You have 4 summaries left

This Week in Startups

Next Unicorns: Building a $1B AI Avatar Business with Synthesia CEO Victor Riparbelli | E1776

Wed Jul 12 2023
Music IndustryTech IndustryVirtual AI-based HumansVideo Content CreationText-to-Video PlatformsCustomizable AvatarsAI-powered AssistantsAI VideoInstructional VideosEfficiency GainsAutomationResponsible AI UseContent ModerationData ScrapingLicensing Agreements

Description

The episode covers various topics including the accessibility and flattening of the music and tech industries, Cynthia's virtual AI-based humans and video content creation, text-to-video platforms for enterprise applications, benefits of video-based learning management systems, customizable avatars and personalized video creation, AI-powered assistants and the future of AI video, AI video for instructional purposes and job performance, efficiency gains, video interfaces, and B2B marketing, automation, job disruption, and responsible AI use, content moderation, licensing, and regulation, data scraping, consent, and licensing agreements, AI advancements, travel blogs, and future events.

Insights

The Music and Tech Industries are Becoming More Accessible

Both the music industry and the tech industry are experiencing a trend of increased accessibility and flattening out. This allows anyone to create great music from anywhere and makes coding more accessible.

Cynthia's Virtual AI-based Humans Have Potential Applications

Cynthia is working on creating virtual AI-based versions of humans. While this technology can be both positive and negative depending on its use, Cynthia has raised significant funding and expanded its vision to replace the entire physical production process for video content creation.

Text-to-Video Platforms Provide Easy and Cost-effective Video Creation

The company launched a text-to-video platform that caters to a market of people who want to create video content but lack the skills and budget. This platform allows companies to easily create videos for training and communication purposes, reducing costs significantly.

Customizable Avatars Enhance Video Creation Experience

A video creation tool offers around 150 avatars that can be customized based on region or language. This tool is particularly useful for creating partner materials and allows users to add gestures, switch background images, and personalize their videos.

AI-powered Assistants and the Future of AI Video

CODA has introduced an AI-powered assistant to recapture time spent on tedious tasks. AI video and generative AI technology open up new possibilities for media creation, and the future of AI video may involve personalized, interactive, and multi-language content.

AI Video for Instructional Purposes Improves Job Performance

Using GPT-4 for instructional videos has made employees more effective at their jobs. Focusing on individual performance and productivity can lead to significant gains without adding headcount.

Efficiency Gains from Technology and the Dominance of Video Interfaces

Companies should focus on the productivity of their existing team rather than overhiring. Video content creation will become as common as writing in the future, with platforms like TikTok showcasing the dominance of video interfaces.

Automation, Job Disruption, and Responsible AI Use

Technological advancements may lead to job elimination, but human creativity and the emergence of new jobs are expected to fill the gaps. Responsible AI use, including consent, content moderation, and education, is crucial to prevent misuse of AI tools.

Content Moderation, Licensing, and Regulation

Content moderation is shifting from distribution platforms to creation points, and companies prioritize responsible content creation. The question of permission and compensation for using language models' training data is complex, and regulators debate the best approach to moderation.

Data Scraping, Consent, and Licensing Agreements

Companies are debating whether to scrape data online or obtain permission. The lack of clarity on what is acceptable creates uncertainty in the industry. Licensing arrangements and crypto technology could facilitate consent and enforce licensing agreements.

Chapters

  1. The Music and Tech Industries: Accessibility and Flattening
  2. Cynthia: Virtual AI-based Humans and Video Content Creation
  3. Text-to-Video Platform for Enterprise Applications
  4. Benefits of Video-based Learning Management Systems
  5. Customizable Avatars and Personalized Video Creation
  6. AI-powered Assistant and Future of AI Video
  7. AI Video for Instructional Purposes and Job Performance
  8. Efficiency Gains, Video Interfaces, and B2B Marketing
  9. Automation, Job Disruption, and Responsible AI Use
  10. Content Moderation, Licensing, and Regulation
  11. Data Scraping, Consent, and Licensing Agreements
  12. AI Advancements, Travel Blogs, and Future Events
Summary
Transcript

The Music and Tech Industries: Accessibility and Flattening

00:00 - 06:38

  • The music industry has become more accessible and flattened out, allowing anyone to create great music from anywhere.
  • The tech industry is following a similar trend as coding becomes more accessible.

Cynthia: Virtual AI-based Humans and Video Content Creation

00:00 - 06:38

  • The podcast introduces the concept of 'next unicorns' - companies that have raised their Series B funding round and have the potential to become billion-dollar companies.
  • The guest on this episode is Victor Ripperbelly, co-founder and CEO of Cynthia (pronounced 'Synthea').
  • Cynthia is working on creating virtual AI-based versions of humans, which can be seen as both positive and negative depending on how they are used.
  • Cynthia recently closed a $90 million Series C funding round at a valuation of $1 billion, with Excel leading the round and NVIDIA participating.
  • NVIDIA's investment in Cynthia is motivated by their research in media technologies that align with Cynthia's work.
  • Cynthia's initial goal was to make virtual avatars for customer support, but they have expanded their vision to replace the entire physical production process for video content creation.
  • They discovered that there is a large market of people who want to create video content but lack the skills and budget, making Cynthia's technology appealing even if it sacrifices some quality.

Text-to-Video Platform for Enterprise Applications

06:09 - 12:14

  • The company realized that there was a huge market of people who wanted to create video content but lacked the skills and budget.
  • They launched a text-to-video platform that caters to this market, focusing on enterprise applications.
  • Their platform allows companies to easily create videos for training and communication purposes.
  • The platform supports multiple languages, making it accessible to a global audience.
  • The ease of use of the platform eliminates the need for video editors and reduces costs significantly.
  • One example is a fast food company that used to send employees a lengthy handbook. Now they can create videos in different languages, improving comprehension and reducing costs.
  • Another example is a software company that uses videos to keep their sales team updated on product information and competitors' offerings.
  • The ability to easily edit and update videos makes this type of video production highly valuable for businesses.

Benefits of Video-based Learning Management Systems

11:46 - 18:32

  • Creating one-minute videos instead of written materials improves information retention
  • Product velocity is crucial for startups to stay ahead of competitors and meet customer needs
  • Crowdbotics offers CTO as a service, helping startups accelerate product development
  • Eliminating the need for media training, hiring hosts, and expensive production costs can save money
  • Video-based learning management systems can ensure compliance and track employee progress
  • Pricing for video creation software is seat-based, with options for different plans based on the number of users
  • The platform provides an intuitive interface with customizable avatars to create videos

Customizable Avatars and Personalized Video Creation

18:04 - 24:29

  • The podcast discusses a video creation tool that allows users to create videos using avatars.
  • The tool offers around 150 avatars of different ages, genders, ethnicities, and body sizes.
  • The avatars can be customized based on region or language to improve communication with specific audiences.
  • The tool is particularly useful for creating partner materials for English-speaking countries and Japanese companies targeting the international market.
  • Users can type scripts in different languages and select accents for the avatar's voiceover.
  • The text box also allows users to add gestures to the avatar's movements.
  • There are currently five gestures available, with more planned in future updates.
  • The gestures are computer-generated and aim to make the avatars more realistic and expressive.
  • The tool can be personalized by adding custom elements like employee names or specific messages.
  • Users can also switch out background images and videos to further customize their videos.
  • The tool provides an end-to-end video creation experience similar to PowerPoint or a camera app.

AI-powered Assistant and Future of AI Video

24:00 - 29:52

  • CODA is an all-in-one platform that helps teams work together and has introduced an AI-powered assistant to recapture time spent on tedious tasks.
  • AI video and generative AI technology open up new possibilities for media creation.
  • The UX of creating AI videos allows for easy customization and responsiveness.
  • Translation of videos becomes simpler with text-based AI video creation.
  • The future of AI video may involve personalized, interactive, and multi-language content.
  • New video infrastructure may be needed to support personalized and interactive videos.
  • The combination of various technologies like GPT and stable diffusion will enable amazing advancements in media creation.
  • Videos can be hosted at a URL instead of being exported, eliminating the need for manual uploads.

AI Video for Instructional Purposes and Job Performance

29:35 - 36:01

  • The idea is to have a library of instructional videos for employees at fast food restaurants.
  • Employees can use an iPad or workstation to access the videos and ask questions about specific tasks.
  • The GPT interface can interpret the questions and provide relevant video instructions.
  • There are many potential use cases for this technology, and it will be exciting to see what people build with it.
  • A real-life example is GoCatless, who built an entire university using Suntizia videos.
  • Using GPT-4 has made employees more effective at their jobs, with potential for continuous improvement over time.
  • The company will continue to grow its team, but GPT-4 will change the shape of many functions.
  • Focusing on individual performance and productivity can lead to significant gains without adding headcount.

Efficiency Gains, Video Interfaces, and B2B Marketing

35:36 - 41:56

  • Focus on the productivity of the existing team rather than adding more people
  • Companies should aim for consistent individual performance gains
  • Writing a job wreck and doing a search takes time and effort, but it's more effective than hiring
  • Being conservative on hiring can lead to a sustainable business
  • Overhiring is a problem in Silicon Valley
  • Utilize tools for productivity in all facets of the business
  • Video content creation will become as common as writing in the future
  • TikTok is an example of how video interfaces are becoming dominant
  • LinkedIn is a valuable platform for reaching decision-makers in B2B marketing
  • LinkedIn has a large user base of senior level executives and C-level executives
  • Efficiency gains from technology can range from 20% to 40%

Automation, Job Disruption, and Responsible AI Use

41:39 - 48:18

  • People are becoming 20 to 40 percent more efficient with the use of technology.
  • There is a concern about job elimination at a massive rate due to technological advancements.
  • The question arises whether jobs will be refilled or if universal basic income will be necessary.
  • Human creativity and the emergence of new jobs are expected to fill the gaps left by automation.
  • Expectations of content creation in the future include a shift towards video production.
  • AI may not match human creativity, but it can enhance productivity and lift all individuals.
  • Job disruption has occurred throughout history with the advent of new technologies.
  • Responsibility lies both with companies like Cynthia and society as a whole to combat misuse of synthetic avatars and fake videos.
  • Consent, content moderation, and education play important roles in preventing misuse of AI tools.

Content Moderation, Licensing, and Regulation

47:55 - 54:28

  • Red content, such as hate speech and swear words, is easy to detect.
  • Gray area content, like promoting risky investments or false health claims, is more challenging to moderate.
  • Only enterprise clients can create news content on the platform.
  • The company takes content moderation seriously and has strict policies in place.
  • Concentration of moderation tools is shifting from distribution platforms to creation points.
  • The company works with Fortune 100 clients and prioritizes responsible content creation.
  • Comparisons are made between the responsibility of AI tool creators and gun manufacturers.
  • OpenAI also practices content moderation but faces trade-offs in user experience.
  • OpenAI has been sued for privacy violations related to data scraping from the internet.
  • The question of permission and compensation for using language models' training data is complex and ongoing.

Data Scraping, Consent, and Licensing Agreements

54:12 - 1:00:27

  • A lot of companies are debating whether to scrape data online or obtain permission.
  • Some companies believe in getting permission for every video used to train algorithms, while others scrape data without consent.
  • Large models like GPT require internet-scale data, making it impossible to get consent from every author.
  • The lack of clarity on what is acceptable creates uncertainty in the industry.
  • Using robots.txt as a model, licensing arrangements could be established for scraping sites with permission and financial commitments.
  • Crypto technology could facilitate the transfer of payments and enforce licensing agreements.
  • Regulators debate whether to moderate the training data or the outcome of AI models.
  • Limiting AI models' output to exclude copyrighted works is one solution to address concerns about unauthorized content creation.

AI Advancements, Travel Blogs, and Future Events

1:00:04 - 1:05:39

  • The implementation of tokens is a bigger project that will take longer.
  • There are inherent issues with running and training on travel blogs.
  • Technology is improving to the point where visiting websites may not be necessary for gathering information.
  • An AI roundtable discussion is planned with the guest Victor.
  • Victor's company, Synthesia, is mentioned and may be hiring.
  • AI technology enables easier product development and democratizes various creative fields.
  • Founder Fridays events are starting in September.
1