Dialogue with MEGA Robots: The Development of AI for Science Urgently Needs New Intelligent Infrastructure | Exclusive Interview by 36Kr

Text | Wang Fangyu

Editor | Su Jianxun

After the 2024 Nobel Prize in Chemistry was announced, the new paradigm of scientific research, "AI for Science", has come into the spotlight and received widespread attention.

AI for Science (AI4S), that is, artificial intelligence-driven scientific research. The decision of the Nobel Prize in Chemistry has largely affirmed the long-term value of this research paradigm and even the entire track. Liu Qingfeng, the chairman of iFLYTEK, even stated directly at a recent conference that AI for Science determines the speed of China's scientific and technological development.

AI4S is not only an AI model. Its ability to create "miracles" benefits from the entire process from data collection to model construction, from experimental design to process control. There is a huge and developable entrepreneurial system and business space behind it.

In particular, data, large-scale and accurate data is still a scarce resource. However, the current situation is that the equipment and consumables in most domestic life science laboratories are still designed for manual operation and do not have the ability to accumulate data on a large scale.

This has become a major bottleneck in the development of AI4S in the domestic life science field. Huang Yuqing, the founder and CEO of MEGAROBO, described the above problem as "the contradiction between the scientists' increasing demand for data and the relatively backward productivity."

Catching this trend, MEGAROBO has been using the strength of the company to build the MEGAROBO Kunpeng Laboratory since 2019, and it began to be put into use in 2021.

The MEGAROBO Kunpeng Laboratory integrates technologies such as robotics, automation, and artificial intelligence. It can not only provide automated tools for life science experiments, but also accumulate data for AI for Science. MEGAROBO defines it as the "data factory of the life science industry".

In September this year, the MEGAROBO Kunpeng Laboratory 3.0 was officially released, focusing on research in the fields of cell gene editing, high-throughput drug screening, organoid automation, traditional Chinese medicine research and discovery, synthetic biology, etc.

Recently, Huang Yuqing, the founder and chief executive officer of MEGAROBO, Zhang Yan, the co-founder and chief operating officer, and Dr. Fang Panfeng, the chief scientist, had an exchange with 36Kr and other media, explaining MEGAROBO's layout in the life science field and thoughts on industry trends.

The following is the exchange record between 36Kr and MEGAROBO (edited):

Media: Why did MEGAROBO build the MEGAROBO Kunpeng Laboratory, and what trend did it see behind it?

Huang Yuqing: When we started the business in 2017, we had a very simple idea to make a robot that is easy to deploy and use quickly, and can be deployed to millions of life science laboratories around the world. But soon we hit a brick wall. We found that in the life science industry, the scientists responsible for research and development do not need a single robot, but an entire solution.

So since 2019, we have decided to transform and start building a super laboratory. In the past few years, we have spent more than 1.2 billion yuan on technology research and development. Up to now, this laboratory has been iterated three times.

Media: What is the relationship between the MEGAROBO Kunpeng Laboratory and AI for Science?

Zhang Yan: The data required by the AI research institute in life science is massive. To obtain massive data, the quality of the experiment must be guaranteed first. The entire process such as cultivation and sequencing must be consistent and controllable, and then the data must be connected, managed, classified, and collected, and then turned into data available for training the model.

However, the previous industry practice is that scientists do experiments manually, and the consistency and quality of the experiments are difficult to guarantee. In addition, the collection of data also relies on primary tools such as Excel. It can be said that a major trend and the main contradiction faced by the life science industry now is the contradiction between the scientists' increasing demand for data and the relatively backward productivity.

The value of the MEGAROBO Kunpeng Laboratory lies in that it can accelerate the model iteration by designing experiments, implementing experiments, testing results, and adjusting the model according to the results to form a dry-wet closed loop, helping humans make more accurate predictions in the field of life science. Therefore, we describe the MEGAROBO Kunpeng Laboratory as the "data factory of the life science industry".

Media: How does MEGAROBO ensure the high quality and effectiveness of data?

Fang Panfeng: Automatic machine extraction is definitely a trend. Otherwise, once it involves high-throughput and large-scale queues, it is actually impossible to complete by humans.

Huang Yuqing: The equipment in life science laboratories usually comes from different brands, and the interfaces are also different, so there is a lot of noise in the data format itself. These raw data cannot be fed to the model, and even the correlation between data from different devices is also a problem. Therefore, MEGAROBO has built a Megalnfomics digital intelligence platform, which includes central scheduling software, data integration software, information management software, bioinformatics analysis software, etc. These softwares ensure the effective collection, structuring, analysis and traceability of data.

Zhang Yan: The experimental throughput of this generation of the MEGAROBO Kunpeng Laboratory can be increased by 40 times compared to the previous generation, the data generation ability can be increased by 100 times, and the data processing ability can be increased by 1000 times.

Media: From selling individual equipment to building the MEGAROBO Kunpeng Laboratory, has MEGAROBO's business model changed?

Huang Yuqing: The MEGAROBO Kunpeng Laboratory is a business model of technical services. Customers put forward experimental requirements, and the MEGAROBO Kunpeng Laboratory outputs the final experimental results to the customers. It can be understood as an upgraded version of CRO (Contract Research Organization) with AI and automation. We not only sell products, but also services. I believe that the proportion of services will be higher and higher in the future.

For the future positioning of MEGAROBO, in addition to providing customers with automated equipment, we hope to become a new generation of drug research and development service provider empowered by AI and automation, and become a new generation of CRO and CDMO.

Media: In the past two years, AI pharmaceutical company XtalPi has also been laying out automation, from software to hardware; while MEGAROBO is from hardware to software layout. Is the automation in the life science field somewhat convergent?

Huang Yuqing: In the end, it is likely to be convergent, which is determined by the needs of customers.

For example, in the process of building a model, a large amount of data is definitely required. In order to obtain this data, the corresponding tools must be used. Our strategy is to independently develop tools first, and in the process of developing tools, we naturally accumulate a large amount of data.

Subsequently, many customers gave us feedback that our tools are excellent and have rich data resources. They further asked us if we could help them better process and process these data. Indeed, in the process of data processing, the close combination of software and hardware is crucial. Therefore, we are also very willing to provide customers with such a software platform to help them solve the problem of data processing.

Based on such observations and practices, I believe that in this industry, top companies will eventually move towards the combination of software and hardware.

Media: After MEGAROBO launched the MEGAROBO Kunpeng Laboratory 3.0, what are the plans for market expansion?

Huang Yuqing: I believe that in any economic environment, there are细分 markets with rapid growth. The key is who can keenly seize these opportunities. MEGAROBO is positioned to provide advanced productivity tools and serve many industries. Therefore, our growth must come from the growth of customers. We must always maintain a keen market insight, accurately judge which industries are experiencing growth, and evaluate whether we have the ability to seize these growth opportunities. Through practice, it has been proved that our model and strategy are effective.

In the scientific research or life science industry, the overall payment ability is relatively strong, and the budget is also relatively sufficient. At present, the industry as a whole is also advocating domestic substitution. So for MEGAROBO, on the one hand, it benefits from the trend of upgrading from manual equipment to a highly automated unmanned laboratory; on the other hand, the data capabilities of the MEGAROBO Kunpeng Laboratory are stronger than traditional equipment. Under multiple market trends, we just need to follow the trend.

Media: Is AI plus automation the development direction of domestic life science automation enterprises in the future?

Huang Yuqing: Yes, just like the autopilot of electric vehicles, to fully exert the potential of AI for Science, the underlying equipment, including automated equipment, needs to be reconstructed. Future scientific instruments and equipment must be designed and developed for such a trend.