AI 2041: Ten Visions for Our Future

Regardless of these theories, I believe it’s indisputable that computers simply “think” differently from our brains.

To the AI, such things were a matter of math, not love.

IT IS BETTER TO LIVE YOUR OWN DESTINY IMPERFECTLY THAN TO IMITATE SOMEBODY ELSE’S PERFECTLY.

RENEWABLE ENERGY REVOLUTION: SOLAR + WIND + BATTERIES In addition to AI, we are on the cusp of another important technological revolution—renewable energy. Together, solar photovoltaic, wind power, and lithium-ion battery storage technologies will create the capability of replacing most if not all of our energy infrastructure with renewable clean energy. By 2041, much of the developed world and some developing countries will be primarily powered by solar and wind. The cost of solar energy dropped 82 percent from 2010 to 2020, while the cost of wind energy dropped 46 percent. Solar and onshore wind are now the cheapest sources of electricity. In addition, lithium-ion battery storage cost has dropped 87 percent from 2010 to 2020. It will drop further thanks to the massive production of batteries for electrical vehicles. This rapid drop in the price of battery storage will make it possible to store the solar/wind energy from sunny and windy days for future use. Think tank RethinkX estimates that with a $2 trillion investment through 2030, the cost of energy in the United States will drop to 3 cents per kilowatt-hour, less than one-quarter of today’s cost. By 2041, it should be even lower, as the prices of these three components continue to descend. What happens on days when a given area’s battery energy storage is full—will any generated energy left unused be wasted? RethinkX predicts that these circumstances will create a new class of energy called “super power” at essentially zero cost, usually during the sunniest or most windy days. With intelligent scheduling, this “super power” can be used for non-time-sensitive applications such as charging batteries of idle cars, water desalination and treatment, waste recycling, metal refining, carbon removal, blockchain consensus algorithms, AI drug discovery, and manufacturing activities whose costs are energy-driven. Such a system would not only dramatically decrease energy cost, but also power new applications and inventions that were previously too expensive to pursue. As the cost of energy plummets, the cost of water, materials, manufacturing, computation, and anything that has a major energy component will drop, too. The solar + wind + batteries approach to new energy will also be 100-percent clean energy. Switching to this form of energy can eliminate more than 50 percent of all greenhouse gas emissions, which is by far the largest culprit of climate change.

Imagine, a $1,000 political assassin! And this is not a far-fetched danger for the future, but a clear and present danger.

In the story of AI and humans, if we get the dance between artificial intelligence and human society right, it would unquestionably be the single greatest achievement in human history.

The popular 2020 documentary The Social Dilemma illustrates how AI’s personalization will cause you to be unconsciously manipulated by AI and motivated by profit from advertising. The Social Dilemma star Tristan Harris says: “You didn’t know that your click caused a supercomputer to be pointed at your brain. Your click activated billions of dollars of computing power that has learned much from its experience of tricking two billion human animals to click again.” And this addiction results in a vicious cycle for you, but a virtuous cycle for the big Internet companies that use this mechanism as a money-printing machine. The Social Dilemma further argues that this may narrow your viewpoints, polarize society, distort truth, and negatively affect your happiness, mood, and mental health. To put it in technical terms, the core of the issue is the simplicity of the objective function, and the danger from single-mindedly optimizing a single objective function, which can lead to harmful externalities. Today’s AI usually optimizes this singular goal—most commonly to make money (more clicks, ads, revenues). And AI has a maniacal focus on that one corporate goal, without regard for users’ well-being.

Deepfakes are built on a technology called generative adversarial networks (GAN). As the name suggests, a GAN is a pair of “adversarial” deep learning neural networks. The first network, the forger network, tries to generate something that looks real, let’s say a synthesized picture of a dog, based on millions of pictures of dogs. The other network, the detective network, compares the forger’s synthesized dog picture with genuine dog pictures, and determines if the forger’s output is real or fake.

While humans lack AI’s ability to analyze huge numbers of data points at the same time, people have a unique ability to draw on experience, abstract concepts, and common sense to make decisions. By contrast, in order for deep learning to function well, the following are required: massive amounts of relevant data, a narrow domain, and a concrete objective function to optimize. If you’re short on any one of these, things may fall apart.

So, is 100-percent detection of deepfakes hopeless? In the very long term, 100-percent detection may be possible with a totally different approach—to authenticate every photo and video ever taken by every camera or phone using blockchain technology (which guarantees that an original has never been altered), at the time of capture. Then any photo loaded to a website must show its blockchain authentication. This process will eliminate deepfakes. However, this “upgrade” will not arrive by 2041, as it requires all devices to use it (like all AV receivers use Dolby Digital today), and blockchain needs to become fast enough to process this at scale.

So, will deep learning eventually become “artificial general intelligence” (AGI), matching human intelligence in every way? Will we encounter “singularity” (see chapter 10)? I don’t believe it will happen by 2041. There are many challenges that we have not made much progress on or even understood, such as how to model creativity, strategic thinking, reasoning, counter-factual thinking, emotions, and consciousness. These challenges are likely to require a dozen more breakthroughs like deep learning, but we’ve had only one great breakthrough in over sixty years, so I believe we are unlikely to see a dozen in twenty years. In addition, I would suggest that we stop using AGI as the ultimate test of AI. As I described in chapter 1, AI’s mind is different from the human mind. In twenty years, deep learning and its extensions will beat humans on an ever-increasing number of tasks, but there will still be many existing tasks that humans can handle much better than deep learning. There will even be some new tasks that showcase human superiority, especially if AI’s progress inspires us to improve and evolve. What’s important is that we develop useful applications suitable for AI and seek to find human-AI symbiosis, rather than obsess about whether or when deep-learning AI will become AGI. I consider the obsession with AGI to be a narcissistic human tendency to view ourselves as the gold standard.

Some of these bots are already arriving in 2021 in more primitive forms. Recently, when I was in quarantine at home in Beijing, all of my e-commerce packages and food were delivered by a robot in my apartment complex. The package would be placed on a sturdy, wheeled creature resembling R2-D2. It could wirelessly summon the elevator, navigate autonomously to my door, and then call my phone to announce its arrival, so I could take the package, after which it would return to reception. Fully autonomous door-to-door delivery vans are also being tested in Silicon Valley. By 2041, end-to-end delivery should be pervasive, with autonomous forklifts moving items in the warehouse, drones and autonomous vehicles delivering the boxes to the apartment complex, and the R2-D2 bot delivering the package to each home. Similarly, some restaurants now use robotic waiters to reduce human contact. These are not humanoid robots, but autonomous trays-on-wheels that deliver your order to your table. Robot servers today are both gimmicks and safety measures, but tomorrow they may be a normal part of table service for many restaurants, apart from the highest-end establishments or places that cater to tourists, where the human service is integral to the restaurant’s charm. Robots can be used in hotels (to clean and to deliver laundry, suitcases, and room service), offices (as receptionists, guards, and cleaning staff), stores (to clean floors and organize shelves), and information outlets (to answer questions and give directions at airports, hotels, and offices). In-home robots will go beyond the Roomba. Robots can wash dishes (not like a dishwasher, but as an autonomous machine in which you can pile all the greasy pots, utensils, and plates without removing leftover food, with all of them emerging cleaned, disinfected, dried, and organized). Robots can cook—not like a humanoid chef, but like an automated food processor connected to a self-cooking pot. Ingredients go in and the cooked dish comes out. All of these technology components exist now—and will be fine-tuned and integrated in the decade to come. So be patient. Wait for robotics to be perfected and for costs to go down. The commercial and subsequently personal applications will follow. By 2041, it’s not far-fetched to say that you may be living a lot more like the Jetsons!

EVERYTHING IS INTERWOVEN AND THE WEB IS HOLY. —MARCUS AURELIUS

To put it in technical terms, the core of the issue is the simplicity of the objective function, and the danger from single-mindedly optimizing a single objective function, which can lead to harmful externalities.

the greatest value of science fiction is not providing answers, but rather raising questions.

Go is a board game more complex than chess by one million trillion trillion trillion trillion times.

Many people think smartphones and apps already know too much about us, but XR will take things to a whole new level.

More data leads to better AI (artificial intelligence), more automation leads to greater efficiency, more usage leads to reduced cost, and more free time leads to greater productivity. All of these will grow into a mutually reinforcing virtuous circle that will continually and rapidly increase the adoption of AV (autonomous vehicles).