2041 envisioned: AI-driven futures according to Kai-Fu Lee
The International Telecommunication Union (ITU) recently connected with pioneering artificial intelligence (AI) expert Kai-Fu Lee, former president of Google China, CEO of Sinovation Ventures, and co-author of the forthcoming book “AI 2041: Ten visions for our future.”
Here, Lee shares insights on how much – and how deeply – AI could shape our world in the decades to come.
ITU: Your new book is a collection of short stories that speculate on ways AI could shape daily life and experiences in different parts of the world.
In one story, AI companions form an integral part of the education system. How would this affect young children, such as those under age 5?
Kai-Fu Lee: Classic childhood companions like Barbie or GI Joe – once inanimate objects – will come to life on mobile phone screens, or through virtual reality (VR) or augmented reality (AR) glasses. They will converse in an increasingly more natural language too.
They don’t have to know everything, just how to converse with you in ways you like and that represent their character.
This can help kids learn things in a fun way – like multiplication or division – before going into school. Or recast problems: math equations could become basketball games. Superheroes and animals can also make education more fun and targeted.
What do you mean by targeted education?
Today, Facebook and TikTok are so good at targeting us with content to keep us clicking.
To use that same tech for good would mean targeting content to students so they are naturally incentivized to learn.
Applying AI at the individual level can help students learn at a pace that suits their individual preferences and passion.
Do human teachers fit into this speculative future?
Teachers can still play a strong role as mentors, helping improve the child’s curiosity and creativity, communication skills, teamwork, and compassion. The human teacher becomes more of a mentor to develop values and skillsets, whereas the AI companion makes learning personalized. That combination will help the next generation grow up better than previous ones.
Are there risks to such deep levels of personalization? Might people eventually prefer communicating with their favourite bots rather than spouses or friends?
That may happen in my lifetime – but it wouldn’t happen to me!
People connect soul-to-soul, because we are made of the same organic materials.
We feel a sense of love and empathy from people, not from robots.
Bots may very accurately emulate what I say and do, and appear interested in human life, but inside they are just matching patterns without self-awareness, emotion or connection. Remind yourself: AI will not love you back!
The second story has to do with deepfakes – AI-generated images in which a person is replaced with someone else’s likeness, with great potential for deception. How do you see deepfakes evolving?
Deepfakes of audio, video, and images are already used to sling mud online. Finding them in courtroom evidence is a likely path.
We are already at a point where human eyes can barely tell the difference between real and fake.
Even computers will have a hard time telling. Whether they can differentiate depends on computing power – which becomes an arms race between bad guys and good guys.
Can anything be done to mitigate the problem?
Authoritative websites – for governments, the United Nations, hospitals and major news organizations, for example –must do everything they can to control quality, especially with content uploaded by users.
Deepfake scanning mechanisms must be built into all websites, just like virus scanning software. Even then, some deepfakes will get through. We need to get used to it.
Perhaps technologies could authenticate content at time of capture, making it impossible to modify as a deepfake.
But upgrading every camera, phone, or other capture module will take 20 years to happen.
Speaking of 2041 visions, one character feels practically blind without her so-called “XR” contact lenses. Is this the future of AR/VR?
The problem in AR/VR today is that headsets are so big and heavy. They look nerdy and also suffer issues with realism, resolution and rendering, and can even make people dizzy.
A lot of these problems will get fixed in the next few years. We will likely end up with glasses that are not thicker or heavier than normal frames, which display more realistic-feeling content than current interfaces.
Games like Pokémon Go will work with such glasses. AR/VR can also be used for education – tours with historical characters, for example. Or for training: learn to fix an airplane in virtual space.
Product research is happening. We should have AR glasses in a ten-year timeframe.
On the future of work, we meet an optimist and a pessimist. The former says technology replaces certain jobs, but humans always invent new jobs to meet new needs. The latter says it’s different this time; all but a few jobs will be replaced by AI. Which is it?
Both. AI is good at doing repetitive routine tasks, like customer service, telemarketing, simple assembly-line jobs, delivering packages, waiting tables, and soon, perhaps, even driving. These will be the first to go.
Companies will purchase AI products and services, so 30-40 per cent of jobs will be replaced in the next 20 years. In the longer term, AI will create new job opportunities – some of which are still impossible for us to conceive now. Still, data labelling needs people; repairs need to be done by humans.
So AI will indirectly create jobs in the service sector. Only humans can do human-to-human services, such as in healthcare.
Those will grow as people live longer, and people will be willing to pay more for human-to-human services.
What about AI in developing countries? Can the benefits reach everyone by 2041? In our recent Machine Learning in 5G Challenge, some participants lacked the computing resources to train their models – some needed up to 20 days. How could developing countries catch up?
Unfortunately, we are going down a path of growing inequality between countries. The UN can try to address this. Developing countries that cannot offer advanced AI education to all might consider allocating resources to programmes for the gifted and talented, or sending the best students on a scholarship to countries with strong AI education. Some of those countries will even pay to receive top students.
This approach was critical for past growth in China. While some students stayed in other countries, some went back to China and became professors or founded tech companies.
Those can form the new tech or AI nucleus in their home country.
Watch the full AI for Good interactive keynote here.
Image credit: Rawpixel via Freepik