Could AI help us create machines to imagine?
Human creativity is the elixir that has propelled civilizations through the ages. It has brought us incalculable breakthroughs in all sectors of the economy, from agriculture to health, from energy to mobility. Our imagination continues to be our savior as the world’s aging population faces difficult socio-economic and environmental challenges.
Imagination is about creating mental models of things that don’t yet exist. This kind of innovation brought us the printing press, the steam engine, the light bulb, the telephone, the airplane, the television and the PC. All of this required unlikely acts of human invention. Yet many companies fail to harness the creativity of their employees, even though they say it’s an essential trait for success.
Martin Reeves, co-author of a recently published book titled The Imagination Machine: How to Spark New Ideas and Create the Future of Your Business, is president of the BCG Henderson Institute, a think tank created by the Boston Consulting Group. He believes that “we have to bet on the imagination because a competitive advantage does not last very long. If you were the leader in your industry in the 1980s, you could expect to be at the top for at least 10 years. Now that period has dropped to one or two years.
He continues, “This means companies can’t just focus on optimizing yesterday’s business model; they must create new ones. We must generate growth through our creativity.
Computers still struggle to mimic this unique human process. Even the most powerful machines cannot imagine a product or service that has never been seen before. They can’t talk about the new iPhone or Tesla of tomorrow. (A lot of humans don’t either, for that matter.) But the tide is turning.
Humans can disassemble and recombine learned knowledge to design new images – think of a red boat, visualize a blue car, then imagine a blue boat, for example. A team of artificial intelligence experts from the University of Southern California (USC) is studying how to artificially emulate the process. This involves a concept called detangling. Humans break down the things they learn and visualize them as colors, shapes, and types. We are then able to recombine these attributes to form new images. The researchers replicated this using neural networks.
“The new approach truly unleashes a new sense of imagination in AI systems, bringing them closer to a human understanding of the world,” says researcher Yunhao Ge, who leads the project at USC.
Disentanglement is not a new concept. Indeed, it is used to create what is known as deepfake content, usually by recombining separate images of well-known people. AI has also been used in this way to invent a new sport called speedgate. Data from 400 popular sports were fed into a neural network. Many suggestions popped up, but the winner involved merging elements of football, croquet and rugby on a pitch with teams of six players, one ball and three gates.
The USC system takes an entire group of images or sample data, rather than one at a time, as traditional algorithms have done. The system examines the similarities between these images, which it then recombines using all the data to create something new. The process is known as new controllable image synthesis.
What is new is that the system can process almost any type of data. This could eventually aid in new drug discovery by combining the learned functions of existing drugs to form innovative treatments. Technology could untangle racial and gender biases to create fairer algorithms or create safer driverless cars by simulating countless crashes. The method may even be able to create new data sets by imagination.
“In the future, AI could be effectively limitless,” predicts Michael Conway, practice lead for AI and analytics at IBM. “He will be able to see connections between things that humans cannot. A more advanced version will be able to look beyond the parameters we have set and offer valuable insights that humans wouldn’t even have imagined. »
While a more imaginative organization is a clear goal for business leaders, the future will likely involve technology, like the system created at USC, which could eventually complement more creative thinking among employees. After all, when the algorithm was trained to create a speedgate, a human still had to choose the most realistic sport from the countless nonsensical options suggested by the system. Augmented imagination is therefore likely to be the next step. The term Reeves uses to describe a company that can gain such abilities is “a bionic organization.”
“What will work best is AI working alongside humans, rather than independently,” says Melissa Terras, professor of digital cultural heritage and fellow at the Alan Turing Institute. “The obvious way forward is to have the AI process large amounts of information and come up with new solutions and have a human working in tandem, refining the results. Eventually we will learn about the idiosyncrasies of these systems. and be able to play them like musical instruments, as well as choose which ideas to develop.This is the “imagining machine” – an act of co-creation, with computing power and intelligence complementing our own. »
Humans are relatively good at counterfactual thinking. It is the ability to imagine something that does not currently exist but could exist in the near future. This is what sets us apart from any current form of machine learning. Yet companies generally do not focus on this type of activity. Instead, they tend to nose dive into quarterly earnings reports, dealing with the here and now or dealing with the immediate past. The imaginative mind is rarely used to help steer companies with a firm eye on the future, but that could change.
“I’m very optimistic about the bionic organization — a more efficient enterprise that harnesses both human and machine cognition,” says Reeves. “The ‘imagination machine’ is not just silicon; it has a human element. If you’re thinking about growth right now, you’re thinking about imagination.
He continues, “Ten years from now, when someone talks about ‘organization’, they’ll be referring to a synergistic combination of algorithmic and human thinking that gets the job done better. This raises all kinds of questions. For example, what technology will be deployed? How do we use it? How to match the bandwidth of machines and humans? How do you ensure that the whole organization serves human purposes and is ethical? What do humans do and what do machines do?
The current era of AI is about using algorithms, neural networks, and deep learning to automate low-value, high-volume tasks. The scale of analysis that computers can undertake puts machines in a class of their own when it comes to this type of work. One school of thought is that applying AI to more of this work could give humans the freedom to be more creative. But not everyone agrees.
“There is concern that an AI-powered innovation sector could repetitively iterate on the world that already exists, rather than transcending it,” says Terras, who is also co-director of the R&D program in computer science. creative from the University of Edinburgh. “Why not let machines do what they’re good at – synthesize and process large amounts of information – while humans manage the complexity and nuance to create products and services that will be successful for mankind?”
We still have a long way to go before AI takes on imaginative human-led tasks and enters the mainstream of creative thinking. But, if that time comes, its potential to benefit society could mark a turning point.
There will be less talk of automation stealing people’s jobs or calls for a tax on robots when technology can create the next pandemic superdrug, for example, or an affordable way to capture and store atmospheric CO2. Then AI will be firmly entrenched in society. It will also take human ingenuity to get there.
“Will we be able to use organizational sciences, brain sciences, social sciences and digital sciences to do a better job? The answer is we already can,” Reeves says. “The focus of the imagination will be the reinvention of the business.”