How Does Beyond Limits, Creator of the Cognitive AI System that Helped a Mars Rover, See the AI Development Trend?
ORIGINALLY POSTED: 9 FEBRUARY 2021
CHEN CIYAN FOR DIGITIMES
Editorial Note: This article has been translated from Traditional Chinese
AI has gradually become an emerging technology gaining popularity around the world, allowing people to realize the impact of AI solutions on their lives, whether or not those solutions are visible or readily apparent in their day-to-day lives. This popularity has also driven many new creative teams in the market to invest in this field more frequently. Chen Chenfei, an analyst at DIGITIMES Research, pointed out that as AI technology becomes more mature, the focus around AI industry development is gradually shifting to lowering the threshold for enterprise development of AI applications. Due to the insufficient degree of solution customization launched by the majority of AI companies, some innovative companies with cross-domain AI solutions have emerged to close the gaps.
Startup Beyond Limits was established in the United States in 2014 with its core AI technology developed at Caltech’s Jet Propulsion Laboratory (JPL) for NASA space missions. Beyond Limits provides enterprise-level AI solutions. Since its establishment, it has achieved success in several industries including energy, power and natural resources, manufacturing, healthcare, and more. The market covers the Americas, Asia, the Middle East, etc., and will officially enter the Taiwan market in 2020. DIGITIMES interviewed David Liu, Executive Chairman of Beyond Limits Asia Pacific, inviting him to explain the establishment process, current outlooks and future layout of Beyond Limits, as well as expectations for the Taiwan market and observations of the AI industry as a whole.
Could you please explain the establishment process of Beyond Limits?
Beyond Limits is an AI company established in the United States in 2014. Its technology is derived from NASA and its affiliated Jet Propulsion Laboratory (JPL) out of the California Institute of Technology (Caltech). The AI system that helped NASA’s Mars Curiosity Rover on its journey in 2012 was built by the Beyond Limits team. AJ Abdallat, CEO of Beyond Limits – who previously served as the Head of Commercialization Strategy for JPL and Caltech – encouraged them to export IP for commercialization in 2014.
In the beginning, NASA built a toolbox for the Mars probe. Since there is a time gap difference of 15 minutes between Mars and the Earth the probe couldn’t be operated in real-time. Under these circumstances, the transfer of data could not be carried out using only numeric AI techniques. Therefore, it was necessary to design a new solution that included an element of human-like thinking (symbolic artificial intelligence) – this is how Cognitive AI came to be.
We chose the energy industry as the first sector with which to get involved. At the time, bp, a large multinational oil group, sought to reorganize their internal operations and required a company that could help them optimize the decision-making process. After several twists and turns, I was introduced to Beyond Limits. In the past, Beyond Limits had experience with NASA probes and solved many communication and system problems in the process. After some discussion, bp believed that Beyond Limits had the right technology to help solve some of the challenges behind their decision-making processes. Subsequent cooperation also led bp to determine that our technology was a good source of support; the company then became a two-time investor in both Beyond Limits’ Series-B and Series-C funding rounds.
As for the establishment of our business model, we will develop several new products in numerous industries and fields at the outset, eventually becoming SaaS modules. That is to say, we are an industrial, enterprise-grade AI software solution provider helping companies solve decision-making process challenges.
In 2014, we discussed with bp exactly what challenges they needed solving, including oil exploration optimization, a very complicated matter. Solving the problem required the creation of an advanced AI solution. At the outset, we first looked at the business challenges bp was facing, provided a customized AI solution to help solve those problems, and finally productized the solution to make it a SaaS service.
With bp’s backing, we began by optimizing their work in upstream from oil exploration to maintaining and managing oil wells and assisting in the optimization of maintenance processes. When it comes to the refinery stage, because processes are very complicated, and in addition to optimizing those processes, we also provided solutions to support chemists in developing new formulation blends. From the refining process down to development work and final logistics, Beyond Limits provided one-stop AI solution services.
Can you briefly describe the characteristics of Beyond Limits’ AI technology?
Beyond Limits’ technology combines numeric symbolic AI, using two AI logics to process data and establish control variables, thereby creating Cognitive AI that provides explainable recommendations to human operators. This technology is also able to continuously learn and optimize using the feedback of personnel, ultimately reaching complementary/collaborative intelligence. One of the characteristics of our technical module is that it is interpretable. When our AI provides you with a suggestion, it will also provide the reasoning behind its recommendation. This is different from the “black box” AI provided by other industry players in the market. When AI cannot provide reasons behind their recommendations, it is difficult for humans to have a sense of trust in the solution.
AI is not a panacea, which is why interpretable solutions are so important. The existence of this feature can be regarded as a necessary element for establishing human trust in the system because people can analyze whether the provided advice is correct, as well as understand its logical inference. When an AI solution is transparent, it results in more certainty in the decision-making stage.
The biggest difference between the interpretable model and the black box is that the black box generally solely uses numeric AI techniques, which require a lot of data as a calculation base. In addition to data, interpretable AI also adds industry knowledge and human expertise. In addition, it can capture human actions against missing data and human thinking, making inferences based on past experience when there is missing or incomplete data, using accumulated knowledge bases to arrive at recommendations.
Presently, there are many new ventures in the market investing in the AI industry, but most of them propose solutions for specific processes. The technology Beyond Limits offers is unique in that we are a full-stack-oriented, one-stop solution provider. We are not a startup that thinks of relationships with other startups as competitive. Rather, we look forward to future cooperation with other specialty players in the space. For example, our AI could be used in coordination with OCR technology, chip makers, data platforms, and more to make on another’s technology that much stronger.
Do you have any layout plans for the edge computing market?
Beyond Limits has previous experience in edge computing. In the future, it will actively seek opportunities to cooperate with hardware manufacturers from all walks of life, potentially in the areas of 5G. Application scenarios of edge computing can be very diverse, such as maintaining the operation of various structures in the energy industry or placing solutions on sensors at the edge; edge computing is an important tool for us. It is not limited to wearable device applications but can solve many internal enterprise problems. Currently, edge computing products are still under development and we are especially interested in healthcare applications.
What is the current development status of Beyond Limits in the global market? Is there a strong demand for enterprise AI solutions?
In addition to the energy industry, there are also expansions to infrastructure, medical care, etc., with the Hong Kong office focusing more on the financial sector. The products we make will not be limited to specific markets – they will be globalized. Some financial solutions currently produced in Hong Kong will also be further promoted to the global market in the future.
The development of some AI chips and smart manufacturing solutions will be the same and we won’t just be selling to Taiwanese manufacturers. After building a good product model, we will then export it to the global market.
Beyond Limits’ Series-C funding round was finalized in 2020 for a total of US$133 million. That same year the company expanded into the United Arab Emirates (UAE) and the Asia-Pacific region simultaneously, with offices in Singapore, Hong Kong, Taiwan, and Japan. There are currently about 15 employees in the Asia-Pacific organization while the Taiwan market is already starting to recruit talent and establish initial interactions with customers as well.
AI implementation often requires education, with awareness at the C-level being the most vital. The reason is that the industry must first recognize the need and necessity of introducing AI. Therefore, in addition to AI solutions, Beyond Limits now also offers digital transformation consulting, which may also become one of the entry points for entering the Taiwan market. We will first evaluate the progress of the digital transformation within a prospective company, and then observe the parts that need to be optimized in terms of business challenges. AI technology is the core of this optimization process. At present, digital transformation is being implemented relatively quickly in the UAE market, cooperating with local tier-one players to cover logistics, medical care, insurance, and more.
Based on current observations, the market demand for enterprise-level AI solutions is very large, which is very similar to the status of past software development. In the past, software changed internal operations and operating conditions for many companies; future applications of AI will similarly change the process of corporate decision-making – this being a common mainstream trend. When observing the process of AI introduction by companies in the Asia-Pacific region it would appear that large organizations in mainland China have introduced the technology very fast as they generally have a decent understanding of how AI will impact various processes in the future.
Can you talk a little about future layouts and planning for the Taiwan market?
As far as Taiwan is concerned, we will seek cooperation with relevant industry players in the fields we have entered such as energy. In particular, Taiwan’s manufacturing industry is very strong, and its management capabilities are also very good. Therefore, we will use Industry 4.0 and smart manufacturing as the entry point to find some partners in Taiwan.
Since entering the Taiwan market in 2020, many demands have actually been observed. In addition to the process and planning of Industry 4.0, the industry should also think about future development policies. I think our services for Taiwan’s manufacturing industry will be roughly divided into several parts. The first is supply chain management, which is a very important part of manufacturing and has a direct relationship with cost; the second is design for manufacturing.
As mentioned earlier, we have an AI platform, but what we lack is expertise in several industries. For example, the importance of bp to us lies in the fact that we were previously engaged in the space industry. By cooperating with tier-one players in the energy sector, like bp, we were able to capture essential expertise and further penetrate the industry. In summary, when we enter a new market, one of our primary goals is to do so with a partner who is involved in that industry. The main task in Taiwan is to find partners in manufacturing. What this means to us is not just about developing products – these industries with long histories often have a lot of important institutional knowledge that needs to be preserved, and this is what we are doing. Taiwan has a lot of management knowledge that is worth capturing and saving. After encoding this knowledge into the AI solution, the technology can play the role of monitoring and providing advice to ensure the smooth progress of processes.
Different manufacturing industries indeed have varying processes – but there are similarities among those processes as well; our system can be placed in various manufacturing processes. The key lies in the data captured by sensors and the human decision-making process. Optimizing the decision-making process is our core work.
In addition to actively looking for smart manufacturing partners in Taiwan, it is also a goal to combine our software technology with hardware applications. Past detector experience has given us edge computing technology, which can be miniaturized and/or work with low energy consumption characteristics; we can use this existing technology to find partners. Taiwan is famous for its manufacturing industry; this will be the best field in which to create more business opportunities and partnerships.
In addition to the above two factors, another reason for entering Taiwan’s market is that talents are also very attractive. We believe that Taiwan has a very good education and cultivation system, with the quality of engineers being top of the line as well.
There are many small and medium-sized enterprises in Taiwan. The solutions proposed by Beyond Limits seem to be mainly applied to large-scale enterprises. Under such circumstances, will small and medium-sized enterprises encounter any difficulties when introducing Beyond Limits or other AI solutions?
Indeed, many of our current solutions may be more suitable for larger companies. However, even for small and medium-sized enterprises, there will still be a decision-making process, and there will also be overlooked problems or fallacies in the process. No matter whether the scale is large or small, there will always be data that needs to be retrieved using a sensor. Keeping in mind the large number of small and medium-sized enterprises in Taiwan, there are many things that we can learn from Taiwanese companies with many opportunities to expand the application scope of Beyond Limits’ AI solutions in the future.
When we enter a new industry to develop customized solutions, our first priority is co-development with industry players. In Taiwan, where there are many small and medium-sized enterprises, the efficiency of development will also be at a higher level. This will be an asset that Taiwan brings to Beyond Limits while we will help industries continue to pass on and preserve the spirit and experience of essential expertise.