Beyond Text: Enterprise AI's Multimodal Moment, feat. Abhay Parasnis

Clara Shih: The world changes very rapidly and certainly every decade or so there's a big step function change that comes along. Whether it was the web, then it was cloud, then mobile, and certainly now we see we are in the midst of the next massive shift with AI.

Clara Shai: Welcome to Ask More of AI, a podcast at the intersection of AI and business. I'm Clara Shih, CEO of Salesforce AI. And I'm thrilled to be here today with Abhay Parasnis, founder and CEO of Typeface, one of the hottest AI startups. They're using LLMs to generate images and videos for brands and doing what we call multimodal AI. Abhay is a longtime technology veteran and he's on the board of Dropbox and Schneider Electric. Abhay, it's so great to see you and have you on the podcast.

Clara Shih: Great to be here, Clara.

Clara Shai: You're one of the most well- respected leaders in the technology world. You've had prominent positions at Adobe, Oracle, Microsoft, and now you're leading one of the hottest companies in the generative AI space. Just an incredible story. I wanted to ask about your background because you spent 20 years in these C- level positions at these leading companies, and then you decided to leave and found a company. What was your thought process behind such a big shift and how's it going?

Clara Shih: I guess is that another way of saying, what was I thinking starting a company? No, first of all, thank you for having me here. Great to be with you and thanks for those kind words. I think, in my heart, I'm a product builder, so even those companies that I was there over the years and they have been phenomenal world- class companies and world- class products and businesses, for me it always all starts with products and looking around the corner and building new breakthroughs that hopefully can delight customers worldwide. And so there has been a common thread, even though the context may be a little bit different in terms of size of a company like Microsoft or Adobe is very different than what we are doing at Typeface relative to where we are in the journey. But there is a connective tissue and a common thread, at least in my mind, which is looking at the next big inflection point. And I'm sure we are going to talk a lot about GenAI as kind of that next big inflection point today, but whether it was cloud inflection point at Microsoft or Adobe, it's a similar trend and similar journey for me. And so I would say the deep passion and conviction around building breakthrough products, looking around the corners and frankly hopefully being surrounded by some amazing world- class people who are passionate about taking on a new challenge, that's kind of the common thread. Then there are obviously lots of different things about startups as you know very well, and from your amazing career journey, there are days when things you take for granted in big companies, you got to roll your sleeves and do all that stuff, nitty gritty, payroll and setting up business insurance plans. But that's all part of the journey and it's all fun.

Clara Shai: It's such a dramatic shift, isn't it? And I am coming the other way, so maybe we can share notes. Many of the enterprise AI startups that have been formed really focus on text. And what's so exciting about what you're doing with Typeface is the multimodal nature, the fact that you're generating brand images, brand campaigns, brand video, what's different and harder about images and video than text?

Clara Shih: Yeah, that's a great question and I do believe multimodal is going to be probably the next big frontier as a industry in the next year or two. The Chat GPT moment for next year, I predict will be more around multimodality, which is telling stories that seamlessly cross from text to image to 3D to animation and all of the above. So I do think that's an exciting future that we are all working towards. Hopefully we are doing our part at Typeface. But to your question, I do think text to image is a whole different sets of challenges come in, and I think frankly, image to video is another level of complexity that is yet to be solved fully. I think there's some exciting early approaches, but if I were to give you my non- technical aspect of what's different, I think the way our brains process, as humans, the way our brains process text and language versus when we see a image, a image has a very deep emotive connection for our brains around the color combination, the harmony. A lot of times you'll see a image, Clara, and you will immediately either love it or something about that image may be off before you even realize what it is because your brain has processed that the color harmony isn't there, or the composition of the two objects in the image are just off. So first I would say there's a additional cognitive complexity with generating amazing images that the way human brain aesthetically finds pleasing, finds very tasteful. Whereas with text, you have a lot more of that nuances with language specific differences, different cultures, different languages, which itself is a huge problem and challenge. With images, it is somewhat more of a universal visual language in some ways, but it has a additional information density problem that you really have to get it right because when you read a paragraph, it may take you a while and each person may interpret it slightly differently in text. But when you show somebody three images, you're going to get a visceral reaction whether this AI generated image is perfect or is this just not going to cut it for your brand. And so I would say the bar of aesthetic quality, taste, the subjective aspects of what you care about your brand are way more pronounced in images. And one of the things we are focused on is giving that kind of creative control.

Clara Shai: And isn't certainly that's the world we're living in now, is so many of us during every break in the day, whatever time we can steal and at the end of the day when we go home, we're trying to catch up on everything that's happened in the world of AI just that day. And even then it's hard to keep up.

Clara Shih: Yeah, it's a fun ride right now. I'm sure you feel the same way. Every week feels like months worth of progress and development. And what's viscerally exciting for me about the current version is it's simultaneously deep technology breakthroughs that technologist in me really enjoys and learning about new science and new research and new papers and academy and industry collaborating at a level that's just unprecedented. But equally, and maybe the bigger excitement, is there is a tangible end user value, an end experience value of people jumping on these breakthroughs. So this is not like some backend database technology or something. People are actually, I look at my 18 year old son and the last one year, the amount of change in his school life with tools like Chat GPT is incredible. It's just one year, not even a year, and it's already permeating every single part of the world.

Clara Shai: It's so true. It's almost like we've never seen a technology before that is simultaneously changing every layer of the stack all at once. It's an infrastructure, but it's also a platform. It's also that end user applications experience.

Clara Shih: And surrounding all of that, the commercial and the business context of how all these existing companies are going to jump on this change. New companies are going to push the limits of new business models. But you're absolutely correct, cloud primarily played first infrastructure level. The mobile share primarily initially played at the experience level, and over time they went through all layers, as you just said. This one feels like it's happening all layers all at the same time, and that's both overwhelming, but probably why it's got to be even more profound than all those previous waves combined.

Clara Shai: Let's go back in time two years. At this point you were the chief product and technology officer at Adobe, and then something inspired you to leave after working in these company leadership positions at the biggest technology companies in the world for two decades, and you decide to start your own company. What was that spark? And tell us about what Typeface does.

Clara Shih: Yeah, it certainly feels like a long time ago now. It's been a wild ride since starting Typeface, but going back to that moment, as you know in my role at Adobe as CTO, it's amazing company, one of the amazing luxuries I had was my role allowed me to look around the corners to guide our engineering agenda. And back in 2016, which now feels like a long time ago, we actually put a stake in the ground that AI is actually going to be a pretty profound change for creativity, for storytelling, for content businesses at large. And I was involved in launching AI platform for Adobe called Sensei, and that was literally in 2016. And so I've been tracking AI, and the question, Clara, has always been, is this one of those shifts where the big moment is always two years or three years away, or is this actually going to be one of those profound shifts that just gets to an inflection point? And in 2021, as you correctly said, two years ago, I actually finally started getting real nervousness as a individual that there is a big wave coming. And as a technologist, I didn't want to actually not follow my heart, not follow my that conviction about being a flag planter. And white companies like Adobe, Microsoft, Google, Salesforce are all doing amazing things with AI, I'd want to continue to invest. The flag planter in me, I was like, " Okay, there's maybe a radical new way to think about building a brand new set of applications that reimagine how companies of all sizes can deal with content, how they create content, how they personalize content, how they deliver and measure their success." And so the genesis was a view that AI can be one of those 10x if not a 100x multipliers for the world of enterprise content. And so I left in March of 2022 last year, started this company. And Typeface, it's interesting to sit here today when so much has happened, with all AI models, foundational platforms, and every company is now exciting and doing amazing work. When we started, Clara, none of these things had happened. So when we started Typeface, we had the conviction and the idea that the world of content is going to get transformed in next five years. And we started with one simple idea, which is still what holds the base of what Typeface is, is as large companies like Microsoft, Google, Salesforce, innovate and invest in building foundational platforms, like the text platforms like GPT- 4 or imaging platforms and AI breakthroughs, there is going to be a need for companies and enterprise customers to be able to personalize these foundation models into their own brand, into their own customer audiences, into their own way of business applications. And that's the role Typeface wanted to play, is to personalize these large AI platforms and bring them to each enterprise within the context of their existing applications and workflows. And so that's the mission, that's the journey we are on.

Clara Shai: You talk a lot about flag planters versus road builders. What do you mean by this and when do you need each type of person at a company?

Clara Shih: The technology space, as you know, the world changes very rapidly and certainly every decade or so there is a big step function change that comes along, whether it was the web, then it was cloud, then mobile, and certainly now we see we are in the midst of the next massive shift with AI. And I think what you want to make sure, especially in large companies, but as certainly as a startup that we are, is you have to have enough people and cultural DNA who are motivated to go where nobody else has gone before. There are different ways to build businesses, but when I talk about flag planters, it's people who get motivated by a long- term vision. They have a glimmer in their eye of something that's going to be real one day, and they are actually not afraid to go down a pathway. There is no existing road, there is no existing playbook, there is no existing business models. That's the flag planter of the world. Then road builders is once a business is off the ground and established, then you need this operational excellence, like really deep day in, day out, how do you grind it through? How do you build it? How do you optimize a business? How do you serve customers for customer success? And I think what's amazing about big companies is they have scale and they have reach and they have a lot of road builders in a way, what they don't have typically is enough of those flag planters to go in new areas. And I certainly, I'm passionate about that part of the software business, is going into new areas. Your question though is very deep and valid in that most big companies, ultimately their downfall happens when they don't have culturally enough space for new flag planters to come in, push the limits, push forward on new businesses that may be disruptive to their existing businesses. And I certainly have been fortunate at Microsoft, if you look at the transition to Cloud with Azure, I was involved in that. At Adobe as we transition to the cloud business. But frankly, I look at company like Salesforce and what you have done, it's not only pioneered the SaaS transition, which itself was a fundamental new kind of a flag planter moment, if you will, the whole industry is now on that. But clearly where we are today is you're on your own version of the next flag planting journey.

Clara Shai: Now when you're interviewing someone, how can you tell whether they're one or the other?

Clara Shih: Yeah, that's a multimillion dollar question. And first of all, I don't know that I have a perfect playbook for you on that one in terms of precisely knowing, but I will say there are a few things you always can tell. First of all, that deep intrinsic passion, whether you're an engineer, whether you're a product person, whether you're a marketer, of are you more comfortable being with the status quo and just doing a optimization and tweaking? And there's nothing wrong with that. But every now and then when you're interviewing, you will meet somebody who has just unbelievable passion and conviction in their ideas. And sometimes they're not even backed by any data, which is often the case with the flag planter journeys, is it's their conviction and in their passion. And they may be right, they may be wrong, but it's that DNA and that passion. So I first of all look for does the person really have that unwavering passion for technology, passion for breakthroughs, passion for willing to be misunderstood for a long, long periods of time? And I know that sounds cliche, but that really is the big thing. Because in flag planters, there is no straight line to success. So you really can't assess them on the accuracy of their ideas as opposed to actually their ability to create ideas and actually have the conviction to chase them in absence of data. And I think that is something I look for, is do you have that conviction? Do you have that passion? Do you have that willingness to go down dark alleys and be willing to be misunderstood and be wrong, frankly? And so I think that's one thing. I would say the other thing in technology industry I do look for, are you really passionate about being on the bleeding edge? Because the rate of change, and as we sit here and talk today, there isn't been a moment like what's happening with GenAI, which is the rate of change is just incredible. And so if you are not really intrinsically passionate about staying on top of it every day, it's going to get overwhelming pretty quickly. And so I think I look for that conviction, passion and ability to deal with a lot of ambiguity and willingness to be wrong.

Clara Shai: Yeah, it's what Reid Hoffman calls the accurate contrarian, but then add to that someone who can move really fast too.

Clara Shih: inaudible Speed. I was going to say that's actually, I'm glad you said that. Is the other characteristic you have to look for in flag planters is incredible adaptability. Can they adapt every single day, every single moment to a new reality around them? Because, and weird thing about this is having the conviction, but then on a moment's notice is being willing to throw your convictions and adopt a new reality. That's kind of the weird contradiction about flag planters.

Clara Shai: That's a very powerful idea. Are you able to talk about any of the customers you're working with?

Clara Shih: Yeah, no, it's been fascinating. Obviously in the grand scheme of things, we are still in the early days both as a company, but industry as a whole. But it's moving fast. And what's gratifying, Clara, for us is not just the amazing product and technology that the team has been able to build in a pretty short order of last 12, 15 months, but we started working with a large number of customers across retail, CPG, travel and hospitality, financial services, and we focus on enterprise customers, which is the bread and butter obviously for you and Salesforce. And the reason we started there is one, they typically are the ones, if it's a foundational shift, these enterprise customers will adopt these if there is a real ROI, real value, and they actually have the wherewithal to see the long- term cut of the shift. So we started there. I can give you a few examples. Obviously companies like LG Korea, they are an exciting example of somebody that's using Typeface and they are completely re- imagining their content workflow where how they do product photography for all their consumer electronics like televisions and refrigerators and washing machines. And it used to be that the workflow was extremely painstaking and costly. Go into studios, do photography in one country, and as you know, they serve 90 countries with their products. And so culturalizing and localizing the content to each market with speed while reducing cost was a huge problem. So this content velocity, mixed with personalization, was their big challenge. And so they are using now Typeface to do extremely high quality product photography and generation of content to deliver into 90 markets while achieving speed and reducing cost dramatically. So that's kind of one end of the spectrum. And we have many other brands like that in retail. Now, financial services and B2B customers, in fact, that's some work we are doing together with Salesforce, which is quite exciting, is workflows like email marketing. If you are a exact target customer, you're using a lot of email marketing and campaigns, which is a bread and butter for a lot of B2B marketers, we are now integrating Typeface deeply with Salesforce Marketing Cloud. And a customer can write in line, use your marketing applications like Salesforce and use Typeface to turbocharge your content creation inside those workflows. And so that's just few examples of the kinds of things we are doing.

Clara Shai: And so how does it work? So a company like LG, they've got their brand images and past campaigns, past photo shoots, and so you're using that to ground or train the models that have also been trained on a public corpus of images to generate new brand campaigns and new images. Is that kind of how it works? Can you share more detail?

Clara Shih: Yeah, no, you got it. And that's exactly how it is. I would just say we do three things. First, we have a metadata layer called Typeface Graph, which allows companies like LG or any enterprise customer to point us to their existing data, their existing content management system, existing asset systems. And we essentially train Typeface engine to learn your voice as a company. It masters your voice, your brand, your customer profile, how you do AB tests, how you measure efficacy of your content. So graph is, think of it like the underlying substrate. Then we do what you said, which is we ground and fine tune some of the best Inclass models. And we work with OpenAI models. We have a partnership with Google with their AI models like Imagine and Palm. We are working with open source models like Stable Diffusion and Hugging Face. And so we bring all of these, we train on that corporate corpus that the customer gives us. And the resulting model in effect is now uniquely trained to speak your language as a company. It knows your customers, it knows your products, it knows your customer journeys. And then the last thing we do, Clara, which was in fact the inspiration came from force. com for us, I've been a long time admirer of what Salesforce did with the ecosystem play. And so for us, flow is a low- code, no- code workflow engine for generative AI content. And so we let companies create rapidly new kinds of workflows, do Instagram campaign, do a Google display ad campaign, do a LinkedIn recruiting campaign, and all of those hide all the underlying complexity of these AI models while still letting you stay in the flow of your work.

Clara Shai: Now, does each of your customers get their own fine tuned version of your models or are you able to have a shared model?

Clara Shih: Yeah, no, that's a great question. And I'm sure we'll talk about, there are a lot of considerations around brand safety, copyright, the lot of legal aspects that come into play here. We have taken, Clara, very clear position from early on in our journey, is every single customization model that we generate is unique to the enterprise customer and they own it. So we don't co- mingle models, we don't try to do shared learning across, which can be tempting and can be powerful, as you know the big foundational platforms do allow that, but our entire value to the customer, is it's completely isolated to you. And you own your data, you own any weights and embeddings we generate, are owned by you as a customer and we don't mix and match those with any other customer.

Clara Shai: So we've talked before about training a model from scratch versus fine tuning one of the many and fast- growing models out there versus not even changing the model weights and focusing on the embeddings and some combination of thereof. It sounds like you're not training anything from scratch, and so you're using these foundation models that have been trained on a public corpus of data, and of course there are a lot of questions around the copyright and the authenticity of that. And so how are you addressing that for your customers?

Clara Shih: Yeah, no, that's a great question. First of all, we do mix and match different techniques as you said. If certain use cases require just a few short learning using standard APIs, and if that's good enough, we start there. If certain use cases or customer workflows require us to go custom embed inside the training pipeline, we do that in certain image generation workflows. So first I would say as a startup you have to be ruthlessly pragmatic. And so we are not really religious about we must train or we just won't train. It's actually what is the use case that we are trying to enable require. And today we are actually along the full spectrum, as you said. We do grounding, we do customer meetings, we do few short learning, we do techniques like lank chain to stitch together operations. So we do all of those. I think to your more foundational question though, no pun intended, which is if you're trained on all these public corpuses in the foundation model, enterprise customers deeply care about things like copyright and brand safety and whatever I'm training, can I actually be safe in using that model? And so the two things I would say, Clara, we have one relatively good answer as a company and then one area I think as an industry nobody has solved that we are hoping collectively to solve. The one we have good answer is with our customization layer, the personalization engine that I talked about, it does give us ability to go to a customer and say, " You give us the proprietary content set and data set that you own, that you have the copyright for your brand, your releases for images and text. And we will ensure that the training run is done only on that data." So if you actually give us as a company your content, we can narrow the funnel if you will. So we call it differential training where we'll only do incremental training runs on your corpus. We'll make sure any offensive data sets or things you don't want are excluded. So that's one thing we today do. That's actually better than what a lot of other companies can do today. And so that's a good answer, but it's only a partial answer because as you said, there is a whole slew of regulations, laws, every country is going to, as you just saw last week, there was one of the first cases around AI generated content, can it even be copyrighted? And so the way we look at right now, Clara, is, be a very trustworthy partner for the customer we are serving. Be direct with them around what we can solve for them, like this custom training, custom proprietary things. And then we also very explicitly tell them that here are all the issues that have not yet been solved that are yet to be figured out as a industry as well as governments.

Clara Shai: What do you say to people who approach this with a fear that AI is coming for our jobs?

Clara Shih: This is one of the things I'm excited about, which is as AI models get powerful and they will, I'm not a believer that AI will replace human creativity when it comes to images, videos, and other forms of creative expression because there's always going to be that human element of me being able to tweak that image. And so what we are doing at Typeface is building a first of its kind collaboration engine, if you will, between a human creative and AI creative and let them collaborate together to get that output.

Clara Shai: In the text modality, localizing means translating it into that local language, respecting the grammar, the sentence structure, the word choice, and it could vary by country or within a country even within the same language. What does localization mean for images? On the one hand, it's universal, like we talked about earlier. On the other hand, what are those considerations that your team is thinking about in terms of making something relevant in India versus in America versus in Canada?

Clara Shih: Not only across different cultures, different geographies, but also different industries. If you look at B2C heavy industries when you're marketing food products or clothing, apparel, very different kinds of challenges around what kind of emotive images, whether the human model and the expressiveness and the cosmetics, clothing, things we all find very personal. The images have to appeal to you not just in the attributes of the product that you are selling as a company, but in the experientially how people will feel it's going to impact their life. And so I think they have to tell a story. I know it's a cliche, but the beautiful thing about this moment we are at GenAI is it's finally getting computers out of this, the rote automation role they have played over the last 4, 5, 6 decades where they've been cheaper, faster, better in doing these automation tasks. And we are finally teaching computers to learn the human world around language, which is what the text world did. Now we are starting to teach computers the human aesthetic, things like taste, what different cultures value, how you show television image in a living room in India for a customer is going to be very different than how you show it in Korea versus how you show it in North America. And so you're almost starting to teach these subtle cultural nuances that have been baked in over decades if not centuries of this kind of visual communication mediums. Now we are starting to make our AI systems and computers actually be aware of that cultural fabric, if you will.

Clara Shai: How far can the specialization go? Can we get to personalized content generation for segment- of- one?

Clara Shih: I think so. I think that's always been the holy grail. As you know, it's the hyper- personalization down to the one. Obviously you have to be mindful of the negative side effects of that also from a privacy and ability to cut out. But I do think so, I think not only in the ability to generate hyper- personalized content and deliver that to the right user at the right moment, which I do think it can get to not only the size of one, but even for the same user. Clara, when you are at Salesforce office during the day, you are a different person than when you're at home, relaxed in New York, in front of your TV. So even it's even more than audience of one, it's audience of one at a specific moment, what is the brand voice you need to hit for that person. And I do think that's happening, but I would be remiss if I didn't talk about the other side. We can't solve all these personalization problems with these AI models without also being mindful of the privacy, authenticity, how do you actually control for unintended consequences of these? One thing I worry about quite a bit and we think about is especially if you think about kids that are going to grow up in the next decade with these AI systems, a lot of the information they consume in their schools and in their inaudible how they learn, is increasingly going to have AI voice play a role in that. And how do you make sure that that is actually correctly curated according to your values as a family, as a society, as a culture. So that's going to be a pretty interesting set of challenges to solve around trust and safety. I know you guys actually talk a lot about that as well in terms of your focus.

Clara Shai: Well, yeah, and we see this playing out all over the world is this fracturing of worldviews and in part enabled by this over personalization and people living in different realities from one another. So it's a very interesting balance to strike.

Clara Shih: Yeah, no, and that's going to be the challenge. And I do think, one last thing I'll say, one of the reasons I'm excited about the work we are doing together, this is such a big shift, Clara, that I don't think any company, no matter how big or small, certainly no small company like us can just go it alone no matter how smart we think the team is and how big our ambition is. A lot of these problems are going to require not only cross- industry collaboration and partnerships to align on standards, align on the common AI ethics, AI biases, but it's also going to require working with governments and regulators. And I think this is going to be, you said something earlier, this is playing out at all layers of the stack infrastructure app and UI. I think it's also going to simultaneously play out at public policy, legal frameworks, societies, all of that is going to play out at the same time.

Clara Shai: You grew up in Pune, India and came to the US and started your technology career. How did some of those earlier experiences shape you as a technology leader and as a flag planter?

Clara Shih: Yeah, that's a very back to the basics kind of a question in some ways. I think as I look back, and I was obviously very fortunate growing up in India, as you can imagine, access to technology now, it's actually much, much more different with smartphones. And they have LeapFrog and the economy has done amazingly well over the last couple of decades. When I left India more than almost 25 years ago but when I went to school there, I was fortunate enough to actually be in some of the best education institutions. There're surrounded by amazing people. But I remember, the first internet connection, I was one of the early lucky ones who had access, but it was like a 14 Kbps modem that would die every few minutes just to get to a Telenet session or a text- based gopher. And I know most of these probably mean nothing to your younger audience of this podcast, but no I think, so if I look back though, there's still a consistent thread. There was this excitement around what the worldwide web and TCP/ IP represents when I was graduating in my undergraduate program. And it felt like there was this amazing new world, even though you are staring through the text lens. And actually there is a weird parallel here, Clara, as I think about it. The first wave of internet innovations I got exposed to was all text- based. It was all text- based prompts. The text- based Telenet sessions, inaudible-

Clara Shai: Sounds familiar.

Clara Shih: There was no Netscape, there was no visual browser. And then when market and team built Netscape, you suddenly saw a whole different level of innovation vector around, " Whoa, this can be a completely different medium to express your ideas." And then when Amazon, eBay, Google, so it's just going to be the same pattern in my mind. I know I'm connecting from back then to now. I do feel like we are about to go from the Telenet gopher era of GenAI text- based models to now the Netscape moment of GenAI models, which is much richer visual storytelling. And then it's up to companies like Typeface and Salesforce and us to imagine what the next set of applications will be built on this fabric.

Clara Shai: That's so true. I couldn't agree more. Well, on that note, you've talked about three big changes that you think are being facilitated and accelerated by AI. Can you talk to us about those?

Clara Shih: Yeah, and you'll have to keep me on time because I can talk about this, I get excited about this. This is more at a foundational level, if you will. What is this AI shift represent at a fundamental level in the role of computing in our lives? If you take a decade long view or a couple of decades out, first to me is the role of computers is going to change from just these automation number crunching assistance that they have been with us for the last 3, 4, 5 decades to machines that start understanding the world around us. When computers start having vision and language and speech and ability to hear, I think they start understanding our world not just through the pure productivity lens, but through all these cultural and social aspects. And so first big change I think over the next couple of decades that I don't think we truly grasp right now is the role of computing machines is going to radically change in our lives. I know it seems hard to believe given smartphones have made such a big impact in the last decade and before that the web has already had a profound impact. I think the role of computers is going to profoundly change in our next two, three decades of humanity because they will start understanding our world around us in a much deeper way than they have ever done before. That's point one.

Clara Shai: It's almost like the Internet of Things. It's finally found its time in the sun because the compute has finally caught up to all of that data that has been being pulled in.

Clara Shih: Actually, it's funny you say that. I had not thought about it in those terms because the first generation of Internet of Things, I really went after more industrial automation type of use cases. And in some ways maybe now this is the real Internet of Things coming into our real world. So it's actually the real things-

Clara Shai: The actual Internet of Things.

Clara Shih: In actual real world. So that's going to be pretty profound. But the bigger exciting thing that goes hand in hand, hand with that in my mind, is the interaction model between us as humans and machines has been evolving over decades. You know obviously we started with punch cards and depending on how much I'm dating myself, I wasn't involved in punch card gen, but the keyboard and mouse, then the touch screens, smartphones brought that direct manipulation. When you get to the world of natural language as a way to communicate with machines, I think the role is going to reverse dramatically. We have spent last 40 years, and you and I have been part of the industry where the entire world was taught programming languages and how to teach humans to speak to computers in weird languages that machines can understand. I think we are about to see a world where machines start learning how to speak to us in our language. And I think you're already seeing that and that role reversal will have such a big impact because the friction of knowledge sharing and learning is just going to collapse dramatically because now natural language becomes the language of machines. And I think you are seeing already there are pronouncements, like the world of coding as we know, or programming, is going to go through a profound change in the next five years, 10 years because once these machines learn our natural language, they can start expressing their compute power through human- centric expressions. And so that's going to be very profound. And then the last one, which is probably where you and I have probably focused a lot in our day jobs is if you have that kind of a change in underlying infrastructure and interaction model, the world of enterprise software and the businesses at large is going to get completely rewired. And this to me is what got me excited about starting Typeface is the last decade I would say was all about moving to the cloud, the SaaS transformation that obviously Salesforce led, the transformation with data with the big data architecture, Spark and Hadoop and everything. The next decade of rewiring the enterprise is all going to be about storytelling. How do you empower every single employee, whether you're a salesperson, a marketer, a support person, finance person, even a legal person, you are actually going to be now empowered with amazing storytelling tools to express your ideas without being encumbered by technology. And that's just going to change what notion of productivity means inside these businesses.

Clara Shai: I couldn't agree more. We have to rewire the enterprise. We have to rewire education. We have to rewire our government and our legal system. We have to rewire everything.

Clara Shih: By the way, education and government. I'm glad you said that. I didn't even touch on those. Those are massive. Think about it. If education of next generation of kids gets completely rewired with these technologies as it will, now imagine what it can do, hopefully in a positive way to accelerate education and get access to latest education to everyone on the planet.

Clara Shai: That's right. And it's up to us because these technologies are not inherently positive or negative. It's what we choose to do with it.

Clara Shih: I agree.

Clara Shai: I feel that weight of responsibility and I can tell that you do too.

Clara Shih: Yeah, and that's what makes it exciting. There's a lot of potential. There's a lot of amazing things possible, but there are a lot of hard technical challenges, hard business challenges, hard policy challenges, and I do think we have the capacity, both individually as companies and then collectively as ecosystems and working together, to tackle these. I'm sure we won't get all of these right, but I am excited about that next decade to go tackle some of these big things.

Clara Shai: What an exciting time to be in technology and what an exciting time to be a flag planter.

Clara Shih: Yeah, no, it's amazing time to be here. An amazing time to be with you on this at this kind of a seminal moment.

Clara Shai: Well, just incredible. Thank you, Abhay, for the conversation today. Congratulations on your recent round of funding. Salesforce Ventures is delighted to be a part of your company and to be a partner. Look forward to seeing what's next from you and the team.

Clara Shih: Thanks, Clara. Likewise. Enjoyed the partnership and it was a fun conversation. Thanks.

Clara Shai: Abhay, thank you for your tremendous insights. I learned so much from this conversation. Just a few of the takeaways. First, the cognitive complexities that images introduce versus text. Two, what it means to make images culturally relevant for different countries. Three, how should we address the copyright issues surrounding images and more broadly creative assets? And four, we're finally seeing the coming of age of the Internet of Things. That's all for this week on Ask More of AI, the podcast at the intersection of AI and business. Follow us wherever you get your podcasts and follow me on LinkedIn and Twitter. To learn more about Salesforce and AI, join our Ask More of AI newsletter on LinkedIn. See you next time.