David Friml on the Future of AI
David Friml is a young techy entrepreneur. He co-created a startup specializing in artificial intelligence (AI) with a bent towards automating tasks. I initially met with him to talk about the growth of machines and its effect on our jobs, but the encounter quickly shifted towards a more philosophical, even metaphysical conversation about the evolution of cosmos and new “frontiers of consciousness.”
Watch the Interview
The Original Idea: “Remove Shitty Jobs”
In 2013, anthropologist David Graeber raised a warning flag about the rise of what he called “bullshit jobs.” In a nutshell, many office jobs with complex and obscure titles consist in doing menial, if immaterial, tasks. Those who do them tend to see them as useless, yet still do them because they need to be paid. Well, according to another David—my host of the day—, good news lie ahead: these jobs can be automated.
Think of immaterial routine jobs, such as recognizing the category where a particular product should fit in, photoshopping pictures, or solving CAPTCHA codes. These off-putting tasks are performed by humans because machines tend to be bad at them. If they were automated, people would not need to do them anymore—and they could use their newly free time to do more personal, properly human things.
Since the beginning of the industrial revolution, many repetitive, mechanical tasks have emerged. With them came the wish to automate them so that actual humans would not have to bear with their drudgery. An early figurehead of Taylorism, the famous economist Adam Smith, mentioned both the division of labor and will to automate the simplest tasks:
Everybody must be sensible how much labour is facilitated and abridged by the application of proper machinery… In the first fire-engines, a boy was constantly employed to open and shut alternately the communication between the boiler and the cylinder, according as the piston either ascended or descended. One of those boys, who loved to play with his companions, observed that, by tying a string from the handle of the valve which opened this communication, to another part of the machine, the valve would open and shut without his assistance, and leave him at liberty to divert himself with his play-fellows. One of the greatest improvements that has been made upon this machine, since it was first invented, was in this manner the discovery of a boy who wanted to save his own labor. (Adam Smith, The Wealth of Nations, I.1.8)
At the same time, though, appeared a movement called Luddite. Workers who were afraid of losing their jobs destroyed machinery that they believed was designed to replace them. Automation makes the employees’ usefulness problematic: if your job is outsourced to a machine or algorithm, what will you do?
Automating Visual Recognition
On this point, David Friml maintains a resolutely optimistic stance. Before the automation of land working, many had to handle ploughshares and sow seeds. Then, huge, but rather stupid machines—that do not need AI—performed most of the job. Harvesters can perform what fifty men did in a lesser time. At the time, farmers reacted just as Luddite machine-breakers. “What will we do? We only know working the fields!”
Eventually, they reconverted into higher-level jobs. Not having to toil the land from dawn to dusk anymore, they could switch to more intellectual, rewarding pursuits, which also tend to be safer. As cities grew, so did journals, plays, books, movies—culture. Now everyone consumes and watches with pleasure what is created by thousands of men and women whose ancestors worked harshly.
Yesterday, physical low-level jobs were automated. Today, immaterial yet still low-level tasks can be delegated to machines, too. David Friml’s startup specializes now in automating visual tasks. In e-commerce, someone needs to see pictures—sometimes by the thousands—of products to label them and put them on sale under the right category. I say someone, not some algorithm, because AIs tend to perform poorly at visual recognition and categorization. However, Friml’s startup is changing that. IT companies can already implement an automated visual recognition tool he and his friends have been designing lately. A well-crafted algorithm can do one of these “bullshit jobs” we do not value anymore.
Instead of these, David Friml guesses, “other [jobs or occupations] will come into play.” These jobs will be related to what machines are bad at: creativity. As people earn free time thanks to machines, they can embrace their inner, most human side. Overall, the less needs we have, the greater the choices. “Many currently existing jobs can be automated, not only in IT but also, say, in logistics or factories.”
Five years ago, I remember that visual recognition was touted as a prime example of something algorithms can not do—or rather poorly. CAPTCHA codes are based on the AI’s inability to read even a barely distorted or crossed-out text. Now, machines can do that, too.
How AI and its Algorithms Became Better and Better
Machines’ “intellectual” abilities greatly improved over the years. According to Friml this is due to the interplay of three factors:
- More and more computing power. As Gordon Moore had observed as early as 1965, the number of transistors integrated into dense electronic circuits nearly doubles every two years. Up to this day the so-called Moore’s law has hold true. Strategy games in the 2000s featured a playable but somehow limited, linear-processing AI, whereas more recent games are much better at being unpredictable. The dynamic behind works as well when it comes to automating tasks that were complex to machines and simple to humans.
- The rise of the so-called Big Data. Machines need a lot of contents to process so they can learn from them. When many sources can be gathered, this is hardly a problem. A computer can be equipped with a lot of camera sensors, get literally millions of images and treat them according to its in-built algorithms.
- Speaking of algorithms, they are, David Friml tells, the third cause: perfected and refined each year, these are now structured as convolutional neural networks that imitate human brains. Thanks to this new algorithmic architecture, cutting-edge programs can now recognize images better than humans.
“This is a great challenge,” David Friml adds. Beyond the particular visual recognition set of modules his startup works on, vocal recognition, decision-making and other modules have to be put together as to build a program or machine with a semblance of autonomy. A drag, though, remains in that “current computing architecture is based on 50 years-old processors,” unfit to current computing power. There would be a clear need for a new type of hardware, rethought as to unleash the potential of current computing power. What would be this hardware or IT architecture of tomorrow? Perhaps, Friml says, artificial brains—as convolutional neural networks already do. Or quantum computers.
Merging with Computers Will Happen—and It Will Be OK
So far I had planned to ask my guest about the controversial, yet unavoidable issue of automation. Many individuals resent machines taking what they perceive as their jobs. But before I could steer our conversation towards the ethical aspects of the issue, Friml opens a whole new can of worms: humans fusing with machines.
When I ask him if the human ability to create and think autonomously, without being pre-programmed before action, does not mean an irreducibility of what is at the core of our humanity, he answers that machines are already doing better than humans at what they could barely do five years ago—like visual recognition—and that the trend is absolutely likely to continue. Both hardware and software will keep improving in ways we cannot confine beforehand.
“Machines are as our children”, he says. “But they will rise at a higher level of complexity and may one day start evolving on their own.” Perhaps, one day, robots fused with humans—or not—will “eat the galaxy.” Humanity as we know it today will become part of an upward-climbing history. “We’ve got our show, now, evolution does what it does.” After all, don’t we already have symbiotic relationships with machines when we literally spend hours every day bent on computers or smartphones, or when we accept having microchips grafted under skin?
Merging with machines may give us the best of both worlds. We could get the processing power of computers, thus gaining the ability to calculate enormous sums instantaneously, or upload our minds and become virtually immortal, while retaining the peculiar human consciousness and creativity. On the other hand the very perspective is a bit frightening, and even among futurists, the idea of fusing with circuit boards is far from consensual. Aren’t we at risk of losing our humanity and objectify ourselves? Are not trans-humans monsters?
According to Friml, we should identify with consciousness and cosmos, not humanity per se. “Everything is always changing. We’re moving towards more complexity and a higher good.” The natural process of evolution may go well beyond what we usually call nature: “current humanity is a stage, and AI is the new frontier of consciousness. Civilizations emerged out of a nature where they did not previously exist, they came out of nothing. AI and merging with machines is the same process” at a later, more evolved stage.
If nature means cosmos and evolution, as Friml seems to believe, then the development of AI is the vanguard of evolution itself. “Everyone is already on the journey. Everyone will be happy once it comes… A century ago an iPhone would have created fear, now people embrace it.” The arrow of progress—natural or technical, those being now the same thing—is fixed. “We may try to divert it” but in the long run, this would be useless. The better thing to do would be trusting the process, going for the leap of faith, and embrace it.
(at 2:40: “do you want to take a leap of faith?”)
The only part of the process where we’d have room for flexibility would bear on how exactly this happens: evolution tends to happen randomly, chaotically, but it can also be taken charge of and steered. Either we would let it happen spontaneously, not to say messily, or we would have a kind of mission control or—immanent—intelligent design.
“The whole of humanity must raise their levels of consciousness.” The major problem would not be technological, or even ethical, but behavioral: “humans are not accommodated for the rapid pace of current change. Our world is changing very quickly.” Individuals may end up harming other people because programs or machines started messing with their habits and ideas. “Singularity is coming at a personal level.” Some are being reluctantly dragged, others enthusiastically join. Through technology, the natural process of evolution is stepping in. “We’ve got smartphones, AI, and so on, but also an increased interest in meditation, mindfulness.” Better prepare for what is already coming then.
We should remember, though, that the more computer power does menial or boring task, the less we’ll have to work for a living and the more time we’ll have to do what we really want to. Can Singularity cure humanity from age-old woes such as greed, material strive, envy and the like?
3 Tips to be Future-Ready
After this dizzying and thought-provoking tirade, I don’t feel very much like engaging a controversy. (If you want another perspective on these same questions, I recommend that you find out about Gerd Leonhard’s work, a futurist colleague whose motto is “embrace technology, do not become it.”) So I prefer asking my traditional concluding question. If you could give three tips or pieces of advice to be future-proof, future-ready, what would it be?
David Friml’s tips are:
- Educate yourself. Read a lot, from diverse sources, about as far-ranging themes as “nature, philosophy, religions, technology, natural sciences…” to grasp the world a little more.
- Be open to new experiences.
- The most important perhaps, embrace the change. Like the process. Trust it. Enjoy the show, take a “cosmic” perspective, avoid getting limited by contingent divides or tastes or controversies. “Today is a wonderful gift!” Whether we want to fuse with AI-powered machines or not, we can feel some gratitude towards the progress which is already behind us, and remember there is more to come.
To know more about David Friml’s work on artificial intelligence and automation of tasks, check here.