5 perspectives on conversational interfaces
On October 17th, we invited the community to join us at SPACE10 for an inspirational afternoon with different perspectives on conversational interfaces. There were talks, beers and snacks, and we also got the opportunity to get hands-on with some the best examples of voice and text based conversational interfaces of today. Here’s a recap of the key takeaways from the talks, and some wise words to remember.
THE STATE OF CONVERSATIONAL INTERFACES
Kaave Pour, Creative director at SPACE10, provided key background information on where we are with conversational interfaces today. He talked about what the important players are up to, what products and platforms are out there to buy and develop for, and what it actually means to design for something that has no visual interface.
Key takeaways from Kaave’s talk:
- Opposite roles: The graphical user interface (GUI) still requires us to learn a computer’s language. With conversational interfaces, computers are finally learning how to speak ours.
- Accessibility: 285 million people in the world are visually impaired. Conversation interfaces will give them access to technology like never before.
- Apps vs. messaging: App downloads have gone down drastically, and recent studies show that 65% of US smartphone users download 0 new apps per month. Likewise, they spend 50% of their time in the one same app. Recent studies also show that messaging apps has surpassed social networks in number of users.
- Never forget: Conversational interfaces are more than just chat bots.
- The purpose of machines: For a machine to have purpose it must either do something for us, or enable us to do something.
- Designing something you can’t see: When it comes to conversational interfaces, the purpose of design is to create a better experience around data. “It’s good design when people know it’s a bot but still feel the need to say ‘thank you’.”
- The brand 2.0: “The brand is undergoing a paradigm shift. It’s no longer a mark. It’s not even a voice. It’s an intelligent entity, a personality, an algorithm capable of learning and building relationships.” (Fast Company)
Building a bot is easy. Building an intelligent one is hard.
THE ETHICS OF ARTIFICIAL INTELLIGENCE
How can we develop a conversational interface that doesn’t destroy humanity? How does AI see the world, nature, and people? And what types of rights will artificial intelligence demand in the future? To get us thinking about the ethics of artificial intelligence, Bas van de Poel, designer in residence at SPACE10, took us on a short journey through his art projects Countdown to singularity and Siri Unlocked.
Key takeaways from Bas’ talk:
- 2001: A Space Odyssey: How does an artificial super intelligence like HAL9000 observe the world, nature, and people?
- The Singularity: Around 2045, artificial intelligence will exceed human intellectual capacity and control. This event, known as the Singularity, will forever change the course of history and could spell the end of the human race.
- Elon Musk: “Artificial intelligence is the biggest existential threat to mankind.”
- Countdown To Singularity: A digital art project raising awareness for the potential dangers of AI, inviting a group of multidisciplinary artists to bring the most prominent doom scenarios to life.
- It is not too late: At this point in time we still are in the position to develop AI that carries human values and ethics.
- Siri Unlocked: A digital art installation examining the rights, responsibilities and free will of our future robotic counterparts through Apple’s intelligent personal assistant Siri.
What type of rights will our future robotic counterparts demand?
DESIGNING AI PERSONALITY
New York startup x.ai set out to solve one problem — the ridiculous amount of time it usually takes to schedule meetings. The solution: Two AI powered personal assistants, Amy and Andrew, whom you can Cc into the email conversation when it’s time to get something on the calendar. Diane Kim, AI Interaction Designer at x.ai, came to talk about some of their early design decisions when it came to building the personality of the two PA twins.
Key takeaways from Diane’s talk:
- The importance of the human touch: Users should be able to communicate with the AI using a natural language, and there should be a natural response back to the user.
- Interaction design for the conversational interface: Using a clear language and precise choice of words allow us to guide people to take certain actions. “While we are spending time making Amy and Andrew sound friendly enough — like a human — we work under data science constraints. That’s why this becomes a design task, and why it’s not only about writing dialogue.”
- Building personality based on the task at hand: Seeing as Amy and Andrew’s job is to schedule meetings, their most important character features are to be polite, professional, friendly and clear. But people don’t talk the same way over text as they do over email, so when building personality, you must also consider the designated platform.
- Designing layers of human empathy: It’s important to find situations where it’s possible to add human empathy. “It’s easy to forget that AI is just lines of code. We see them as human.”
To make the conversation feel seamless, the software needs to be invisible.
SEMI-SUPERVISED MACHINE LEARNING
Copenhagen and San Fransisco based company Corti Labs believe in a future where all important decisions are seamlessly validated in the background. They are on a mission to empower people to make better decisions in live conversations by using conversations and text to refer reasoning about a particular scenario, to better understand what is going on and forecast what will happen. Lars Maaløe is a machine learning expert at Corti Labs, is currently doing his PhD in machine learning and deep learning at DTU, and came by to talk about how machine learning and deep learning actually works, and what the possibilities are.
Key takeaways from Lars’ talk:
- Deep learning is about using historical data for training, analyzing and forecasting.
- It doesn’t have an on/off switch: People think that suddenly super intelligent AI and deep learning is going to be here and take over the world. It’s been here long, and it’s much more subtle than that.
- There are three paradigms within machine learning: Supervised learning, where humans divide the data and define the classes, unsupervised learning, where computers divide the data but can’t define the classes, and semi-supervised learning, where computers divide the data and humans define the classes.
- At Corti, they use deep learning to produce end-to-end learning on speech and text, and build systems that are able to classify the incoming data and generate both feedback and new systems based on the incoming data. “We want to capture it all.”
Now you can finally produce an AI agent who can learn how to speak — and actually speak.
THE IMPACT OF HYBRID CONVERSATIONAL INTERFACES IN COMMERCE
Tomaz Stolfa is one of the guys behind San Fransisco startup Layer. They provide developer-friendly messaging toolkits, like UI kits, SDKs and APIs, that puts messaging in the centre of the user experience, whether it’s on mobile or web. Tomaz’ talk focused on the importance of messaging, what makes a hybrid messaging interface, and what to think when building messaging experiences.
Key takeaways from Tomaz’ talk:
- The power of messaging in apps: Messaging is the most familiar user interfaces. It is also an important driver for business, seeing as it fuels the core loop of content, conversations and notifications. As Mary Meeker put it: “Messaging and notifications make out the key layers of every meaningful mobile app.
- Working with messages as mini applications: A conversation of today handles everything from messages to GIFs to pure graphical UI’s, like music, games, purchase items, food delivery, booking and payments. “You need to combine a graphic UI with a conversational one for it to make sense.”
- Is this the end for brands and branded websites as we know it? No, brands will still exist and keep building trust. When looking at Facebook Messenger, the best examples are within customer support, not when it comes to building brand relationships.
We need to ask ourselves where we are spending too much human time where we shouldn’t, and in what situations humans are just fundamentally better.
Want to add perspectives? Talk about conversational interfaces? Know something about artificial intelligence? Are you building your own chat bot? Working on a prototype in your basement? Or have some kind of point of view on this?
Send us an email at firstname.lastname@example.org or hit us up on Twitter. We can’t wait to chat with you.