Tech Junction IV: Deep learning, AI, and what it promises for the future

Artificial intelligence – the point at which information science, neurology, and electrical engineering combine – creates fascinating questions for humanity.

A prominent scientist in this field is Dr. Yang Ming, Co-founder & Vice President, Software of Horizon Robotics, a Beijing-based startup researching and developing Artificial Intelligence and Robotics. Yang is an expert in computer vision and machine learning, and was a founding member of the Facebook Artificial Intelligence Research (FAIR) organization, leading Facebook’s deep learning research initiative on face recognition.

Yang gave a presentation at a Tech Junction event in Beijing, followed by an audience Q&A hosted by AllChinaTech Co-founder Wu Nan.

Ming
Dr. Yang Ming, Co-founder of Horizon Robotics

What’s Artificial Intelligence?

Yang defines Artificial Intelligence (AI) as “the study and design of intelligent agents, in which an intelligent agent perceives its environment and takes actions that maximize its chance of success”. In practice, this involves three steps: sensing (observing the physical environment or taking in other sensor signals), thinking (reasoning how to perform the task), and acting (performing the task using autonomous decisions).

He stresses the word agent is used rather than computer, given the range of devices that can or will have AI as a component.

“In 5 years, every smart device will be running AI technology, ” he said.

Deep learning and neural networks

The initial prototype of artificial neural networks was devised way back in 1957, but their importance has exploded in the last 5 years. Essentially deep learning is a process of machine learning that is modelled on or at least, in some ways mimics the human brain: that software can simulate connections between virtual “neurons”. The links between these neural networks are formed through data, rather than directly programmed, allowing the computer to “learn”. The more data you feed in, the better the computer can “learn” and perform.

According to Yang, 2010 saw a massive breakthrough of computer speech recognition driven by deep learning; this prompted google to start its own “google brain” project. In 2012, there was another breakthrough using deep learning for image recognition. In 2014 the idea of autonomous driving became widespread.

AI related business

The interest that AI has attracted from the internet industry is huge.  Yang quoted the CEOs of Google and Facebook both talking about a desire for practical AI assistants to help them make decisions.

Companies involved in AI R&D involve many of the massive internet companies, but also many smaller companies and startups. Big companies include Apple, Amazon, Baidu, Google, Tesla, et cetera. Obviously, the direction of their R&D is in line with the business model of each business. Right now, Uber is researching AI for autonomous vehicles, but not face recognition. Semiconductor giants like Qualcomm, ARM, Nvidia, and Xillinx are also researching deep learning.

When it comes to startups, Yang cited research by VentureScanner stating that there were at least 957 of them working in AI worldwide, with a combined USD 4.8 billion in funding.

The most promising fields for AI

Discussing direct applications of AI to human society, Yang suggests the following areas: Digital/virtual assistants, autonomous vehicles, smart robots and devices, healthcare, and manufacturing.

A digital assistant could be an organizing program attached to something like a computer, but with a sophisticated AI making its suggestions more realistically useful to a normal life, just like Baidu’s AI robot Duer. In terms of smart devices, Yang is especially optimistic about the phenomenon of AI combining with mobile computing and the “Internet of Things” phenomenon, where internet connectivity is available on a vast range of everyday appliances.

MingnadNan
Yang Ming and Wu Nan at Tech Junction

Following Yang’s talk, he had a quick Q&A with us:

AllChinaTech:  Are the AI risks for mankind exaggerated? Why or why not?

Yang: My personal position is that the risk is exaggerated. There are several reasons for this. Firstly, nowadays, AI are programmed by humans. Not a single human, but a bunch of human programmers, all of whom check each other’s work.

Another reason is that the AI that so far, the AI that we have are all designed with a specialized functions; they’re not so-called “strong AI” or “general intelligence”. They are designed for specialized tasks: for example, autonomous driving AI is complex in many ways, but it essentially just wants to drive you from place A to place B. Even if it were technologically possible, you wouldn’t want to put emotions or consciousness in your car.

The third reason is that the research community has no idea of how to represent emotion or self-awareness in AI. Essentially, we would need to create something digital…but we are simply just not there yet. In the current form of AI, there are no risks.

AllChinaTech: From the perspective of individual startups, do AI projects require burning a lot of cash to succeed?

Yang: That depends on your strategy. You can start from something concrete. Every startup wants to change the world, but you can change the world a little bit at the time.

You certainly don’t want to burn your money in the beginning, but you can start from something small and meaningful, and start making money from that.

Alternatively, if you manage to raise half a billion dollars, then burning money is not a problem (laughs).

AllChinaTech: Finally, what is the next big thing from Horizon Robots?

The next thing that we’re up to – I’m not sure if it’s big – we’re working very hard to have our own co-processor launched.

About Tech Junction:

A meetup group in Beijing with a stated goal of exchanging ideas in short, sharp dialogues among startups, tech and media professionals in order to shed light on the tech scene in China. Our mission is to establish a tech community to connect with thinkers and dealmakers around the world. This monthly meetup is co-organized by AllChinaTech and DayDayUp. Find out when the next Tech Junction is on meetup.com.

AllChinaTech is a Beijing-based startup media platform, dedicated to providing timely news and analysis on the Chinese tech industry in English.

DayDayUp is a co-working space in Beijing creating an international collaboration community for entrepreneurs from China and all countries.

© 2022 All Tech Asia. All Rights Reserved.

AllTechAsia is a startup media platform dedicated to providing the hottest news, data service and analysis on the tech and startup scene of Asian markets in English. Contact us: info@alltechasia.com


%d bloggers like this: