Interview with Dr. Ling Shao, President of Terminus International How far away are we from AGI?

Recently, Terminus Group, the leading global smart service provider, has appointed IEEE Fellow, Dr. Ling Shao, as its Chief Scientist and President of Terminus International, who will further enhance the strength of the company’s R&D capabilities and develop the company’s international expansion and localization.

Dr. Shao is an internationally renowned AI scientist. Before joining Terminus Group, Dr. Shao served as CTO and Chief Scientist of the National Center for Artificial Intelligence (NCAI) at the Saudi Data and AI Authority (SDAIA). From 2018 to 2021, he was the founding CEO and Chief Scientist of the UAE’s Inception Institute of Artificial Intelligence (IIAI). He is also the initiator of the world’s first AI university - Mohamed bin Zayed University of Artificial Intelligence, for which he served as its founding Provost and Executive Vice President. Dr. Shao has published over 300 scientific papers at top AI conferences or journals including CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, TPAMI and IJCV. For four consecutive years, between 2018 and 2021, Dr. Shao was selected as a “highly cited researcher" by the Web of Science.

Taking Dr. Ling Shao’s perspective on Terminus Group’s business and AI industry development, an interview was recently conducted with him. Dr. Shao indicated that the main reason he decided to join Terminus Group was that his research field matched closely with the company’s future ambition and business development. Dr. Shao also shared his views about the current challenges in the AI industry, as well as future developments and breakthroughs. According to Dr. Shao, the main reason why many AI companies struggle to be profitable is that AI solutions require a lot of customization and they are not scalable; however, large pre-trained models can now, in part, solve this problem, with research advancement - in designing relatively smaller pre-trained models and efficient model compression techniques - helping to accelerate AI profitability.

Interview Q&A:

1. What can you tell us about yourself and your background, before joining Terminus Group?

Hello, I'm Ling Shao. I consider myself to be an AI researcher. I was brought up and had most of my education in China. I did my first degree at the University of Science and Technology of China. Then I went to the UK to do my master’s and PhD, at the University of Oxford. After finishing my PhD, I worked at Philips Research in the Netherlands as the senior scientist for four years. Then I went back to the UK and worked for several universities as a professor of AI.

In January 2018, I came to the UAE and started the Inception Institute of Artificial Intelligence (IIAI) where I was the founding CEO and Chief Scientist, until December 2021. During that period of time, I also initiated the world’s first AI university, Mohammed bin Zayed University of Artificial Intelligence, for which I was the founding Provost and Executive Vice-President. Earlier this year, I also worked briefly as CTO and Chief Scientist of the National Center for Artificial Intelligence (NCAI) at the Saudi Data and AI Authority (SDAIA).

Just recently, I joined Terminus Group as the Chief Scientist and as President of Terminus International.

2. What made you decide to join Terminus Group？ What's your impression about Victor AI, Founder and CEO of Terminus Group?

Victor is one of the main reasons why I decided to join Terminus Group. Before joining, Victor and I had several video meetings where he introduced his vision about Terminus and his thoughts about AI and IoT, which align very well with my own ideas. We both think that, in the future, AI devices and AI systems should be inter-connected. A typical example is the autonomous vehicles. If the vehicles are interconnected with each other and with the other road users, such as pedestrians, autonomous driving will become much more feasible.

I was impressed by Victor's general knowledge about the AI industry, although his background is mainly on investment. He also knows a lot of senior AI experts and researchers, in China; some of whom are also my acquaintances.

Another reason I decided to join Terminus is that there are a lot of application scenarios with Terminus products and solutions where AI can play a key role. As an AI researcher, I’m extremely excited about applying my research to new areas.

3. Which Terminus' products appeal to you the most? Which technologies are you looking forward to integrating with Terminus’ business?

Terminus Group’s robots and AI CITY products are very interesting to me, both of which require comprehensive AI solutions and are very challenging. My research areas are mainly computer vision, deep learning, and vision and language. These are the key technologies that can enable sensing and interaction, which are two of the major functions of our robotic and AI CITY products. So, I am very excited about integrating some of my research findings with Terminus products and solutions.

4. What are your roles and responsibilities?

My first role is the Chief Scientist of Terminus Group. I'm going to build a team of AI researchers and engineers to enhance the AI capabilities of the group. Some of the team members will focus on long-term research and others will work closely with different business units to solvechallenging AI-related issues, across both product and solution development.

My other role is the President of Terminus International where I will oversee all our activities outside of China including R&D, product development and localization and sales. We are also going to establish partnerships with some local government agencies, companies and universities.

5. Terminus Group’s AI Computing Center provides core AI elements (AI models, data, computing power) to SMEs. What do you think of the AICC concept？

I think it is a great idea. As you know, AI is not straightforward – it requires GPUs, big data and the talent to develop AI models. Not every company has the need or the resources to develop their own AI from scratch. Terminus Group's AI Computing Center can serve as a platform for startups, SMEs and academic institutions to more quickly and cost-effectively adopt AI technologies.

6. During Expo 2020 Dubai, Terminus Group’s robots received a lot of attention. What do you think of the AI technologies behind them and what will make them stand out in the market?

Terminus Group' robots were indeed highly recognized at the Expo 2020 Dubai. But, of course, our product coverage is much broader than just robots. In terms of AI technologies, a lot of computer vision and speech-related techniques have already been implemented in our robots – for navigation, object recognition and so on. We will keep improving our vision and NLP models to enable advanced sensing and more natural interaction in our future robotic products.

Since Terminus is an AIoT company, all our robots will be connected with each other and with other objects and will be operating in an AIoT environment. This will make our robots stand out from competitors.

7. Where are we now in terms of AI development?  How far away are we from AGI (Artificial General Intelligence)?

The current AI systems are mainly based on deep learning. On the one hand, deep learning has made a step change to the performance of many vision, speech and NLP-related applications. In some areas, such as face recognition or playing go, AI can already outperform humans. On the other hand, the current AI is based on brute-force and big data. In the past few years, large pre-trained models have become very popular, mainly in the US and China. Once a large model is pre-trained, only limited domain data is required to fine-tune the model for a new scenario. This has totally changed the paradigm of AI model development, from customization-based to mass production.

Models such as GPT-3 can produce human-like documents, novels, poetry and so on. Some AI experts think this is a big step towards artificial general intelligence (AGI).  Personally, I think it has nothing to do with AGI because it does not have common sense, it cannot think or reason at all. The excellent performance is mainly due to the massive data we use to train the models, and the models are usually huge and contain hundreds of billions of parameters. So, it is basically memory with some generalization.

In short, we are still far, far away from AGI or maybe AGI cannot even be achieved in the future.

8. Although the AI industry continues to boom, most AI companies are still facing losses. What difficulties need to be overcome in AI profitability?

I think the main reason why AI companies are not profitable is that AI solutions require a lot of customization and they're not scalable. Now the large pre-trained models can partly solve this problem, but not every AI startup can afford to train their own large models. Therefore, research advancement in designing smaller pre-trained models that still work well and efficient model compression techniques can help to accelerate AI profitability.

9. What do you think are the most promising AI applications that will change the world in the next 10-20 years?

I will give a few examples. First, AI for healthcare, including AI for diagnosis and intervention and AI for drug discovery.

Second, robotics, especially the companion robots which can take care of the elderly – because we have an aging population.

Last but not least, autonomous vehicles will be ubiquitous within 10 to 15 years.

I think all these application areas will be promising, in the coming years.