In this episode of In the Blink of AI, we meet Angela Shi, the CEO and co-founder of Empathetic AI, a Sydney-based startup revolutionising the tax industry with artificial intelligence. Angela shares her journey from being a finance professional with over 14 years of experience to becoming a first-time founder tackling complex tax challenges through their AI assistant, Luna. She discusses how Luna, a domain-specific large language model trained on decades of Australian tax legislation and rulings, assists tax professionals by simplifying complex scenarios and increasing productivity. Angela delves into the technical aspects of building Luna, the importance of human expertise in AI development, and how Empathetic AI addresses data sensitivity and compliance. She also opens up about overcoming personal challenges as an immigrant and introverted female founder, offering valuable insights into entrepreneurship, continuous learning, and finding meaning in her work.
Transcript Synced · click any line to jump ▾
Georgie Healy: Founders scale faster on Deel. Set up payroll for any country in minutes. Hire anyone, anywhere. Get visas handled fast and get back to building. Visit deel.com/dayone. That's d-e-e-l.com/dayone.
Angela Shi: I have my own struggles. Of course, everyone have their own struggles, right? So Entrepreneurship is not for everyone, but if you choose to be an entrepreneur, chances are along the journey, the valuable lesson you've learned is much, much, much more valuable than, you know, the downside you experienced.
Georgie Healy: Hello and welcome to In the Blink of AI, where I speak to the brightest AI startups and innovators each week. I'm Georgie Healy, and this week I'm speaking to Angela from Empathetic AI. She's the CEO and co-founder. There is no way I would've been as excited to do this episode if I hadn't seen Angela on the stage earlier this year. She presented as one of the finalists of over 100 AI startups for the CSIRO AI Sprint. It wasn't just her amazing power suit that really drew me in. It was the fact that what Empathetic is building is such a clear and incredible example of AI that needs to be built. Tax, so boring, so dry. I apologize to anyone that's into it, just not for me to go deep and learn how that product works. But in this case, I could speak to Angela about this for hours. What Empathetic does is it leverages a model that they've created in-house called Luna. And Luna's a large language model. You type in a prompt about very, um, complex tax questions as a tax professional, and it seeks out, um, or it leverages almost 100 years of tax legislation, tax literature to give you an answer. And you can imagine the amount of time and effort savings that is on tax professionals where they can use that time in a more meaningful way. So this episode is super cool, super fun. I love Angela. I will be always watching what they do because even since this episode was recorded, they have just been doing incredible things in the ecosystem. They're constantly building, constantly moving super, super fast. So I know you'll enjoy it. Thanks so much for listening. Hello, Angela. Thank you so much for joining me today for In the Blink of AI. I'm really thrilled to be able to chat to you and perhaps we could kick to kick things off by you telling me about Empathetic AI.
Angela Shi: Yeah, thanks for having me, George. My name is Angela Shi. I'm the founder and CEO of Empathetic AI, and we are a Gen AI startup based in Sydney. Empathetic AI was founded in February last year, so it's been, you know, more than an 18-month period of time journey for me being a founder, a first-time founder. So it's such a challenging but rewarding journey. So I'm super excited to be the guest today here and to share my two cents with the audience.
Georgie Healy: Amazing. Now, Angela, you launched not long ago, but we were speaking offline about how fast you have grown. Actually, the first time I was made aware of Empathetic AI was a bit earlier this year at the CSIRO AI Sprint, where you were not only a finalist but placed in the top 3, and now you have over 2,000 users. What has been your approach to growing so quickly?
Angela Shi: Um, I mean, I, I don't know whether or not it can be called quickly, but I identify the pinpoint deeply in my past, you know, 14 years being a finance professional myself. Um, so I started my career as a junior accountant in a fintech company, and, uh After that, started to climbing the corporate ladder. So, used to be the junior accountant, senior accountant, finance manager, financial controller, and eventually become CFO. So luckily, all the companies I worked at, they are all technology-driven, heavily technology-driven. So that's why I personally, I got very fascinated about technology itself. But having said that, Because my role is finance, so I deeply understand, you know, finance in fintech or finance in financial services is heavily, heavily data-driven. And we were at the stage of automation but not reached the level of intelligence. So I always think that what if I can use some, you know, intelligent technology to really tackle some complexities in the finance world. So that's the initial inspiration of creating something using generative AI to help finance and tax professionals. And I guess because I really understand the pain points, so I naturally get along with my community members. So it's been a journey. So we actually spent quite a bit of time at the backend, you know, building the product. But as soon as the product has been launched, We were very lucky, fortunate to have a group of, you know, domain experts around us being the volunteers helping us test the product and continuously providing the feedback to us. Keep iterating the product very, very quickly by having a small team. So, I intentionally try to maintain a relatively small team so that we can move faster. And also I do believe that, you know, the AI application in finance and tax need a lot of humans, you know, expertise and domain experts to get involved in order to kind of, you know, really create something they need. And I believe that, you know, naturally, because I had a lot of interactions with the domain experts. So that's why relatively quickly to build a community, but we are still growing.
Georgie Healy: Yeah. Amazing. You touch upon a small team, but this clear, deep domain expertise for a very vertical AI solution in the tax space. Stalking your website, you mentioned that your product is for complex tax scenarios. What's a complex tax scenario?
Angela Shi: Yeah, so that's a great question. So initially the inspiration for me to create Luna, the AI tax copilot, is because when I was working as a CFO in the corporate. My last job was Apex CFO, so I actually covered different jurisdictions. So one of my job, you know, to— the CFOs do two things. One is increase the revenue and the other thing is reduce the cost. So for cost reduction, part of the job is to optimize the tax planning across different jurisdictions. So during that process, I worked closely with tax advisors. So I found, you know, this experience is quite time-consuming. And because we have a vast amount of data we need to leverage in order to have the clear understanding of, you know, each country's tax requirement, tax legislations and laws to really optimize the tax planning. So that's why we start using Luna to tackle this complexity by helping the tax professional do the research 'Cause before, in the past, they need to manually do the research for the tax legislations and laws. But having said that, Luna can also answer some of the simple questions like the data retrieval. For example, you ask Luna, "I earn that much and what is my tax, taxable income?" And stuff like that, just simple scenarios. But because our business model is B2B right now, and we target the tax professionals, which they already have a lot of tax knowledge in their mind, but it's just time-consuming for them to do the research to find the relevant legislations and laws accurately. So that's why we said complex scenario. But having said that, Luna can definitely help the individuals like you and me to kind of, you know, help us better understand our tax liabilities.
Georgie Healy: Yeah. Amazing. And you've beat me to my next question. In a few sentences, maybe you can let everyone know who or what is Luna.
Angela Shi: Oh yeah. So I haven't introduced Luna. Yeah. So Luna is the text copilot, or we call it text assistant, AI text assistant. Personally, I like Luna, this name. So that's why we named it Luna and then we create AI persona avatar for Luna. Luna can help you. Luna is like your smart tax assistant, but minus stress. Then it can answer all different types of tax questions. At the backend, it's more like a vertical domain LLM, vertical domain large language model, specifically for taxation. We actually trained the entire ATO's legal database since 1936, and it includes all different kinds of legislations and laws and rulings and, you know, cases. And then Luna can help you not only just to provide the answer, but also, you know, provide the reasoning behind. Because a lot of the time when you reach out to your tax advisor, not only you want to know how much tax I need to pay, but also you want to know how can I optimize my my tax, so obviously without breaching the law. So that's something Luna can help really provide the reasoning and as well as the references check to make sure that this is the best advice you can have and also is backed by the legislation and laws.
Georgie Healy: Amazing. Now when you say that, it sounds almost like it's B2C, but actually this is for tax professionals, right? So maybe you could talk about the relationship that tax professionals would use Luna and what kind of scenarios they would do. So—
Angela Shi: Yeah, absolutely. So actually tax is a broad definition. So within tax professionals, we have tax lawyers, tax accountants, tax advisor, or tax manager in the corporate. So we have different professions within this ecosystem. So each profession, they might refer to different, you know, sections in the legal database. For example, tax lawyer might refer to the legislation laws and tax accountant might refer to the private rulings, let's say. So that's why based on their professions, we can actually customize the hierarchy of the data retrieval to provide the most accurate, you know, references to back them up. So, and also for the individuals like you and me, and then if we're uncertain about our tax obligations and and we can simply just ask Luna. And Luna compared to the general model like ChatGPT, Luna goes very deep because it's a vertical model for the taxation only. But having said that, because the solution is scalable, so which means that we can expand to financial planning, insurance, and others. So all the industries across this industry similarity is, is all about financial optimization. Mm-hmm.
Georgie Healy: And I find this fascinating. Maybe we can dive a little bit deeper into how you built Luna. You talked about, you know, using legislation and laws since 1936. What kinds of data exactly is that? You know, I'm desperate to know the inputs that would go into a model as sophisticated as Luna.
Angela Shi: Yeah, sure, absolutely. So actually, believe it or not, there are some public information on ATO's website already. So both you and me and all the public can have access to that data, so it's all listed on ATO's website. And also ATO have the ATO community where you can ask questions, so there are certified experts answer your questions. So all this data are publicly available, but we are lucky enough to collaborate with some universities as well so that we can see, for example, some professional bodies like CPA or CA, And then we can take reference to the standard questions and answers and compare, you know, Luna's output with the standard answers for certain questions as a benchmark. Because the interesting thing is there's no certain benchmark existed in the market for the vertical domain. So that's why we try to build the first vertical domain evaluation mechanism by comparing, you know, the standard answer. answers certified by the expert versus Luna's output. By using that way, we, we kind of improving, you know, Luna's output step by step.
Georgie Healy: I can't tell you how many people, myself included, um, that just find tax, A, very complicated and, you know, we just don't know where to start and B, we just don't want to learn. So it's really incredible that, that you broken that barrier, you know.
Angela Shi: I think, You know, a lot of people, including myself, even I've been, you know, a finance professional for many years, most of time I feel, you know, tax and finance is boring. I have to be very honest. So that's why I think certain things, if we can leverage AI to do it, not only we can free up our time and some repetitive work, and also in the meantime, it can allow us to focus on more strategic work and more interesting work, right? So, you don't want to spend 5 hours doing the research on the legislation just to find a reference. So, for things like that, you can literally just ask AI to do it for you. So, by having such, you know, AI assistant tool, it really just, you know, just free up your— increase your productivity and make you focus on more strategic work.
Georgie Healy: I completely agree, and I'm so glad you didn't get offended by that.
Angela Shi: No.
Georgie Healy: You discussed data desensitization on your website. What is that?
Angela Shi: Oh yeah, so finance and tax, this industry naturally, a lot of data very sensitive. So, and as a professional, you wouldn't be able to, you are not allowed to disclose such information to a third party. So obviously when you ask the question, you can ask like ABC company instead of mentioning the real company name. But if you really want to, you know, do a bit of a copy and paste and just test Luna's output, and then there's a desensitization function that you click a button. So certain sensitive information is going to be removed and replaced by generalized information. Like empathetic AI, if it's sensitive information and then I click a button and it'll be replaced by ABC Company, for example. Ah, yes.
Georgie Healy: As an ex-consultant trying to share my accolades, I would have to say worked for Large automotive company. So, okay, I completely understand what that is. Amazing. How nuanced does this model have to be? We had a previous guest who was working on an AI model with the complexity of cognitive wellbeing, which honestly sounds just so incredibly complex in terms of different inputs and different scenarios in order to create that model. When it comes to tax though, it seems almost like there's not that significant range of complexity, or am I missing something?
Angela Shi: Well, I would say it's quite complex. So that's why it took us a bit of a time, but I intentionally create something complex. I mean, at the backend, but at the frontend, the user interface is relatively, it's super user-friendly. So because I don't want to use one complexity to replace another complexity. So I want a user to simply just use it and then get the perfect output.
Georgie Healy: That's it.
Angela Shi: But at the backend, we actually, we've been doing a lot of work. So the work is on us and we can help the user to free up their time. So that's what we've been doing. In terms of the complexity, yes, because text is a combination of data wording and as well as calculation. So there's another layer on top of the purely just wording. If we only talk about the legislations and laws, that's purely just language, right? So, but on top of that, you got to have the calculation right, because a lot of the time when you ask your tax advisor, so you actually want to— Yeah. Them to have a plan, a strategy, how they can help you optimize your tax planning. So it's not only just advice, but also a bit of calculation as well. So that's why we spent a lot of time to test Luna. And obviously this is the— I wouldn't say it's perfect yet, but it's improving because there's another layer on top of the language itself. Yeah.
Georgie Healy: Incredible. But considering you're the CEO, not the CTO, it's incredible to be able to dive this technical with you, Angela. What happens— and I don't wish this upon you— but what happens if legislation or regulation in Australia changes? How often does it change? What does that do to your model if that happens?
Angela Shi: Yeah, great question. So Australian taxation is actually on a case law basis. Which means that there's new case every single day.
Georgie Healy: Oh, God.
Angela Shi: Yeah. So part of the reason you can use the advanced technologies such as AI to reduce your complexity, because the AI model can capture the real-time changes. So on ATO's website, you can see the rulings might get updated on weekly basis and different sections gets updated. Frequently. And so that's why the AI model can really pick up the updated information. So that's a benefit of using, I guess, the advanced model so that you don't need to keep your knowledge up to date all the time. But it's essential for the professionals to keep their knowledge up to date. But you can leverage the tools so that it can capture the most up-to-date information for you.
Georgie Healy: Amazing. And last one on all the metrics and data. How do you approach missing data?
Angela Shi: How do I approach the missing data?
Georgie Healy: Yeah. If you've got a query or a question from a tax professional that asks Luna, for example, and it's not in the existing model, how does, how does the model approach that?
Angela Shi: Well, so that's a great question. So we wouldn't say that we covered everything. But obviously, because at the backend we can keep feeding the model with new information and obviously we don't wanna confuse the model in the first place. So we do a lot of human-in-the-loop evaluation first to justify, you know, to evaluate Luna's answer, whether or not it's reached a certain level of the accuracy. And then we keep feeding Luna with the new verified data so that it keeps improving the algorithm at the backend, keep improving the quality of the output, I would say. But back to your question in terms of missing data, like ChatGPT, like the general model, it wouldn't cover every single thing on the internet. It works like a human being, you know, text advisor. If you ask a question to 10 tax advisors the same question, so they might give you slightly different answers as well as a human being. So a lot of questions in the finance or in the tax space are open questions, like how can I optimize my planning? So it's open question. So it's really depend on how Luna or even the human being, you know, tax advisor interpret this question and try to search the best answer or try to get the best answer for you based on your scenario. But of course, for us as a product producer, we try to keep feeding Luna with more and more data, verified data, I would say.
Georgie Healy: Incredible. You mentioned human-in-the-loop verification, super interesting concept. At what stage can you phase that out, or does one never phase that out? Like, does it depend on the product?
Angela Shi: I think finance and tax naturally, this industry have a very high barrier to entry because we all know financial services or fintech and tax practice, they are all license-based professional services. So I think naturally they have very high barrier to entry. So that's why they need higher accuracy. And then also you need to really spend time and effort to build the trust. So that's why we actually took a lot of time to train Luna and then to build the trust and then to get the input from the domain expert. So they are part of Luna's creation as well. So we borrowed their wisdom and transformed their wisdom into the backend algorithm, backend coding. I think for industries like finance and tax, And perhaps in healthcare, such data-driven industry as well, I think human expertise or the domain experts plays a very, very important role. It might be even more important than the technology itself because even we have the fundamental model, but how you translate people's knowledge into the coding, there's a big gap between business leader and AI technologist. So yeah, so that's why for me, my background is the business side. So as you mentioned that I'm not a CTO, I'm a CEO, so that's why I also try to bridge the gap. But I think for industries like finance and tax naturally need a lot of the human's input.
Georgie Healy: Yeah, it sounds like you're trying to get the best quality answer and not just the most technically driven answer or— you know, we must only use the model, which sounds perfect to me as someone that wants to, you know, if I'm searching for a solution in tax, I would prefer accuracy over, you know, a cool tool. Maybe text-to-image can have 5 fingers on one hand and 6 fingers on the other hand, and it's not that big a deal, right?
Angela Shi: Yeah, I mean, yes, certain industries, there's a Higher compliance requirements, I would say. So it's not only just product building, it's all about compliance, regulation, and legal as well. So you need to meet certain criteria. I'm conservative in terms of if we launch something that has not really reached the level it can really help people, I'm very concerned. So that's why we we actually leverage a lot of domain expertise knowledge so that to make sure that Luna's output is accurate enough. But still, I wouldn't say it's perfect. We still have a lot of opportunity, a lot of space to grow, to improve.
Georgie Healy: That's what every founder says, actually. Of course, none of them are perfect yet. And it's, you know, this is called In the Blink of AI because everything's moving incredibly quickly. And, um, it's, it's genuinely fascinating to keep checking in with you every few months because the product has evolved so much every time we talk. Um, switching gears a little bit to your journey, Angela, because it is quite a fascinating one. You mentioned working 14 years, um, already in the fintech space. You've held very prestigious CFO roles. What inspired you to take the leap when as far as I'm concerned, it would have been quite a stable lifestyle. Why did you take the jump?
Angela Shi: Well, that's a great question. So for me, I'm the type of person, I like challenge. I am very proud to say that I enjoy the lifestyle of, you know, living in a challenging but rewarding environment because I'm easy to get bored. But this founder journey is is much harder than I anticipated when I started. So I'll be spending 15 years in fintech, actually, you know, the financial leveraged financial instruments industry, which is quite niche. So I think after 14 years, it was a time for me to make some changes. So that's why I actually quit my job and then went to Harvard to do my leadership program. And then in the US, I met some incredible people and it kind of broadened my perspective to the world that people can be really, really different. And also for me, my personality is that I want to make something that can really have the social impact. And part of the reason I quit the job is that I think I have more control of, you know, the direction of my life because even you're at the C-level, you're kind of acting as a functional role regardless you're a CFO or I guess CEO as well. I don't know. I've never been a CEO before I founded the company, but I was a CFO in a few companies. So, I think I really need, you know, to make some changes so that I can lead my personal life to a direction that I really want to be. And then it can really bring some social impact to the world. Yeah. So that's why Luna start from the tax, but obviously this is a starting point because tax and finance, it have, you know, very massive amount of data. Yeah. So it's naturally very important for training the AI model. But also in the meantime, I think everyone in Australia, if you ask them, do you feel— how do you feel about tax? You know, pretty much everyone would say, yes, it's complex. So it's a topic that relevant to everyone.
Georgie Healy: Yeah.
Angela Shi: But it's so boring that nobody stand out and say, can we do something to really tackle this complexity to help people? Also in a mind, so that's combination of everything. So that's why we created Luna.
Georgie Healy: Yeah, there's definitely like— you would need to be passionate to do, to do what you're doing, right? You'd need to really think that it was going to change people's lives in a meaningful way. You mentioned that the journey was harder than you expected. Is that because you— and I think this is probably incorrect— you thought it would be easy and it was harder than easy? Or was there something in particular that you weren't expecting that's hard about being a founder?
Angela Shi: I think part of the reason is that AI technology itself have a very fast development. So, and then in Australia we are building AI applications so that we're not training, you know, the fundamental model, not like the US. So finding the product market fit is something super important. You can't say because you wanna create some social impact and then you create something that nobody like or nobody use, So that's why you need to be realistic, but also in the meantime, you need to look at the long term to kind of adjust your direction or strategy and try to be very agile. So in the process of finding the product-market fit and iterating the product all the time, getting the domain experts' involvement, and then keep upgrading the product and then maintain the team. So we used to have a team across 3 cities in Australia in the past, but now only 2 cities and then 1 team overseas as well. So being a founder, you're actually an all-rounder. So yeah, I can easily just change my title from CEO to all-rounder of Embed AI. So I like this title and it keeps me busy, but also it's a very rewarding title. Because it made me realize that, okay, so marketing is not that easy and product design is not that easy. So, but I never touched certain areas when I was a CFO in the company. I only do finance. So it's a huge learning curve for me. And also in the meantime, I got to have certain level of the AI knowledge myself when I talk about my product, I got to have a clear mind in terms of how you train the model and how you get to the point, the certain level of the accuracy, the backend algorithm whatsoever. So even I'm not a CTO, but still, so it's a huge learning curve for me. So that's why it's harder than I just started, but within this almost not really 2 years, but more than 1.5 years. So now it's 19 months already. I would say I learned— if I was in the corporate, I, I would spend like 5 years to learn that much.
Georgie Healy: Yeah, you're so, you're so right. Like, the, the founder journey— you're passionate about solving an incredible problem, and you've probably got domain experience in the area. And this is across all founders, but something you really raise that strikes a chord with me as an AI founder in this time, at speed on top of all of that responsibility and managing a team and all the rest of it. You're trying to learn something that is in its infancy in many ways and moving so quickly and is very expensive for compute. So yeah, that's a really great take, Angela, and something that maybe not everyone considers, um, that an AI founder is going through. You came to Australia as an immigrant, you've told me, and self-identified as quite shy, which to me, who saw you on stage, is the first introduction to you I've ever seen before, and was a standout on the stage. Perhaps you could share a little bit with the listeners about not only how you kind of overcame perhaps that shyness and how you got that strength of character along the way.
Angela Shi: Well, I came to Australia as an international student. So yeah, I told you. And growing up, I was such an introvert. So if I can finish the conversation within 1 or 2 sentences, I wouldn't elaborate into like any longer. I was such a shy person. And 'cause I grew up in an environment where being quiet or being humble as a female is something, you know, people really like in that environment, and people don't really like the outspoken, you know, females, and they would think it's too very aggressive. My parents are still in China, but they said that I changed a lot. Yeah, compared to when I was a kid. Of course. So you changed based on the environment, right? So you're living, so Even I was in the corporate, I was a finance professional. I don't need to talk to the customers. I only need to talk to my CEO and my team member and certain people. I've been working with them for a long time. So that's such a familiar environment, comfortable environment. So along the journey, along my founder journey, I really have to break up a lot of comfort zones to really raise my hand, raise my voice, get myself heard, you know, try to engage with like users, domain experts, stakeholders, and try to put myself out there. So, I always said that you have to be comfortable for being uncomfortable. So, that's super important. And the other thing I've learned is that I know it's hard for being a female founder. I mean, so, there's a lot of voices. Yeah. Talking about, you know, female founder and we need support and we need funding. I get that. I actually, I have firsthand experience, but the thing I always want to mention is that we shouldn't have the victim mentality all the time because your energy gonna affect your luck. That's my genuine belief. I think regardless how exhausted or how tired I am, I try to have a certain level of energy when I talk to people because the conversation, you have to make the counterpart engaged to push forward the conversation, right? So, not just because you are tired or you just fight with someone else and then you make people feel uncomfortable. So, I always think that being a founder, you choose a journey that is challenging. but you made the decision by yourself. So it's challenging, but it's more rewarding than most of the journeys. I still feel that, you know, choosing to be a founder or female founder in AI or female non-tech founder in AI is the most important decision I've made in my entire life. So I'm still very, very grateful that I was bold enough to quit my job and went to Harvard for leadership program and then broadened my perspective to the entire world and came back to Australia and then founded Unfed AI. So I have my own struggles. Of course, everyone has their own struggles, right? So it's just entrepreneurship is not for everyone, but if you choose to be an entrepreneur, and then chances are along the journey, the valuable lesson you've learned is much are much, much more valuable than, you know, the downside you experienced. Because a lot of people ask me, are you afraid of failure? So my answer is that it really depends on how you define failure. Is that I shut down the company and then I got bankrupt? That could be defined as a failure, or how do you define failure? For me, I think I already learned a lot along the journey and put aside of my startup, AI startup, I've been doing a lot of other stuff as well. I'm a content creator for AI application in finance in different universities. I've been doing the workshops and get engaged with the finance and tax professionals. So yeah, so I never been so happy for my journey. Yeah. But of course it's challenging. Yeah, I have to admit that.
Georgie Healy: Thank you so much for sharing. But I agree with you. I had a failed startup myself. I created an app. I had a bunch of users, but no one would actually pay for it. And so I felt like a huge failure, such an idiot creating this app, spending all this money on something that I didn't really get product-market fit for and should have done a lot more iterations and questions in the early days. And then it's like, well, I learned how to build an app and I learned how to speak to customers and I became an investor. So it is how you position it, right?
Angela Shi: I completely— that's 100%. I think the judgment criteria have to be very diverse in terms of, you know, being an entrepreneur. So if you say I have no paid users, paid customers, but I have a lot of users, I don't think that it can be called failure because the way you can get users, at least, you know, people interested in this product. So I think if you only look at the returns or the financial return, obviously this is something important. But also in the meantime, if you look at the long term, they might find this very, very useful, very helpful for their lives. And I genuinely believe even I don't do AI application in finance and tech, someone else is going to do it because eventually this data-driven industry is going to be transformed. Yeah, so I'm more than happy to be part of it. So that's my mentality. So even Luna wouldn't be able to serve, you know, maybe millions of people, I might fail, but in the future, I mean, eventually, you know, the workforce is going to be transformed anyways. Yeah. So if it's not by Luna, but by something else. So I'm genuinely happy to be, to take the first mover advantage in this ecosystem and then also leverage my finance, you know, knowledge or expertise to really educate the finance and tax professionals how we can use it, how we can leverage AI. I think that's super, super meaningful for me. Yeah. And believe it or not, I think finding the meaning in our lives is the most, most difficult thing. If we can find something you are passionate about and in the meantime is meaningful, that's incredibly hard. So yeah, so far I still very much enjoy it.
Georgie Healy: Well, I'm very glad it's you, Angela, and I'm very glad that you've built Luna and I love her name. So to finish the interview, Um, what I love to do is some rapid-fire questions. These are usually the spicier questions, and you kind of just say the first thing that comes to your mind. How does it sound? Are you up for the challenge?
Angela Shi: Yeah, sure, absolutely.
Georgie Healy: Should founders be introverts or extroverts?
Angela Shi: Depends. Both.
Georgie Healy: You can't say it depends.
Angela Shi: I think depends. I'm an introvert.
Georgie Healy: Okay. What is the best AI tool for professionals?
Angela Shi: Luna.
Georgie Healy: You can give a second answer, but obviously Luna is number one.
Angela Shi: So for the workforce, I guess there are so many fantastic tools for transforming the workforce. And I think especially in Sydney's AI ecosystem. Oh, by the way, so I just want to mention Build Club because Today Build Club announced their raise and I think Build Club have a lot of brilliant, brilliant AI engineers and building something incredibly useful for the professionals. So I recommend them, highly recommend.
Georgie Healy: Shout out to Annie who was on the previous episode. Well done. Yeah, she teased that before it launched, which was very exciting. Okay, I've got 5 more rapid-fire questions. The first AGI question I've ever asked, actually, do you think AI will surpass human intelligence?
Angela Shi: We're still far away from the AGI, I believe. And also, that's something, you know, we have the regulations and governance. And if something, you know, we fear is more risky than the benefit it can really bring to the society, society, then we can have the regulation to have some restrictions. Because I came from financial services, so it's more like risk-based, and we have a lot of regulations to kind of place the governance. So I think AI can do the same. So that's why I think humans still play a very important role. Yeah. And I believe we are a bit far away from the real AGI.
Georgie Healy: Yep. Do you ever use AI tools for fun?
Angela Shi: Of course. I use ChatGPT every single day. And yeah, so the reason is I try to see, you know, if ChatGPT can really mirror my tone of writing or speaking. And yeah, so I've done a lot of fun testing.
Georgie Healy: Yes. Now on LinkedIn, you're a top AI voice. Are there any other AI influencers that you recommend or that you follow?
Angela Shi: Well, yeah. So one of the top voice I always follow is called Alex Wong. I think she's based in Australia, but she's also a thought leader, global thought leader. She has almost 1 million followers on LinkedIn, and then she shares very, very valuable resources for AI learning stuff. Yeah. So, and she recently just joined Generative AI platform as a head of education.
Georgie Healy: Do you think there's any AI tools or platforms that are more hype than they are value?
Angela Shi: Never thought about that. Yeah, I never thought about that. No, I don't know about that. Okay.
Georgie Healy: And what part of your life would you outsource immediately to machines?
Angela Shi: Repetitive work. I hate repetitive work. Yeah, I think human beings have to focus on more strategic work, more fun and interesting, and then we shouldn't be the, you know, the cheap labor, especially when we first joined the workforce. We were always told to do the repetitive work, like in accounting, have a lot of journal entries, reconciliation, stuff like that. I think we can assign to AI to do it.
Georgie Healy: Yeah, that mind-numbing stuff. Yes, I agree. So, Anjali, you have been such a good sport, especially with my spicy questions at the end. I would love to give you the floor now to shout out to anything that you would like to share about Empathetic AI or Luna or anything upcoming that you would like the audience to be aware of.
Angela Shi: Oh yeah, sure. So actually next week I'm going to South by Southwest the entire week and We'll be joining, we'll be attending different sections and like the short talk in conference. And we are one of the pitch finalists in the Superpower Workforce Transformation category. And we also going to be joining Build Class Hackathon brainstorming section and also the Tech Village. So, and yeah, super exciting. for next week. And also in the meantime, we're in the process of raising capital as well. So since we launched Luna version 2 just last week, and we also officially kicked off our collaboration with CSIO's Data61. So yeah, so responsible AI team. So exciting news. So it's gonna be the second stage of updated AI in Luna. Since I started 18, 19 months ago. So it's a second stage and we are super excited to looking forward. And as I mentioned that we are in the process of raising the capital. So we have the expression of interest page being published already. So if you're interested to get involved, feel free to let me know. Amazing.
Georgie Healy: I'll make sure that we have the link to that, that raise in the show notes. Look, Angela, this has been such a pleasure. Thank you so much for joining in the Blink of AI. I really love talking to you about data desensitization, whether we need to be afraid of failure when starting a startup, even like work-life harmony. I love that concept so much. So thank you so much.
Angela Shi: Thank you for having me. Thank you.
Georgie Healy: Bye.
Angela Shi: Bye.
Georgie Healy: Thank you for listening to In the Blink of AI. You can check out the show notes for anything discussed in this week's episode, and we will be back next week. This podcast was produced by Day One with music by Dan Hansen and visual artwork by Sophie Tyrell. If you loved the episode, please tell your mates, and I love AI news. Please share your thoughts and suggestions to Georgina Rose Healy at Gmail.com.
