October 14, 2016

Design And UX For An AI-First World

Table of Contents:

Remember when people were writing about designing mobile-first products? Now, at the frontier of one of the possible next platforms for our digital interactions, we discuss how we at Clarifai think about designing AI-first products.

There’s an age-old adage in UX that says a machine should never ask a question that it can work out on its own. Alan Cooper and his partners, authors of About Face, widely considered the Bible of interaction design, described this as a machine that is like a considerate person — one that takes an interest in us, is forthcoming, anticipates needs, is conscientious, and is perceptive, among many other attributes.

“A machine should never ask a question that it can work out on its own.”

 

brain

 

Today, as artificial intelligence develops at an accelerating clip, these attributes of a considerate person are increasingly attainable through machine learning technologies, and it is no longer as critical for elements of UX to carefully create the illusion of such a person. Instead, a new set of questions arise:

  1. What should the relationship be between human and machine?
  2. How do we make AI accessible to and inclusive of everybody?
  3. Given that current AI technologies rely on large amounts of data to learn, how should AI-first products treat data and privacy?

At Clarifai, we believe AI is the next platform for our digital interactions. We see an opportunity to revisit many user needs with an AI-first approach, much like how many web needs – such as communications, media (photos, videos), and even business directories – were revisited with a mobile-first mindset. Beyond this, we are also confident that a new set of possibilities will emerge, much like how geolocation-driven products emerged with the mobile platform.

“A new set of possibilities will emerge with AI, much like how geolocation-driven products emerged with the mobile platform.”
ui

 

Similar to the transition to mobile, there will be attempts to “cross the chasm in two small steps,” so to speak. That is, some will attempt to sprinkle AI into their products as many sprinkled mobile into their products, creating AI-enabled systems like people had created mobile versions of their websites and called them mobile apps. There are probably many instances where this sprinkling is good enough, as was the case with mobile (mobile sites are good enough for restaurants), but here at Clarifai, we’re interested in the set of experiences that will be transformed with an AI-first approach.

To begin, one of the fundamental questions of thinking about AI-first design is defining the relationship between the machine and the person. There have already been a few stances taken:

The Oracle

The Oracle will know what a person needs before s/he knows it himself. In this relationship, the machine is superior to the person as it is infinitely more intelligent. This is what Google wants to build.

The Henchman

The Henchman exists to serve the person’s whims. If the person wants to leave 15 minutes late for a meeting, it isn’t going to chastise because it assumes the person knows best. In this relationship, the machine is subservient to the person, much like the Genie to Aladdin; its intelligence and power is solely used to fulfill its master’s wishes. This is what Apple wants to build.

The Partner

We see a third path here, one where machine and person work side by side. The machine’s goal is never to replace the person but rather to augment him or her, allowing the two to work symbiotically in concert.

But what does this mean? For the foreseeable future, there are many things people are better at than machines, such as empathy, metacognition, and efficiently sensing and encoding new information. Machines, on the other hand, are very good at other tasks, such as organizing information, interpreting data, and deriving insights from situations too complex for people to fully comprehend. We believe this relationship will continue to evolve, but there will always be something both parties bring to the table.

“At Clarifai, we believe the machine’s goal is never to replace the person but rather to augment him or her, allowing the two to work symbiotically in concert.”

 

ar

 

With this belief in mind, the design team at Clarifai works hand in glove with the engineering team. There isn’t a bias towards a user-centered approach or a tech-centered approach. Instead, we aim to meet in the middle. One side of the room considers our tech, what it’s capable of, and asks, “what is possible now that this exists in the world?” The other side of the room looks at what people want and asks, “how can AI best help serve these needs?”

We leave you, then, with three of the top challenges on our minds as we think about designing for an AI-first world.

Building human-friendly AI

New technology is always scary, especially new technology at the magnitude of intelligent machines. Making our products relatable and approachable is always top of mind for our design team. This is particularly relevant for our iOS photo app, Forevery.

Making teaching AI accessible to and inclusive of everyone

Right now, people code in order to tell machines what to do. We believe as AI advances, people will instead teach machines what to do, and it is important for the teaching to come from a diversity of viewpoints. It is our design team’s goal to make this teaching experience enjoyable, inclusive, and human. You can already experiment with teaching AI in an easy and accessible way via our Custom Training UI.

Protecting data and privacy

In the current paradigm of AI, machines need a lot of data to learn. As a part of teaching machines, we think about how best to design our products to get the data to make machines smarter. That said, we are also deeply aware that the acquisition of data needs to be done respectfully and therefore constantly challenge each other to think creatively about what privacy looks like in an AI-first world.

We’re excited to tackle these questions and more as they arise and look forward to sharing our learnings with you! In the meantime, we’d love to hear what your thoughts are on AI-first design!

 Artificial Intelligence Glossary, 2021 Edition