingo Goes AI-First, Runway's New References Feature, Model Context Protocol ExplainedDenoised · 33 min
Listen in VO