The Voicebot Podcast

2022 was the year of synthetic media. The mainstreaming of deepfakes and voice clones, along with the rise of text-to-image AI models, assured synthetic media of a breakout year. Then ChatGPT came along. It changed the conversation entirely and consumed news media and social media cycles for weeks. The GPT-3.5 model was better than expected, and the fine-tuning that delivered ChatGPT showed that large language models were ready to up end a lot of assumptions about what technology in general, and AI in particular, can do.  Joining host Bret Kinsella to break down the top synthetic media news of 2022 are Rupal Patel of Veritone, Michal Stanislawek of Utter.one and Hearme.ai, and Eric Schwartz of Voicebot.ai. Get ready for an in-depth discussion about everything from digital waste to the meaning of mortality. Along the way, the group discusses OpenAI, DALL-E, Midjourney, Stable Diffusion, GPT-3, Google LaMDA, virtual humans, synthetic voices, America's Got Talent, and more. 

Enterprise voice AI has been overshadowed for years by the tech giants' activities. That meant the consumer applications often drowned out what was happening in the enterprise. At the same time, most enterprises were moving slowly. That has changed over the past two years. Enterprise adoption of voice and conversational AI solutions is growing steadily and expanding into new use cases. Today we will talk about the contact center, restaurants, automotive, and media sectors. We also go into some detail about large language models and how enterprises are thinking about ChatGPT, Omnichannel, and more.   Susan Westwater is the founder of Pragmatic Digital and Strategy Director at Vixen Labs. She is also the author of the book, "Voice Strategy: Creating Useful and Usable Voice Experiences."   Jason Fields is the chief strategy officer at Voicify, a leading platform for voice experience creation. He was formerly a senior vice president at Rightpoint and an adjunct professor at Emerson College.   Braden Ream is the CEO and co-founder of Voiceflow; the leading conversation AI design collaboration platform. Braden also was recently named to the Forbes 30 under 30 list.   Susan, Braden, and Jason are each working on the front lines with enterprises, so you will get some fresh and practical perspectives. Enjoy!

This is the first of Voicebot's voice AI 2022 year-in-review episodes, and today we focus on consumer solutions. There was no lack of news this year, and industry insiders Peachy-Jean Retizos, Tom Hewitson, and Eric Schwartz join me to break it all down for our sixth annual year-in-review show. Amazon's layoffs that impacted the Alexa and devices groups dominated industry discussions late in the year. However, it was just a few months earlier that a similar move by Google drove industry news cycles. The two tech giants have set the tone for voice AI in consumer solutions since 2014, so the pullbacks were big news. We lead off with these stories and how they are reshaping the voice AI consumer sector. While the moves are generally viewed negatively in the market, we also talk about the positive elements and how they were not exactly unexpected. There is also a discussion about what is getting additional focus in 2023 and where the new paths of growth are emerging. However, the tech giants are not the only game in town. SoundHound became a public company in 2022 and became the first large-scale voice AI pureplay to tap into public financial markets since Nuance. Synthetic speech engines had a notable year in terms of customer growth and acquisitions. And large language models are taking natural language in an entirely new direction, plus we hit on a few other topics. It's been a pretty amazing year with highlights and lowlights, and it was good to get some front-line experts in to hash it all out. Enjoy!

Jesse Shemen is CEO of Papercup, a company he co-founded in 2017. The company transforms audio and video media into multiple languages to broaden its reach. Papercup estimates that 99% of all content is only available in one language. Using AI tools in conjunction with human translators and a synthetic speech engine, Papercup is working with companies such as Bloomberg and Insider to make their content available in the native language of their international audience. Shemen has a finance degree from NYU. He previously was a venture lead at Octopus where he helped launch a wealth technology platform for financial advisors, and a co-founder of Deloitte Ventures UK. 

Chris Parkinson began working on the idea behind RealWear while at Kopin back in 2007. In 2015, he founded WearNext to explore routes to further technical development and commercialization of the productivity tool for connected industrial workers. That ultimately led to co-founding RealWear in 2016.   The company presents itself as providing the first hands-free and fully ruggedized head-mounted tablet solution. But it's not quite a tablet. It's a headset for voice interactive hands-free access to data, information, and applications.   Earlier in his career, Parkinson was a senior engineer at Alien Technology and a researcher at Battelle Pacific Northwest National Laboratory. He earned a PhD in computational and theoretical chemistry from the University of Manchester.

Dom Esposito is a top YouTuber that worked with Hour One to create a digital twin. That's a virtual human clone of himself. In this interview, we show Dom's clone and talk to the real Dom about his motivation behind the project. We also discuss the process for creating the clone and use cases he thinks will be most impactful. Dom Esposito began creating tech review videos on YouTube way back in 2012. He was previously a writer at 9-to-5 Mac and AppAdvice and a Creatives Producer at ClearChannel, now known as iHeart.  

Natalie Monbiot from Hour One joined me for the recent Synthedia event to present several new use cases in language learning, media, and entertainment that are expanding the market for virtual humans. She shows how synthetic media is being used at Berlitz, Defiance Media, and people creating entertainment on YouTube.  Monbiot joined Hour One as head of strategy in 2019. Prior to Hour One, she was an SVP at Publicis, where she worked on new technologies and the Samsung account. She was an SVP at UM Worldwide before that, and earlier in her career worked at IPG Media Lab. Monbiot earned a Masters’s Degree from Oxford. To hear an earlier interview with Natalie and the full Hour One origin story as of summer 2021, check out episode 219.

Maaike Coppens is the author of the new book Design Conversationnel published in French by Eyrolles with a forthcoming English edition. Maaike and I first met at an event in 2018 in Paris, and that provided a springboard to discuss how the priorities and expectations around conversation design have changed. One important topic we discuss is the rising focus on task completion for voice assistant applications as opposed to likeability and building affective trust. Much of this is driven by changing consumer behaviors and preferences.  Coppens is the vice president of design at OpenDialog AI, the developer of the open source conversation management framework. Previously, she was a senior user experience and conversation design consultant for Accor Hotels, Applause, and XAPPmedia. She also worked as a conversation designer at voice-first game maker labworks.io and is an Alexa Champion.  

John Campbell is the founder and managing director of the voice AI agency Rabbit & Pork, a division of TIPi Group. Previously he was head of performance marketing and SEO at another TIPi agency ROAST. The agency's start was automating the collection of answers from Amazon Alexa and Google Assistant for a variety of questions. That data led to several reports which captured the attention of brands, and from there, Rabbit & Pork began building Alexa skills, Google Actions, and other voice interactive experiences in addition to voice SEO projects. Campbell goes into some detail in our interview about how Rabbit & Pork automated voice search result collection. He also discusses where voice assistants source their answers to common questions, the strategies brands use today to improve their ranking, and approaches that do not work. In addition, we discuss how the Alexa Answers service works today, how it is evolving, and what that indicates about voice search. Beyond search, Campbell goes through the differences in building for Alexa versus Google Assistant and how Google's pullback from the third-party Conversational Action ecosystem has impacted how brands are approaching voice assistants today. We conclude by discussing a variety of industry use cases for voice and chatbot solutions.

Zohaib Ahmed joined us at the Synthedia synthetic media conference in September to discuss Resemble's work re-creating the voice of famed pop artist Andy Warhol. A Netflix documentary was under development that centered around Warhol's diaries. The creators thought it would be more impactful if the viewers could hear the ideas and experiences of Warhol in the artist's own voice. So, they began searching for a synthetic speech provider that could create a voice clone.   Ahmed goes through the creation process and some of the challenges faced along the way. Plus there are a couple of nuances that you will find interesting.   Zohaib Ahmed is the CEO and co-founder of Resemble AI. He appeared previously on the podcast in episodes 251 and 103. Ahmed previously was a lead software engineer at Magic Leap and Hipmunk and a developer at Blackberry.   Voicebot's Eric Schwartz interviews Ahmed in this episode and Bret Kinsella adds commentary at the end about the evolution of synthetic speech.