Ng Enda: Artificial intelligence is the main driver of innovation today

Author: Andrew Ng

Source: MIT Technology Review

Image source: generated by Unbounded AI tool

Advice for aspiring innovators on experimentation, failure, and the future of artificial intelligence.

Innovation is a powerful engine that drives social progress and economic growth. Antibiotics, lights, refrigerators, airplanes, smartphones -- we have these things because innovators created things that didn't exist before. MIT Technology Review's Innovators Under 35 list recognizes those who have achieved great things early in their careers and have the potential to achieve even more.

Over the years, I have been engaged in AI research and AI product development, and have been fortunate to be involved in some impactful innovations, such as using reinforcement learning to fly helicopter drones at Stanford University, and launching and leading Google Brain to drive deep learning at scale. , as well as creating online courses that led to the founding of Coursera. I want to share some ideas with you on how to do this well and avoid some pitfalls that can cause serious harm during the creation process.

AI is the primary driver of innovation today

As I said before, I believe AI is the new electricity. Electricity has revolutionized all industries and changed the way we live, and artificial intelligence is doing the same. It reaches into every industry and discipline, and the advances it has made have helped countless people.

Artificial intelligence, like electricity, is a general technology. Many innovations, such as medical, space rocket or battery designs, are only suitable for one purpose. In contrast, AI can be used to generate artwork, serve web pages relevant to search queries, optimize shipping routes to save fuel, help cars avoid collisions, and more.

Advances in artificial intelligence create opportunities for everyone in every sector of the economy to explore whether or how AI can be applied to their fields. Learning artificial intelligence can create more opportunities to do things that others have never done.

For example, at the AI Fund, the venture studio I lead, I have been fortunate to work on projects applying AI to maritime, relationship coaching, talent management, education, and other areas. Because many AI technologies are new, their applications in most fields have yet to be explored. In this way, understanding how to leverage artificial intelligence can provide you with numerous opportunities to collaborate with others.

Looking ahead, several developments are particularly exciting.

• TIP: While ChatGPT has popularized the ability to prompt AI models to compose emails or poems, software developers are just beginning to realize that prompting allows them to build powerful artificial intelligence in minutes that once took months. Smart application type. A large wave of AI applications will be built this way.
• Visual Transformer: Text Transformer - a language model based on the Transformer neural network architecture invented by Google Brain and co-authors in 2017 that revolutionized writing. The Vision Transformer, which enables transformers to adapt to computer vision tasks such as identifying objects in images, came out in 2020 and quickly gained widespread attention. The buzz around visual converters in the tech world these days reminds me of ChatGPT’s buzz around text converters a few years ago. The field of image processing will also usher in a similar revolution. Part of this revolution will be visual cues, where the cue is an image rather than a string of words.
• Artificial intelligence applications: The media pays great attention to artificial intelligence software and hardware infrastructure and developer tools. But this emerging AI infrastructure won’t succeed unless more valuable AI businesses are built on top of it. Therefore, although there is a lot of media attention on the AI infrastructure layer, there will be greater development in the AI application layer.

These areas provide rich opportunities for innovators. Furthermore, many of these areas are within reach of a broad technical elite, not just those already working in artificial intelligence. Online courses, open source software, software as a service, and online research papers provide everyone with the tools to learn and start innovating. But even if these technologies aren't within your grasp yet, many other avenues for innovation are wide open.

Be optimistic, but dare to fail

Still, many ideas that initially seemed promising ended up being squibs. If you take innovation seriously, failure is inevitable. Here are some of my projects that you may not have heard of because they were all failures:

• I spent a long time trying to get planes to fly autonomously in formation to save fuel (similar to birds flying in V formation). In hindsight, I didn't do a great job and should have used a larger aircraft.
• I have tried to get the robotic arm to unload dishes of all shapes and sizes in the dishwasher. But in hindsight, I did it too soon. Deep learning algorithms for perception and control were not yet complete at the time. *About 15 years ago, I thought unsupervised learning (i.e., letting machine learning models learn from unlabeled data) was a promising approach. However, the timing was not right. However, as data availability and computing power increase, this approach is finally working.

The failure of these projects pained me, but the lessons I learned were instrumental in the success of other projects. Through my failed attempts at V-flying, I learned to plan projects better and put risks forward. The effort to unload the dishes failed, but it led my team to build the Robot Operating System (ROS), which became a popular open source framework that is now used in a variety of robots from self-driving cars to robotic dogs. Although my initial focus on unsupervised learning was a bad choice, the steps we took were critical to scaling deep learning at Google Brain.

Innovation has never been easy. When you do something new, there are always going to be people who are skeptical. When I was younger, I encountered a lot of skepticism when starting most projects that ultimately turned out to be successful. But that's not to say the doubters are always wrong. I've also encountered doubt in most unsuccessful projects.

As I gain experience, I find more and more people agree with everything I say, which worries me even more. I had to actively seek out people who were willing to challenge me and tell me the truth. Luckily I have a lot of people around me these days who will tell me when they think I've done something stupid!

On the one hand, skepticism is OK and even necessary, but on the other hand, society has a strong interest in innovative results. This is also a good reason for us to treat innovation with optimism. I would rather be on the side of the optimist who wants to give it a try and might fail than on the side of the pessimist who doubts the possibilities.

Responsible for the work

As we focus on artificial intelligence as a driver of valuable innovation across society, social responsibility is more important than ever. People inside and outside the field are seeing all the harms that AI can cause. This includes both short-term issues, such as biased and harmful applications of technology, and long-term risks, such as the concentration of power and potentially catastrophic applications. It is important that we have an open and intellectually rigorous dialogue about these issues. That way we can all agree on what the real risks are and how to reduce them.

Over the past millennium, successive waves of innovation have reduced infant mortality, improved nutrition, increased literacy, raised living standards around the world, and advanced civil rights, including for women, minorities, and other marginalized people. protection of the group. However, innovation also contributes to climate change, exacerbates inequality, polarizes societies, and increases loneliness.

Clearly, the benefits of innovation come with risks, and we are not always able to manage these risks wisely. Artificial intelligence is the next wave, and we have an obligation to learn from past experiences to maximize future benefits and minimize harm for everyone. This will require commitment from both individuals and society as a whole.

At a societal level, governments are moving to regulate AI. For some innovators, regulation can be an unnecessary constraint on progress. I don't see it that way. As we move into an uncertain future, regulation can help us avoid mistakes and deliver new benefits. I welcome regulation requiring greater transparency into the opaque operations of big tech companies; this will help us understand their impact and guide them towards wider social benefits. Additionally, we need new regulations because many of the existing ones were written for a pre-AI world. New regulations should clearly spell out the outcomes we want, and the outcomes we don't want, in important areas like health care and finance.

But avoiding harm shouldn’t just be society’s top priority. It also needs to be a priority for every innovator. As technologists, we have a responsibility to understand the impact of our research and innovate in beneficial ways. Traditionally, many technologists have had the attitude that the shape of technology is inevitable and there's nothing we can do about it, so we might as well be free to innovate. But we know that's not the case.

When innovators choose to work on differentiated privacy—which allows AI to learn from data without exposing personally identifiable information—they make a powerful statement about the importance of privacy. This statement helps shape social norms adopted by public and private institutions. Conversely, when innovators create Web3 encryption protocols to launder money, it is also a powerful statement - and in my opinion, a harmful statement - that governments should not be able to track how money is moved and used.

If you become aware of unethical behavior, I hope you will raise it with your colleagues and superiors and engage in constructive dialogue with them. If you are asked to do something that you think is not good for humanity, I hope you will actively work to stop it. If you can't do that, consider leaving. At the AI Fund, I have terminated a number of projects that I considered financially sound but ethically unsound. I urge you to do the same.

Go ahead and innovate! If you're already in the innovation game, keep going. What great things you will achieve in the future is unknown. If your ideas are still in the dream stage, share them with others and get help turning them into practical successes. Start executing and find ways to use the power of innovation to do good.