About this episode
“We always believe that there is an opportunity to make the world a better place by simply adding in AI to take out some of the menial tasks that folks do.”
— Mike Liu
Mike Liu, the founder and CEO of FreeFuse, started his entrepreneurial journey after earning a PhD in engineering from Texas A&M University, where he developed a passion for lifelong learning and 3D printing technology. His work in 3D printing was the catalyst behind FreeFuse, which he founded in February 2020, aiming to revolutionize the industry with his vision. As a leader, Mike is committed to fostering a supportive environment that encourages his team to excel and innovate, while he focuses on setting the strategic direction and objectives for the company. Known for his love of meeting new people and engaging in meaningful conversations, Mike is always open to connect and share insights about his journey and the future of FreeFuse.
Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.
In this episode of AIMinds, we had the pleasure of hosting Mike Liu, who shared his experience on harnessing AI to maximize teaching and learning experiences. Mike spoke about his journey into the tech world and the challenges he faced while building his AI-centered company, FreeFuse. His thoughts on building with AI and the value of proper positioning shed light on the intricate process of product development. Here are some extracts from this episode:
Mike's journey from engineer to AI startup founder is a testament to the power of innovation.
The birth of FreeFuse, leveraging AI for educational content, and the challenges faced in positioning and educating users make for an inspiring story.
Balancing novelty and familiarity in the user experience, a crucial factor for AI startups.
Gain valuable insights from Mike and learn about the dynamic intersection of AI, education, and entrepreneurship. If the future of AI and its impact on shaping new possibilities in the realm of education is your cup of tea, look no further.
Fun Fact: In the early testing phase of FreeFuse, Mike went to great lengths to gather feedback, including creating a low-fidelity version of the product and having friends participate in test sessions via Zoom. This highlights his commitment to gathering user input and iterating on the product.
Show Notes:
00:00 Deciding between current project or exploring new opportunities.
06:06 Deepgram improves accuracy and product quality.
09:16 Testing Zoom functionality, but data proved useless.
12:17 Challenging to build vision, resources, and testing.
14:33 Feed concept brings both novelty and stress.
18:51 Startups face challenges in finding AI resources.
20:20 ChatGPT popularized AI, dispelling misconceptions.
23:31 Wrap up
More Quotes from Mike:
“It's sometimes hard to have a vision of what you want to do, especially in the early days, because you're very excited about all the possibilities. At least we were. And we're now reaching some parts of that roadmap now, but it definitely took a lot longer than I thought, because you never really consider what you need to do in terms of resources, aligning the right people to actually build out what you're doing.”
— Mike Liu
“Now a lot of these people had educated themselves about what AI does. It's not just like this scary T2 Terminator stuff that people want to make it out to be. And then they really saw the value, which is kind of hilarious because it was literally the same things we said. It's literally one thing that's different, which was the introduction of a more widely used commercial service for consumers.”
— Mike Liu
“If you say that you're different from some other product that's in your market, or you call yourself a new category, if all of your features are the same as the other group, well, I mean, it's the whole adage of if it walks like a duck, talks like a duck, then it's a duck, right? So if you have all the same features and all the same offerings and you try to tell everyone you're something else, it's going to be a really hard sell.”
— Mike Liu
Transcript:
Demetrios:
Welcome to the AI Minds podcast, where we explore the companies of tomorrow built AI. First, I am your host, Demetrios, and this episode is brought to you by none other than Deepgram, the number one speech to text and text to speech API on the Internet, trusted by the world's top startups, conversational AI leaders, and enterprises like Spotify, Twilio, NASA, the one that puts rockets into space, and Citibank. Today we are joined by none other than Mike. What's going on, man? How you doing?
Mike Liu:
Hey, I'm doing great. Thank you, Demetrius, and I appreciate you guys inviting me for today's conversation. It's always a pleasure to be able to hop on podcast and discuss the future of what we see in terms of utilizing AI. And I think that in terms of philosophical approach.
Mike Liu:
I think we're one of those types of people who have looked at it as an augment to what people currently do, as an opportunity to actually help sort of build out proficiency and people feeling more comfortable about it, because I think a lot of folks have this inkling of, is it dangerous? Is it going to take away from what I am doing right now? And so we always believe that there is an opportunity to make the world a better place by simply adding in artificial intelligence to take out some of the menial tasks that folks do. And so that's a little bit into an insight of what we do here at freefuse, but happy to be here. And of course, answer all of your lovely questions.
Demetrios:
Excellent. So I do want to get into freefuse, the tool that you are creating. But before we jump into that, I would like to go down a trip of memory lane and hear about how you got into tech.
Mike Liu:
Sure. So I'm actually an engineer by trade.
Mike Liu:
The only problem with engineering that I saw, and maybe this is because I have some level of impatience with outcomes and things like that. But the only problem that I saw as something that I wanted to do was part of my PhD thesis. I was at Texas A. M. University. I was getting my phd, and I thought we had a really great idea for a new type of metal outlook right now. When I found out what we were doing would take, I don't know, like a decade to get pushed through, there was kind of that moment where you have to say to yourself, okay, where do I want to go with my life?
Mike Liu:
Do I want to pursue this? And this could be great, but the payoff is ten years, or do I want to look into some new opportunities that I feel like are easier to iterate on, I can get much more data from actually talking to people, seeing how their experiences work, and whether that's a better path for me. And so I had kind of those two options, and I think both had their positives and negatives. I was leaning more towards, of course, the actual project that we were working on at first. But the more I thought about it, the timeline of actually sticking with that and then not actually seeing if it could possibly get pushed through as a new type of material that people could use was pretty disheartening.
Mike Liu:
And for all that work, the downside risk seemed really great. And so I just felt like it made the most sense to actually utilize my experience in the teaching realm to be able to look at a problem that I felt like could be solved. And of course, like a lot of probably recent companies, a lot of things were born from the pandemic times, right? And they were born from a lot of folks recording their lectures at our school. And so I just got the idea that we should utilize the videos that we had, break them down into these decision tree modules, and utilize AI to do so.
Mike Liu:
Because the problem I kept on hearing was the professor doesn't have time. They might be on multiple boards, or they might be doing a bunch of research and writing grants and papers. The tas definitely don't have time. I come from that world. I know that they barely have time to do anything, let alone being a ta, let alone being like a research assistant.
Mike Liu:
So I knew that they didn't have time. And then, of course, the students just obviously had the negative parts of the experience, which were, okay, I have all this material, but nothing's useful for me to review. So we blended the idea of breaking everything down into shorter, bite sized pieces of content. So essentially like much more palatable micro content that you could kind of peruse through in your own personalized pathway. And so that's what the first version of the free fused AI did. It was kind of like the first little kernel of the offering that I thought we could bring forth, which is what we have today. But a lot of that was built off of language models.
Mike Liu:
And so I spent a lot of time actually trying to figure out, okay, so we know how to use AWS, maybe we'll just go and utilize that. It's the easiest thing, early stage startup. Try not to waste as much money as you possibly can, at least in a non practical way. So that was fine for probably the first, I would say, pilot programs, and going into, let's just say, the first times we were maybe selling free fuse as a platform subscription, but over time we realized that and the way our system works is that we utilize the actual audio transcript in order to make the micro content cuts.
Mike Liu:
So by having less accurate transcripts, we therefore have a less useful product. So I was actually introduced to Deepgram as part of a search for improving our accuracy and improving what we were doing model wise. And so a lot of that actually came into the forefront from a suggestion that this individual had. And so that's kind of how obviously you and I got connected, Demetrius. And from there, it's kind of leveraged all of that knowledge that I've had from being like an engineer all the way to now into what we've done product wise and how we think about building the product and how we kind of have a certain sort of idea towards quality in terms of what we do. So it's been very exciting. I feel like all of those elements have played a role. And so it's been great to be able to obviously be here and talk more about those stories.
Demetrios:
So talk to me about building the product. And of course, looking back on it, I imagine you can draw a straight line, but going through the trenches and when you were building the product, what were some of these iterations that you were playing with? How did it actually look in practice when you were going in the day to day and thinking about, should we make it a product that does x or should we make it a product that does y? People are asking for these features or those features. What kind of North Star did you have there?
Mike Liu:
Yeah, that's a great question. Well, I want to share the story of how I even started testing this thing to begin with. Right. I'm sure a lot of folks who listen to this have read blue ocean strategy or I forget if that's the original OG book or if that's the second book that they did. But the very first time we did it, I was like, okay, let's try and add as many differentiating points as I possibly can.
Mike Liu:
So the original concept was, okay, we're going to give people choices as interaction points, and we're going to add in voice recognition.
Mike Liu:
It was not very well thought out, like a half baked idea.
Mike Liu:
But I wanted to understand how people interacted with choices. I guess it wasn't just very good data to start with because we added the extra little wild card of voice recognition. But I have to tell this story because it's so funny. I actually took, I forget what these things are called. I have one here, but they're one of the Amazon assistants, but it has like a video screen on top of it. So we actually built a little low fidelity version that we hard coded. And I have some of the nicest friends in the world. They were willing to sit down for 1520 minutes through my crazy kind of plans of what this could be.
Mike Liu:
And they were nice enough to sit through, see what it was, and actually use it. But the funny thing is, I was in Texas at the time because I was obviously a teaching assistant. And then I also know these folks that I knew were primarily in southern California.
Mike Liu:
So, of course, the beauty of Zoom, and thankful the school gave me a free Zoom account because I really just maxed out that account as much as I possibly could. But I would actually go on Zoom and have people actually yell the direction through the interface just so I could see if it was maybe generating some kind of signal there. It's like the stupidest test probably ever imaginable. There's not even really that much relevant data because it's not even in the actual conditions in which it would normally be used. But it's just so funny because I feel like there was a part of me then that thought that, hey, this is a super valid test. We could see what kind of joy people have when they actually make these decisions and personalizing their experience. I found very quickly, probably, I don't know, a quarter or two later, that none of that data was useful at all. And so I think that what a lot of people underestimate is the quality of your testing criteria.
Mike Liu:
Like the actual thoughts around, how am I going to test if I'm right or wrong about this? And then more importantly, what are you testing? And are they actually high leverage things? Right, because you could test the color of a button on a random page. That's pretty low leverage. But if you're talking about, will the color of a button affect whether people purchase this through, like, a sales automation, then that's a lot different of a story.
Mike Liu:
So I think that there's equal parts like iteration, and there's equal parts like having that really solid product roadmap.
Mike Liu:
And I think in the early days, there was the idea that, okay, this is the first piece, which was the auto editor, and it has a video interface and you can have a library, but there needed to be more.
Mike Liu:
And so the idea was, okay, so you have interaction points. The next logical step is to be able to have interaction points as part of a p to p experience.
Mike Liu:
Okay, well, where does that all actually fit in. Well, you know what? It makes sense to build out a community, right? And so those all kind of came out of the initial ideas of, okay, what are we actually doing here with this experience? And based on what we're actually doing, what are the logical things that people would also do alongside it.
Mike Liu:
Because we want to at least make it so that there is some level of expectation that can be met, so that people can feel like, okay, this isn't like a completely off the wall experience that I've never had before, which had some level of shock value and benefit. But I think that there was, like, some quote from a guy. He was, like, an artist. It's the Maya principle, the most advanced yet acceptable.
Mike Liu:
And what we wanted to do was take in what would be the acceptable parts, which were the more familiar stuff, and then obviously innovate in ways where people could feel like it was a fresh experience.
Mike Liu:
And so that's what we felt like we were kind of building towards in those early days. But it's sometimes hard to have a vision of what you want to do, especially in the early days, because you're very excited about all the possibilities. At least we were. And we're now reaching some parts of that roadmap now, but it definitely took a lot longer than I thought, because you never really consider what you need to do in terms of resources, aligning the right people to actually build out what you're doing. And there's a lot of other stuff as well. So, that being said, I think those are some of the key elements to building properly, which is like having a vision that I think you can tailor over time. But equally as much, knowing that you're testing the right things so that if you get a signal for what you think is a good feature, it's not like pixie does.
Mike Liu:
It's an actual, real thing.
Demetrios:
I love this idea of the Maya principle. Most advanced yet acceptable, so it doesn't confuse users because it's so advanced. They're like, what do I even do with this? How do I use this? But it is advanced enough to where people think, oh, wow, this is delightful. And you're almost going for that delightful aspect, but not the surprising or overwhelming aspect of it. And that feels like a really good north star to shoot for.
Mike Liu:
And to add to that just really quickly, I think that what it really does is it actually activates a lot of people. Talk about the three different parts of your mind.
Mike Liu:
Like the paleocortex, the limbic system, and the neocortex right? And of course, the stuff that's the oldest part of your brain, I think it's like the cerebellum and the brain stem and stuff like that. They kind of call that, and I think in some circles, the lizard brain or the reptile brain, right. And the whole idea behind it is if you have enough novelty, it gets that person's paleocortex actually activated. And because of that, that's why a lot of people really like feeds or like lots of novelty when they come into a user experience.
Mike Liu:
That's why the feed concept, when it was first introduced, was great, because you could literally just sit there and refresh as many things as you want to, right? But there's a give and take, right? You can activate novelty, which that part of your brain really likes contrast, or you can activate fight or flight, which would be bad, right. It's too different. Automatically shut down. I'm not going to look at this. And those are just things that happen. And it's kind of part of, to me, one of the things that makes it so interesting to design for people, right. Because if you think about it, everybody kind of comes in with a lot of those same things that they might deal with, or we all have the same paleocortex response, right. It's the other stuff that actually kind of sort of changes our experience per se.
Mike Liu:
But it's so interesting to have to design for people for that reason, because now you have to take into account all of these psychological and neuroscience elements to be able to actually give people a pleasant, relevant, and most importantly, useful experience.
Demetrios:
So, continuing on this theme of how you dealt with adversity and how you ultimately created the product that you wanted to create. I know that everything wasn't roses from the get go, and I would love to hear about what some of the challenges you found were, as you were a building with AI, and maybe even more specifically, like any other challenges in creating the company. And anything that comes to mind, really, as you look at the last x amount of months of trying to get everything working properly and create that well oiled machine, maybe it's organizational challenges, whatever is right there and comes to mind.
Mike Liu:
I'll mention three things. So I'll go with. First, like building with AI, right? So when I really started to look at the concept of AI, I think it was like the summer of 2017, I was still at a and m looking to finish out my master's degree, actually, at the time. And I started looking at what was really possible in this field, because I had the idea that I wanted to build a company, but I wanted to make sure that I was utilizing technologies that I feel like could make people's lives easier. And as an overarching concept, obviously AI is within sort of that category of things.
Mike Liu:
It can obviously do. I first started learning about robotic process automation, RPA, right. And to some degree, that's what sort of the auto editor does for us, right. It's basically an RPA. It'll do this task, you don't have to do it anymore, and you're able to get a particular result.
Mike Liu:
Normally a person would do that. It sucks. They don't want to do it, and so therefore we've saved them x amount of time.
Mike Liu:
The interesting thing about doing that as well was I was able to also, while I was at school, access tons of science papers. So I was able to read tons of what folks were doing in terms of the field. And while I didn't obviously understand some of the computer science aspects of it, I could understand sort of the thought process behind what people were doing with this and allowing me to know, okay, here are some of the great minds in the field, and what is it that they're really talking about? What are the things that are actually keeping this science and this field from growing at that exact moment?
Mike Liu:
And I think that idea comes from being a PhD myself and going through and figuring out, okay, this is actually the thought process that goes into how you're going to approach some of these problems.
Mike Liu:
And I think that that's such a useful framework to have in terms of understanding the world around you and more importantly, questioning what's possible.
Mike Liu:
So challenge wise, actually, I would say this, and this is the second part of this, which is I think that it's also sometimes difficult to find the right resources as an early stage startup for AI focused work if you don't already have that AI team member with you or like they were a part of the early stages of the company, I think that that can be a challenge for a lot of folks. And then you have to think about, okay, well, do we send this over to a shop that's actually going to do it? Do we build a relationship with a vendor partner? So there's all of those considerations, and if you don't have a highly technical team or at least somebody who might be able to even understand or manage working with some of those folks, a lot of those projects can go sour. So I'm very grateful that we have a great engineering head who's worked with lots of different types of vendor partners, including offshore. And so we've developed sort of that 24 hours dev cycle, right? So I think specifically the interesting part about, I would say the last 18 or so or even 24 months has been that people didn't really understand. I think what we were actually offering AI wise, right? Like we were telling them, this is what it does, this is what our AI does. And for whatever reason, I feel like once they started using, they understood it. But when you tell them the first time, it was almost like, well, I don't really care. I don't really care about this AI thing.
Mike Liu:
When chat GBT started to get popular and we started to share that we have AI offerings and what we do, the number of people who now wanted to know or go back into conversations was amazing, right? Because now a lot of these people had educated themselves about what AI does. It's not just like this scary t two terminator stuff that people want to make it out to be. I even think there's some movies that were released about ais, not like evil eye or anything, but other stuff, right? So to me, I feel like it was that sort of flipping of the switch that got people more educated about what we were really doing. And then they really saw the value, which is kind of hilarious because it was literally the same things we said. It's literally one thing that's different, which was the introduction of a more widely used commercial service for consumers.
Mike Liu:
So that was, to me, a very interesting part of even pitching the value of that part of the platform. But I think overall, and a lot of people, I think maybe discount or underestimate this aspect of it, which is how you position yourself, right? And so by having a certain suite of features and user experience elements, we can position ourselves as something completely different than what we were originally getting positioned against, which was LMS. And we really didn't want to necessarily compete against LMS because they just do a lot of stuff that we really don't care to do. Now, can you use free fuse as a turnkey? Lms? Absolutely not. Everybody needs all of those features that an LMS provides, right? I mean, I used to use lms all the time. They're bloated with tons of stuff people never use.
Mike Liu:
So on our end, we really wanted to focus more on the sort of community and unique p to p experiences rather than, yeah, hey, you can have a gradebook, right? That didn't make sense to us. So looking at how you position yourself, also tying back to the north star of how you develop, knowing how you position yourself and knowing where you want to position yourself maybe in the next two, three, five years, allows you to then also have the right north star in order to make sure that you develop the right stuff.
Mike Liu:
Because if you say that you're different from some other product that's in your market, or you call yourself a new category, if all of your features are the same as the other group, well, I mean, it's the whole adage of if it walks like a duck, talks like a duck, then it's a duck, right? So if you have all the same features and all the same offerings and you try to tell everyone you're something else, it's going to be a really hard sell. So I would say that those things are somewhat underestimated. I'm sure everybody has things like hiring challenges, scaling challenges and things like that. So I figured that this would be actually a useful thing, perhaps for folks maybe starting out.
Demetrios:
Well, Mike, I appreciate this a ton and I'm very excited that you are part of the Deepgram startup program and we get to join on the adventure with you. Thanks for coming on here and chatting with me and I look forward to hearing updates on how everything goes with freefuse.
Mike Liu:
We appreciate you and happy to be here and hopefully I'll be able to come on again and talk sooner.
