Skip to main content

The innovation business

“What we’re trying to do is impossible,” says Paulo Dimas, VP of Product Innovation at Unbabel. The crowded room of engineers, designers and product managers chuckles.

It’s a sunny Friday afternoon, moments before the weekly All Hands meeting, and the Labs team gathers to go over their week. They pass around Japanese candy, from one of Paulo’s most recent trips to Japan. Some of the engineers inspect the red packaging, attempting to decipher the strange symbols. No one in the team is fluent in Japanese, but they have one anyway.

A young engineer continues touring the conversational app prototype, going over the many challenges they presently face, and Paulo explains that, sometime in the future, this app can be used for voice morphing, a technology that allows anyone to record sentences, and reads back translations in their own voices. What they’re trying to do is impossible, yes. But that doesn’t seem to stop them.

Paulo is wearing a brown, long-sleeved polo — discreet, like the man himself. He’s not one to draw too much attention to himself, but he’s also not shy or inscrutable. On the contrary — he speaks with the childlike enthusiasm of a beloved professor. As soon as you ask him what he and his team are working on, his eyes light up. It’s obvious he loves his job.

Everyone shows what they’re currently working on — multilingual bots, transcription and translation, an internal bot that scans for common questions and turns them into FAQs —, and Paulo constantly offers feedback, advice, and a few jokes here and there, much welcomed as the weekend approaches.

Ever since Paulo was a kid, he knew he wanted to become a scientist. He remembers watching Space: 1999, a science fiction TV show than aired in the 70s. “Everyone wanted to be John Koenig,” he goes, the iron-willed hero and commander of Moonbase Alpha, “But I wanted to be Victor Bergman, the science advisor. And he was old, well into his 60s.” Even when he was a kid, he wanted to be this old archetypical science guy.

He developed his first commercial product at the age of 14. By 16, he was researching interactive systems with INESC, an R&D institute dedicated to advanced research and development of Information Technologies, Electronics, Communications, and Energy. He went on to co-found startups, develop award-winning projects, submit patents. But he realized very young that it was frustrating, if not altogether pointless, to build things that no one used.

Striking that balance between innovation and utility would become a sort of underlying theme throughout his life — a key source of tension and motivation.

It’s become especially relevant now that he’s in charge of product innovation at Unbabel, doing what he refers to as “inventing the future.”

You have to balance your crazy ideas that are looking into the future with the business and with the need to deliver. At Unbabel we can combine both things. We have the spirit, we have the vision to do amazing stuff into the future, but also things that people will use, because we have the pressure to raise the next round.

He leads the Labs team, and they’re responsible for coming up with new and exciting applications for the Unbabel technology, and vision. That could be improving on an existing application, or coming up with something the consumer has never even dreamt of.

Labs was one of the first teams at Unbabel that was designed to be cross disciplinary. There’s AI engineers and researchers, mobile developers, designers, front and back end developers, product managers. “In a sense, we combine all the disciplines in a way that allows us to deliver any kind of experiments for future products”, says Paulo Dimas.

You might think that this motley crew of coder-dreamers works at a complete remove from the operational side of the business. But the Labs team’s job isn’t just looking into the future.

Labs deals with three different tiers of innovation. The first, the one that is more grounded in the context of the business, is what Paulo calls super-powering. That’s about looking into our existing customer service translation products, and think about ways to improve those products, to give them an edge.

The second tier is game-changing — new products and formats that can benefit from Unbabel’s technology. Multilingual live chat and video transcription and translation, for example, now part of our solutions, were originally born out of Labs’ projects. When a project becomes viable, when it stops to be an experiment with potential and starts to become a full fledged product, with clear business potential, Paulo starts incubating a team. He explains:

One of the things we learned is the importance of incubating a team inside labs. When we decided our team has all the disciplines, when we have that product launch on the roadmap, when we feel like this is going to be really big, then we start incubating a team inside Labs, a team that will afterwards become autonomous.

The video team, originally incubated inside Labs, has become bigger than Labs itself. They must be doing something right.

High risk, high reward

Only then, in the third tier, come the “moonshots”, or what Paulo refers to as universe-exciting. The things that grab people’s attention, excite investors, headline events.

“It’s what we call a kind of 10X goal. Something that it’s not yet possible, but we believe that we will learn a lot using this kind of mindset,” Paulo says, referring to a popular notion inside the Valley that claims that only by setting goals that are 10 times bigger than what you think are reasonable, and by taking actions that are 10 times greater than what you believe are necessary, can you succeed.

This framework has gained massive appeal lately, but it’s hardly a new way of thinking. Almost 60 years ago, in 1962, President John F. Kennedy gave a rousing speech at Rice University, during the peak of the space race between the USA and the USSR. “But why, some say, the moon? Why choose this as our goal? (…) Why climb the highest mountain? Why, 35 years ago, fly the Atlantic? (. . .) We choose to go to the moon in this decade and do the other things, not because they are easy, but because they are hard.”

At the time, no one knew how to achieve that. NASA had barely shot an astronaut into space with a military rocket, much like a human cannonball, let alone devised a plan to go to the moon and back. When you’re working towards an elusive goal, you can’t lean on the tools, technology, and even assumptions you currently have. You’re not going to get there through sheer force of will. You’re forced to challenge your assumptions, and come up with creative solutions. Seven years later, in 1969, although JFK never lived to see it, Buzz Aldrin climbed out of Apollo’s lunar module and joined Neil Armstrong on the surface of the moon.

Labs goes through this creative, assumption-challenging process on a daily basis. Paulo recalls a time when Vasco Pedro, our CEO, and a PhD in Language Technologies from Carnegie Mellon, teased him about their voice morphing goal: “You would need 20 PhDs to pull this off.” Labs doesn’t have 20 PhDs, but they’re combining multiple pieces of the puzzle with open-sourced work of other researchers. “You just need help with the building blocks,” Paulo says.

But when dealing in uncharted territory, it’s hard to have frameworks that measure progress, or success. Paulo and his team are strong believers in OKRs, or objectives and key results. OKRs are a tool pioneered by Google, and have become very popular within the startup world. By coming up with your own goals for what you want to achieve that quarter, and aligning them with your team members’ and the companies’, they can create alignment and measure progress around quantitative, actionable goals. In the Labs’ case, that would be, for example, the number of prototypes created, or number of potential customers engaged for a prototype.

From B2B to brain-to-brain

Few ideas are as “universe-exciting” as the idea of transmitting thoughts from one brain to another. More than an exciting idea, it’s become a sort of an inside joke at Unbabel. It’s hard not to think of Paulo when the subject of brain-to-brain comes up — there have been too many presentations, chats at Friday’s All Hands, and company retreat skits not to immediately make the connection.

“I believe that the future of language is going to be directly connected with the brain”, Paulo says, a vision both himself and Vasco share. Tim Urban, Wait but Why’s quirky, long-form writer and stick-man illustrator, agrees. In his 40,000 word essay on Neuralink’s mission, the evolution of knowledge, language, and brain machine interfaces, he points out that when we communicate, we’re effectively using 50,000-year-old technology. The same species that replaces their smartphones for a new, shinier model on average every two years.

Language is neither a fast nor lossless form of communication. As we compress our thoughts into speech, we lose a lot of information. We lose context, intention, and a lot of useful metadata that would help the receiver understand us better. The receiver then incorporates that lossy data into its own set of preconceptions and experiences, where the message can gain a completely different meaning. Frequently, the loss of context is irrecuperable.

But that’s not the only problem. For Paulo, speed is another issue: “If we think about it, the way that human beings communicate is very limited by our muscles. We can only speak 150 words per minute. If we go directly into the brain we can be much faster.”

And we may be going there soon. Paulo believes we’ll be able to communicate using our brains in this very century. The technology isn’t quite there yet, but innovation doesn’t happen in a linear way. It happens gradually, then suddenly. “Innovation always comes in an S-shaped wave”, he gestures with his index finger. It starts slow — the technology is still in its inception, it lives mostly out of university labs and R&D departments. Then, it starts growing, developing. Maybe it’s the increasing computing power. Maybe the price of its components just dropped. Maybe there’s a new approach to the problem, or someone invents a new technology that allows this one to flourish. Then, suddenly, boom. It’s no longer a crazy idea in a research lab. It’s something adopted, and promptly taken for granted, by a big part of the population.

“We are staring at that moment when the S curve is starting to grow exponentially,” he says. The trigger is about to happen, and there are three things conspiring to make it happen.

The first one is improvements in machine learning.

Each day, electrical impulses run through the 60,000 miles that make up our own nervous system, controlling everything we do — the cold buttons we push each morning as the coffee maker warms to greet us, a slight adjustment to the black, boxy specs sliding down your nose, the nervous laughter as you introduce yourself to an auditorium full of researchers. We’ve known for a long time that these commands come from our brains, and yet, they are extremely hard to decode.

Neuroscientists have been making painstaking efforts and great strides to improve the resolution of these signals through non-invasive methods. No method is ideal, but one that is fairly simple, and readily available, is an electroencephalography, or EEG. EEGs use a small number of electrodes, usually between 14 and 50, placed along the scalp — typically in a very characteristic cap which bear a striking resemblance to an 80s swimming cap — to register brain activity. Erratic waves, at least for the unsuspecting eye, are then displayed across the screen, representing our brain patterns and various awareness states, and can detect a myriad of conditions, from anxiety disorders and depression to epilepsy or diabetes.

Its resolution is not the best, given that EEGs can only detect the sum of the charges from millions and millions of neurons. But with recent advances in deep learning and pattern recognition, we are now able to better decode the signals of our neural networks using artificial ones.

The second reason is the increasing number of simultaneously scanned neurons. Electrodes were typically made by hand. But with advances in semiconductor technologies, they were able to move from single to multielectrode arrays, allowing plenty more neurons to be scanned on a single use.

As Neurosurgeon Ben Rapoport explains Tim Urban, “the move from hand manufacturing to Utah Array electrodes was the first hint that BMIs were entering a realm where Moore’s Law could become relevant.” Over the last 50 years, we’ve seen the number double approximately every 7 years. In the 80s, we could barely scan a few dozen. Today, as Paulo learned in a recent conference at NYU, the number is closer to a thousand.

“By scanning more parts of the brain in real time with resolution, we’re going to generate much more data that machine learning can decode and make sense of what we are thinking about or thinking in terms of imagined speech. That’s where we believe things are going to move,” Paulo adds.

The last factor is the development of non invasive techniques. Up until this point, techniques that would be able to address high, neuron level resolution, typically involved neurosurgery, and brain implants, something, Paulo jokes, not everybody is willing to do. New techniques are being developed that are trying different approaches to scan the brain in noninvasive ways, and Paulo is quick to enumerate some of these: “Some are using optics, near infrared light. Others are experimenting with a kind of a mesh network of nanoparticles that are injected within the brain and then create a network that could interface with individual neurons.”

Neuralink, as it turns out, is using flexible “threads”. A couple weeks ago, on July 16th, Elon Musk revealed its technology to the public for the very first time. These threads are thinner than human hair, and can contain “as many as 3,072 electrodes per array distributed across 96 threads.” They’re a lot safer, too. Musk hopes to implant Neuralink’s system in a human patient by the end of next year.

But it’s not just companies or research labs who are in on it. DARPA, the US government agency responsible for developing these new emerging technologies, is chasing brain-to-brain too. One of their challenges is to scan one million neurons by 2020. Considering we’re halfway through 2019 and we’re not even close to one million, it doesn’t seem like that’s going to happen. Brain-to-brain is a moonshot. And by any way you put it, it’s almost completely detached from any kind of commercial application — at least for now. But, like DARPA, Labs also likes to have these stretch goals.

Gazing at the night sky

Inventing the future is all part of the labs experience. It’s a story as old as science itself. Throughout mankind’s history, many men have stopped and gazed at the night sky, wondering about the unknown. One of these men was Galileo.

400 years ago, in 1609, Galilei Galileo, a rising star in the Italian scientific community, heard rumours about a peculiar instrument which had just been invented by Dutch spectacle makers, one that could show distant objects as if they were close by. Immediately, he started working on his own version, experimenting with different lenses and distances, figuring out how to increase the magnifying power, and producing increasingly powerful telescopes. A few months later, he presented his telescope to the Venetian Senate, to be used as a spyglass able to detect enemy vessels before they could ever see theirs.

But then, Galileo turned his telescope from the earth to the sky. That year, he drew the Moon, its characteristic rugged surface, and all its phases. A few days later, he discovered four moons revolving around Jupiter. He observed that Venus went through phases, much like the moon. He saw Saturn’s ring. He found stars no one had ever seen before, stars invisible to the naked eye. There was a vast world just waiting to be discovered.

Crediting any one individual for his achievements in the history of innovation is a fool’s errand. Innovation happens a lot more often from a collective effort of inventors building on each other’s work through generations, than by a singular moment of genius. This is true for the Scientific Revolution well underway during Galileo’s time, and it’s been true ever since, throughout the intellectual development of the Enlightenment, the mechanical fever of the Industrial Revolution, and the digital revolution in the late 20th century.

Galileo built on the work of Copernicus, René Descartes and Francis Bacon, who built on the work of Roger Bacon and Robert Grosseteste, who built on the work of Ibn al-Haytham, Al-Biruni, and Ibn Sina, who built on the work of Aristotle, Archimedes, and Epicurus. Galileo didn’t invent the scientific method. He inherited it.

He was, however, a perfect ambassador for it, and his empirical approach was the reason he was chosen to embody one of Labs’ principles. Rather than cherry picking evidence that would confirm a certain belief, he was open to follow whatever conclusions his experiments would lead him to. “It’s about not bringing our biases to anything, working with data, with facts, with reality,” says Paulo.

R&D with the ancients

Experimentation is one of Lab’s principles. They chose principles to represent the three values of his team, values that would carry them through many hiring processes, on-boardings, and brainstorming sessions and design sprints.

But the first principle they chose, lo and behold, was the one of “first principles,” traced back to Aristotle. All his intellectual contributions stemmed from self-evident truths — the foundation of all knowledge.

Aristotle believed we should start with our beliefs, and work backwards until we find the underlying truths with which those beliefs are built. First principles help us break down complex ideas into smaller parts, and again into smaller parts (it’s turtles all the way down), until you get to the core building blocks. Doing so can open new perspectives and solutions to otherwise very complex problems.

But real innovation takes more than defining the foundations of our knowledge, and building on them through experiments, gathering data, following the experiment wherever it goes. It takes a perfect blend of fondness for risk, or perhaps aversion to monotony, a certain animosity towards the status quo, and maybe even a much needed speck of self-delusion, to believe that you can succeed where others, and possibly even yourself, have failed.

One of the principles that the history of innovation has teached us is that innovation always comes, and this is a quote from Eric Schmidt, by the way, the former CEO of Google, innovation — and this is the executive saying this — always comes from small teams having new ideas that managers and executives don’t understand.

Paulo and his team believe all innovators challenge assumptions and preconceptions of the age. They must have a certain rebel spirit. And so they chose it for their final principle. To represent it, they picked Gandhi, a choice that isn’t as unanimous as the principle itself.

Paulo shows me a specific picture of Gandhi. In it, he is walking on this ordinary-looking street, surrounded by men in their suits and ties and elegant shoes, while Gandhi is only wearing a simple white cloth and some dirty sandals. It seems to be raining — the ground is glossy, and those men in their suits and ties and elegant shoes are sporting umbrellas. However, if he was slightly annoyed with the lack of appropriate outerwear, the photo doesn’t show. He looks calm, collected. Like he doesn’t need any umbrella.

To Paulo, that particular picture illustrates the spirit of innovation. The ideas that managers and executives don’t understand. Gandhi is different things to different people: an idealistic pacifist, a revolutionary martyr, a political inciter, and even an apologist for India’s caste system. He was a complex human being, as most of us are, and even the scholars and experts may never come to any conclusion as to which of those things he is, or if one individual really needs to be any one thing at all.

Discovery tour

As we continue the tour of the demo in that Friday meeting, an account setup process pops up. Paulo interrupts, sweeping his arms across the air. “No, no accounts the first time! Just let them start using the app, they can create an account later.”

User experience is something the Labs team keeps in mind from a very early stage. The way their ideation process works is through customer discovery programs, processes that start by the discovery stage of the product. They select four or five potential customers, or reference customers, as they call them, with which they start working to test their ideas and prototypes. Sometimes those customers come from suggestion of the sales team, sometimes they reach them themselves, but the idea is always to learn more about their needs and challenges, so they can figure out how to meet them.

“By engaging the customer in these discovery programs, in three months to six months, we can deliver a solution that will be targeted towards their problems.”, Paulo says. After that period of time, they interview the reference customers again, to see whether the prototype really matches their needs. The idea behind it is to test market viability as soon as possible, and then, to iterate it, fast.

During this process, Labs always addresses the four main vectors of product ideation:

Paulo explains that they see these four vectors as the risks of the product:

On a customer discovery program, we try to de-risk each of these vectors. If this is something that the market will want, the customer will use, that is feasible to build and viable for Unbabel as a business, then we move it into the delivery stage.

And ideally, this will be done very fast. Labs often uses design sprints, a methodology that was created by Google Ventures a few years ago, based on principles of human-centered design, behavior science, and business strategy, all neatly packaged so you can compress months of endless debates and back-and-forths into just one week, going through all the steps of ideation, prototyping, decision, and user testing.

Instead of a regular product cycle, where after months of tinkering, you launch a minimal viable product and test it in the market, you have a shorter one, where you’re moving forward with customer insights and clear data from the prototype iterations. Design sprints also create a lot of creative tension in the team, which ends up generating plenty ideas.

Design sprints even come with their own DIY guide for running your own.

On Monday, you’ll map out the problem and pick an important place to focus. On Tuesday, you’ll sketch competing solutions on paper. On Wednesday, you’ll make difficult decisions and turn your ideas into a testable hypothesis. On Thursday, you’ll hammer out a high-fidelity prototype. And on Friday, you’ll test it with real live humans.

Google Ventures

All of these processes help keep Labs grounded, aligning their work with the business’, and the market’s needs. It brings a tangible, concrete value to this notion of innovation. It allows them to look up and dream about the impossible, while still keeping a foot on the ground.

The siren call of innovation

But it’s not easy to introduce what they’re working on to the rest of Unbabel — to the engineering teams, product marketing, sales. “I would say moving innovation to the rest of the company is always a challenge in any company,” he explains. It’s crucial to have processes in place to move prototypes into production, to communicate the work they’re doing, and what it means for Unbabel, and the other teams. Which, he admits, sometimes lags behind.

To address this lack of visibility, they’re building a Labs page, where everyone across the company can see what they’re currently working on, but also past, and future, projects.

No matter how fast things move, there’s always a lingering feeling that they could be experimenting more, and faster. It’s the weight of the Innovator’s Dilemma, a seminal book chronicling why market leaders fail when confronted with a new technology paradigm.

Companies naturally want to allocate their resources on tangible customers’ needs, on increasing profits, on features or integrations that can quickly return their investment. But when new technologies emerge, they don’t have a carefully studied market fit. The markets aren’t big enough, the ROI isn’t big enough. These new technologies can’t be evaluated in the same terms as technologies that have already matured.

Constantly reinventing your products, even competing with yourself, is a fundamental mindset in the tech industry, which is why Paulo always looks back at Facebook’s culture book: “If someone is going to kill Facebook, it’s going to be Facebook.”

That’s why Labs has three different tiers of innovation, and that’s why the stretch goals, like brain-to-brain, are important. Although they are very distant from the daily lives and responsibilities of all other employees, and can sometimes be seen as the last item on the operational list, they are part of what gives Unbabel the edge.

Some ideas will fail, sure. And surely, some have. But that doesn’t concern Paulo too much.

“One of the things that we talk about in Labs, is that 50% of the ideas should fail. If they don’t fail, we’re not taking enough risks.”

The post The innovation business appeared first on Unbabel.