OK, let’s get started.
Sure. Why don’t I give you my card?
I have to give you my new card.
You’ve got a new card?
(pause)
Cool and I’ve got mine, which is from last year. [laughs]
Great, thank you. I’ve heard a lot about you, [laughs] even checked you out on YouTube. I liked the conversation that you had with Professor Yuval Noah Harari…
The historian.
…with the RadicalxChange.
That’s right.
That was pretty cool. Anyway, let me give you a little bit of background about myself and the company that I represent. Beyond Limits is a AI company, based its headquarters in Highclere, England. The original IP and the technology platform is from NASA, the JPL Caltech.
The team, mainly we’re from the JPL spinoff. What we did was the team was the same team that developed the AI platform that drive rover to Mars, the Curiosity Rover 2012, that was the AI prep was mainly developed by this team. AJ, our CEO used to be the head of commercialization of Caltech. He has done quite a lot of commercialization project for Caltech. It’s been out there, technology platform.
AJ persuaded the Caltech there’s a commercial usage for this particular platform that have already been deployed in rover and also using some other military applications as well. We know, cut beyond him, is purely a commercial company, basically get a license, exclusive license from Caltech for about 70 IP brands since 2014. From that point on, we develop another 100 something to IP brands.
Right now, we own about 210 IP brands. The first industry that we got into is energy. The BP was one of our first technology adopter and also later on became our investor on Series B. My involvement, I run an investment fund company for capital headquarters in Seattle. I came into in series C.
We just closed the series C last year for about 135 million. After we closed that, our plan is essentially to globalize the platform. Now, we have Bendel, which is our main R&D center and also providing services in development products in the US.
We have a UAE team that’s serving the Middle East and EMEA market. I’m the one who’s leading the APAC expansion. Basically, right now what we have done is we set up operations in Singapore. We have a team in Singapore. Chungyi is our general manager for Taiwan. We are building a team over here.
We have a small team in Hong Kong. In Hong Kong, we are focusing more on fintech. Primarily, the team in Hong Kong, most of them have financial technology background. They are using our platform to find partners in that area, focusing on, for example, AML solution, small medium size loan processing, essentially using our AI platform to drive that. We also have a small team in Japan.
Within a relatively short period of time, we’ve done a fair amount of work trying to set up the infrastructure in Asia. I personally spent a fair amount of time in Taiwan. That’s when I learnt most of my Chinese. I was a general council for a Taiwan comapny for about nine years. Then I moved back to the States and started my investment business about 10 years ago.
Basically, in the past 10 years, I’ve been trying to identify unique technology platform. Hopefully, either me as a venture investor, and then later on try to branch out to other parts of the world, primarily in Asia.
That’s the general background on Beyond Limits. We look at Taiwan from multiple angle. One is definitely the certain products that we already developed, especially on the energy sector. Chungyi has been very helpful connecting us with a lot of the energy players over here. We just came from [Mandarin] . We just arrived for about an hour ago. That’s just been pure business development.
The other part of it, we also want to look for strategic partner in Taiwan. When I was in Taiwan, I was pretty tied into the hardware design and manufacturing ecosystem. Then from our viewpoint, we would never be a chip company, for example. We would never be a device company. However, our AI algorithms, especially with the Nasa Rover AI Program, there’s a lot of edge AI technology in rover.
That’s right. Sensor fusion.
Exactly, sensor fusion. We actually have IP brand focusing on that. One of the things that I’ve been talking to AJ is that we are not going to be developing those devices. I think Taiwan is the best place to find partners, both on chip development as well as any kind of more intelligent device development. It could be applied in healthcare. It could be applied in any IoT infrastructure, and things like that.
I’ve played with the predictive model you have on the website.
OK, you did?
Yeah, I did.
Did you do the COVID one?
Yeah, I did. I did this one.
That one, we did it with Cleveland Clinic. We did that in about six weeks here.
Oh, really? Wow.
We built that thing in six week. We gave it up for free to the States.
It’s a pretty basic SIR model.
That’s right, yeah.
It looks pretty well researched.
Cleveland Clinic, they are the one who provided medical donation, whatnot, and all that. Then, we build the model. We also have other healthcare projects going on right now. Some of them relate to integrating our AI algorithms on sensor collecting biometrics information to do some simple diagnosis at the edge.
Body area network?
That’s right. A lot of them has to do with COVID management. Especially in some other areas, even in developing countries, it’s about allocating hospital resources. To the extent that you can understand the current, the physical conditions of the, not exactly patients, just the general populace.
Then, you can manage how to direct the right populace to the hospital, and say how to match together. “You’re OK, you can probably stay at home and and take care of herself or not.” We are developing solutions like that, with our healthcare partners. Definitely, if there are solutions that we think we can bring in to Taiwan, we would love to.
We also, very much, like I said, the second purpose for us to be in Taiwan is to innovate with local Taiwan partners. Both on hardware design solution, developing different solution, different verticals. Then, the third purpose is to build talent pool. We will be hiring folks, AI engineers, data scientists over here.
In Kaohsiung?
No, overseas in Taipei. This time our client is in Kaohsiung. Our product focus is focusing other production management for heavy industry. That’s why we were in Kaohsiung today.
OK, more like this is in the customer’s side.
For client, basically.
In this day and age, you can be anywhere…
Sure, of course.
[laughs] …to be honest. I heard a lot of great things. There are things that are happening in the southern part of Taiwan, Kaohsiung, and Tainan.
The Mayor of Kaohsiung is quite enthusiastic about this.
Sure. We will be happy to even look at talents over there as well. For us, it’s not particular geographical. Taiwan, you can have Kaohsiung and…
It’s a larger municipality. [laughs]
Exactly. It’s very doable. We are pretty open. Also, we are not talking about building an army of AI engineers. Even for our company right now, it’s 150 people and then we get a lot of things done. Right now, in the entire Asia operations, about 15 people and this year, 2021, we are expecting to expand to about 35. Then, to 20 people, almost most of them are the technical.
We do plan to put a fair amount of them, the technical resources, in Taiwan because we can support the Japan market as well. We already have a couple of pretty big projects in Japan that we would need the technical resources to support. Very quickly, a very brief background about our purpose, why we’re here? Chungyi, you thought that it would be [laughs] good for me too.
Yeah, well we’ve talked. [laughs]
From my perspective, we do need to ask the government support. To build up any of that in Taiwan from scratch, we want to look for governments, even the support or a co development partnership or anything. This is the first thing. The second thing, it’s a win win situation for both Taiwan and our company. On one hand, we had about 210, the license…
IP process.
IP process.
IP process, yeah.
Then, we combine all these technology into a cognitive AI. The good point is, we are just like a hub. We can embrace all different kinds of the AI talents, no matter it’s numeric or symbolic. Either is good. They can see the holistic picture and being able to work at global project. Through the process, of course, we can imagine the AI industry in Taiwan will have a boom.
On the other hand, we do want to look for a government who’ll support, whatever kind of support. This is our purpose to visit you.
Support is one. The other one is I, personally, have been impressed by what, Audrey, you have done, and also Taiwan has done this past year in this challenging environment, and then continue to maintain the economy. Big part of it is the use of technology, and use of technology in the right way. I had that conversation with AJ and also Mark, our CTO, all the time.
A lot of times, you need to find a right environment. Last year was not exactly the right environment, even in my own country [laughs] to do this kind of experiment. Taiwan is a good place for us to test out some new models, how to adopt our AI, especially our cognitive AI, which comes with a specific feature, which is explainability.
That’s something that a lot of AI, you don’t have that. That part, it is very core in our philosophy, in developing our AI platform. We think that if AI is going to be adopted in a society for this purpose, you has to develop that trusting relationship between a machine and a human. In order to do that, you have to have explainability.
That goes with the transparency idea. What you have done in Taiwan, a lot of time using technology to make it more transparent, make technology even more accountable, so to speak, to people. That’s something that I’m interested in exploring, what kind of projects that we can do together with the government. Definitely then, to that extent, it will be natural.
There might be some mutual support, in a sense, to try to do something together.
One of your main points about explainable AI is, and I quote, “evidence-based queries in natural language.” That means, it’s not only comprehensibly natural in language, but also, people can query it using that in their phrases?
I’m not exactly the best person to talk about that aspect.
Sure, of course, but what’s the experience for nontechnical people with this system?
Basically, what we did was, we have a cognitive AI agent. That is the core. Essentially, we can take a lot of inputs from the numeric AI, so whether it’s coming from CNN, our end, different sets of data and all that. We take those input. We also have IP that can help us qualify heuristics, the industry practice, certain kind of past experience.
There’s a lot of capturing of the institutional knowledge. Say, for example, this past two weeks or so, we have been talking to a lot of different industries and their problems. A lot of the problem has to do with 傳承, their experience. They’re not passed on.
Their tacit knowledge.
It’s a tacit knowledge, it’s all lost. Then you have the young folks. Maybe sometime, especially in the industry, energy sector is not exactly the most sexy industry. It’s hard to attract really good talent. Then, how are you going to be able to train those folks and bring them up to speed as fast as you can? AI can play a role in that.
You can also have people who are less skilled to be able to do more skillful work. To some extent, there’s a little bit of that social impact on that as well. That’s part of what our AI engine can do. We could go deeper into it. I could probably even bring Mark to elaborate on the platform.
For example, we have these hypothetical generators, the fact that we don’t need a complete data set in order to have the cognitive AI agent to give you a recommendation.
Part of the reason is because we can take a lot of rules and heuristic from the symbolic AI to generate different hypotheticals, to fill up that gap, so that you can have a complete model, so that the machine can give you a certain recommendations. While he is doing it, the machine can also explain the audit trail. They give you the audit trail for that particular recommendation.
We have demonstrated we have a product that’s quite well built. It’s called Process Automization Advisor. We took on a lot of the operator’s knowledge in how to run a prod. Most of the time, we’re focusing on where the human decision making process is key. Again, like I said, it’s almost like having a super operator next to you.
That’s right. It’s like a interactive decision system that you can go back and change the hypotheses. You can ask a Shifu, or a skilled person, what if this situation’s not like this, but rather that? They will also apply their expertise.
Absolutely, yeah, you got it. Then, you click on it. The thing is, what we do is that we don’t override a human decision. We only give you advice. For the operator, he can click on it and understand what the logic. Let’s say there’s a tripping event because you’re deviating from the plan. The objective could be economic, could be safety, could be a combination of different objectives.
The operator will click on it, then you will see that the machine have gone through a couple recommendations. Each of them come with a certain logic and, finally, recommendation number one. You can override it. You can come in and say, “You know what? I look at all this, I probably prefer recommendation number three.”
There’s a little bit interaction there. Maybe it gives you some insight, you come up with different way of doing things. That feed back into the loop, so you continue to be self learning that way.
OK, that makes total sense. Thank you.
If we think about why Taiwan, and how we can help Taiwan, the strategy is very clear. Taiwan are seeing a world of treasure. We almost have all kinds of heavy industry, and all kinds of traditional maker industry.
When we look through what happening in China, for example, a fabric industry, or manufacturing, east China or southeast Asia, or whatever, you can certainly find certain source from Taiwan since 1980s. Luckily, we still keep the roots. The strategy in Taiwan is through the co development process. We keep those treasure in our system as a constraint.
If possible, we want to co sell with either local partnership to oversea. All kinds of partners we have in Taiwan, it’s more like our product development partner. It’s not purely a client or a consumer, they are partner. We utilize and keep the treasure from Taiwan and then sell to oversea. That is our strategy.
If we want to go through this process, each single domain knowledge, we need a local expert to translate those domain knowledge into comprehensive language our computer science can understand, and to co develop the product.
Originally, that is how we expect technology to help the society or to help the world. Unfortunately, in the majority of the case, this kind of synergy working pattern, never come true because people from different domains, they don’t speak the same language. It’s very hard for them to collaborate.
Luckily, we made a successful case in US. We know how to make it happen. Given the fact, like Taiwan, we have the treasure of all different kinds of maker, and we want to make it happen in Taiwan.
One of the things I have been thinking about, we visited Microsoft and talked with them on we can cooperate to sell solutions over here. When I say sell, frankly, we have business considerations obviously. The big guys, we definitely have to talk to them, try and sell our solutions, help optimize the processes.
But there’s all these small medium size enterprises. They are doing day in day out manufacturing, design, and whatnot and all that. They can also benefit from the AI solutions. How do we do that? Just brainstorming this whole trip. I just got out of US, and then just talking to folks and trying to brainstorm.
One of it is right now, if you sell a complete solution to some of these folks, you might be just too overwhelming for them anyway. They’ve been running a mom and pop shop, making these gadgets and whatnot and all that. I’ve been doing it for generations.
However, I am also facing this potential threat that I will not be able to compete with the big guys, but I know something. I know something specific in this industry.
Domain knowledge.
The domain knowledge. It’s very, very crucial for that domain knowledge. If I can create a knowledge editor, I can extract that domain knowledge and build it into some kind of domain specific language. Then have a knowledge editor that the person can input their own [Mandarin]. That can be expanded. Then you are selling, not only just your hard labor and your…but how you are generating IP, for yourself.
I think that, for one thing, it will help even the [Mandarin], even the enhancement. At the same time, it also may cultivate a pretty vibrant, flourishing environment. That requires some infrastructure building because different verticals, you need to develop different kind of knowledge editor.
That’s something that I’ve been talking to Mark about, and it’s called knowledge editing. You convert the domain knowledge into domain specific language that the economic agent can understand.
Ultimately, you cannot have everything done by grandeur or beyond limits. You want to democratize it a little bit so that the client, at their level, can actually continue to use this tool. To continue to build up their own knowledge building and capture.
That’s the thinking about it, putting some time in. That requires some infrastructure. We’re still a startup, even though we raised a fair amount of money, but at the same time, we survive. We need to make sure that we get the bread and butter.
Instead of always looking for this seven figures project that we want to get, this is more scalable. This is something that you can sell to more popular…the public to use.
OK. Just to check my understanding, the current operational efficiency schemes deployed in existing customers, deploying in beyond limit fields, is all centered around energy or is it also other centers?
Mostly centered. We are mostly energy, and we are moving into healthcare, as I mentioned. We are going to move into different verticals. That’s part of the reason that we do the CVC. So, the CVC is about globalization and also moving to different vertical.
Like in UAE, we are pretty close with UAE. Part of the reason is because one of the investors, G42, is a kind of UAE investment fund that support…I’m sure that, Audrey, you will probably understand a little bit about the background. Through the Abu Dhabi investment fund, we are doing a fair amount of digital transformation in different vertical sector.
Healthcare is definitely one of them, power and utilities is another one, and I mentioned financial, fintech.
Yeah. G42 is also into fintech.
Yeah. They are very, very much into fintech. In Taiwan, you definitely want to find some industrial 4.0 partners and as I mentioned, in industrial 4.0, the solution can be benefiting not only just the big guys, but also the small media design enterprises as well. It is our goal to actually branch out to other verticals.
We are probably not to the point of getting into the 2C market, like advertising and whatnot. We are not really touching that part yet. We are still focusing more on the industrial enterprise side.
Right. No chat bots just now. [laughs]
We actually were talking to [Mandarin] because they have a chatbot that can process multi language like Cantonese, Mandarin and whatnot. We were thinking about using it for the banking operations like customer relations and customer services and all that. There was a little bit discussion about how we can actually improve the chatbot.
I see. You are more like the Wolfram Alpha to their Alexa.
That’s right. We were having a little bit discussion with them. We were having a discussion with ASTRI in Hong Kong and also A STAR, in Singapore. A STAR in Singapore, they want to play Malay and the other language and other dialect in there as well.
OK. I think you are off to a great start because operation technology efficiency in the energy sector, it’s very evidence driven and there is a lot of [Mandarin] tested knowledge that is at the constant risk of being lost. It’s a great sector to begin with.
I don’t quite understand the SME, the small medium enterprise angle because there is less of the tacit knowledge or experience to be captured but more of a day to day interaction patterns. It’s more on the 2C side rather than internally under operation technology side.
There might be some part of your technology that I have yet to grasp down to what enabled small media enterprises customers, but on the operation sectors including energy but also, as I mentioned healthcare. This is a pretty good fit.
Yeah, I know. The SME is not easy. Trying to find angle is not that easy.
Yeah, because a lot of SMEs here, they already have this inclination to basically self host their software even to get the cloud deployment going. It’s not that easy and mostly out of habit rather than any concerns about server security or things like that, but some of them, it’s more about the psychological safety thing.
If I’m a SME owner and if I go to your website, which says, in no uncertain terms, on the bottom of the website that, we’re only providing “temporary use of online non downloadable software,” [laughs] that’s actually a warning sign for many SMEs around here.
Yeah, I know. There has to be some on prem title solutions which…
I understand for a sense of fusion and things like that, of course, that’s possible and maybe needed for edge computation, but, for the SMEs to deploy this whole sense of array maybe outside of fair budget.
Yeah. You are very, very to the point. That’s great insight. Absolutely.
I’m not in charge of the Taiwan Steel or Oil [laughs] or Water for that matter.
For example, if we want to, because we know the Taiwanese government support the Microsoft and the Amazon to build up presence in Taiwan.
Municipal governments, yes. They’ve helped to set up the incubation centers and training centers and so on, that’s right. In particular, the Startup Terrace in New Taipei City partnered with them and also Shalun too.
We like to know two things. Firstly, from your high level perspective, how will you advise us to work together with Taiwanese government to empower the society and to utilize our platform? That is the first thing.
The second thing is we also like to know, given the fact you are in the central government, based on your understanding, where may we find a resource to support us, realize to build up our own AI team and realize the local R&D in Taiwan?
The reason why I emphasize the municipal governments is that most of the talent base, be it training or recruitment or things like that, they do offer subsidies and so on on the municipal level.
On the central government, what we mostly do is to make sure that there is a favorable, for example, immigration law. For example, all your teammates who are not residents of Taiwan can nevertheless apply for a Gold Card. Anyone who have the potential to contribute to science and technology are now eligible for a Taiwanese Gold Card.
Long as they are willing to do this 14 day quarantine, which we just went through, they can apply for a Gold Card outside of Taiwan. It just takes a month or so. Then they’re free to travel here and even enjoy healthcare and bring their families.
That’s what we in the central government do because we understand for many people to get together with their customer is important. Later on they can do — I don’t know — mix reality or something.
Having this initial face to face understanding is important, so we make the Gold Card, the immigration law, the foreign Taiwan app and things like that and make it very favorable even during the pandemic for the people around the world to contribute to this kind of talent development.
Like the AWS or Amazon Web Services and Microsoft, it’s a cognitive service or something, they make specific deals with municipal governments who are looking to make, for example, the test fields, the sandboxes, and so on.
They can relax their local rules about, for example, self-driving vehicles on 5G spectrum, millimeter wave and things like that, so they can test out their AI models in real life, so to speak, but in a risk contained way.
The explainable part is the most important part in this because during the sandbox application’s trial run, of course, accidents will happen. If people can account for all the accidents and simulate what’s the best ultimate path back in time, that makes it much more easy for the society to co domesticate with new AI based solutions. Most of that happens on municipal level.
Even the Sandbox and Prime was done at a municipal, even on the recreating level.
That’s right. The regulatory office has to do it, the Department of Industrial Technologies of the Ministry of Economic Affairs. All the self-driving vehicle applications, it’s reviewed technically by the duet in the MOEA. The sandbox field needs the approval by the mayor of whichever city or municipality that this test field is going to do.
It has to clear both. In reality, the most day to day collaboration, and incubation, and de facto subsidy during sandbox application, they’re executed, implemented by the municipality. Of course, sometimes the municipality asks for the central government to also help in getting, for example, science and technology budget.
That’s channeled into these kind of experiments. Still, the implementation is done by the municipality.
I see. Is there any ranking of which municipality is actually more receptive to innovative projects to be tested out and all that?
The six municipalities, they are all very much into this sort of thing. Also, like we talked about explainable AI, they are also in charge of explaining the cybersecurity implications to the counties and cities near their municipality or require a regional alliance [Mandarin] .
The head of the six municipalities or the head of the…is variously called the Bureau of Information, Department of Information — sometimes it’s called Digital Bureau now in those municipalities — are the main go to people.
Sometimes that Taipei City also have a small city office. It’s a project management office that is specifically designed to work with private sector.
Any preference in terms of whether there’s any particular preference for working with domestic companies rather than international companies like…?
Well, I don’t think there is a clear heuristic whether to work with domestics startups or oversea startups. It all depends on how viable or scalable the solution is. If it’s catering to one particular county and has very little likelihood to scale out, then if I am a mayor, I would prefer to work with local startups and retain the talents here.
On the other hand, this is just like a pilot run. The model depths generates has a high possibility to extend to the world, then some sort of connection to international community will have belonged.
(pause)
It’s for discussion. Given the fact we are trying to reach as many heavy industry as possible and co develop our product with those heavy industry, those heavy industry, for certain reason, mostly they are in Kaohsiung.
Kaohsiung is a great starting point.
Yes, it is. In this case, we do want to work together with the Kaohsiung government, but we’d like to know if, because we don’t have any contact point in the Kaohsiung government, could you help us for the referral?
Kaohsiung has its own smart city management office. What is public information… Zach probably has the contact to the PMO in Kaohsiung, but it’s public information.
OK. Thank you.
You’re going to do some research, who you should go and talk to in Kaohsiung.
If you already have some…I read about, for example, pinpointing drilling opportunities, inspecting pipes with self navigating robots, and things like that. I think the major energy companies, state owned or otherwise, in Taiwan already are quite versed in these scenarios.
Your valuable proposition would be that, your technology make it more explainable or easier to tune or easier to launch. You will not be talking to people who have no idea what you’re talking about. They all have their own AI things working on precisely those technologies.
Thank you so much.
All right.
All right. That has been a very good exchange. A lot of good insight.
Thank you for the exchange.
Thank you.
Also, from the central government level, we would also like to know what kind of project do you think we can work together?
Around end of this month or early March, we will start the call for solutions for the Presidential Hackathon. It may make sense for you to take a look at the previous Presidential Hackathon…
Presidential Hackathon.
…and see what cross sectoral teams that maybe you’ll be interested…
I was at the AmCham luncheon. You mentioned about a water leakage…
That’s right. That’s one of the inaugural Presidential Hackathon cases.
Right. We have a sense of appraisement solution to identify water leakage.
That’s right. Last year, one of the top five teams used smart meters and did Taiwan energy company, the Tai Power, that use AI to find out whether it’s the fridge, or whether it’s the air conditioner that’s consuming a lot of power at peak hours. A lot of those AI insight tools is already being deployed.
If you can narrow down your value proposition to either the explainability angle, which I like, or to the faster knowledge transfer interactive knowledge bank, or workplace, or workbench metaphor.
That’s another angle, but I think it’s easier if you focus because from what I can glance from your website, you have a lot of technologies. [laughs] It sometimes gets dizzy just looking at the breadth of technology.
I’ll take that feedback back to the CMO. [laughs] Sometimes, I have the same issue, which is a lot of information on our website.
That’s right. It’s just so much. [laughs]
Indeed. Great.
All right. Thank you.
Thank you very much for your time.
Thank you for your time.
(background sounds only)