Designing AI for Humans

Editor's Notes

• #4: Everything is moving so quickly and continuing to change.
• #5: Truly choose your own adventure
• #6: It will change medicine. I broke my navicular bone last year. They shoot horses who break that bone. My recovery took well over a year and I will never be the runner I used to be because of a misdiagnosed xray. AI will make that a thing of the past.
• #7: AI will change education. We talk about personalized education and helping people who are more visual, tactile or aural learners – what about teaching aids that can learn what types of learning are most effective for each of us at what point and with what subjects?
• #8: And it will change war. UPDATE: DARPA responded to a request from Fusion that “brain-neural interfaces” have not yet been implanted in soldiers, though test devices have been implanted in the brains of volunteers already undergoing brain surgery. We’ve changed the headline to reflect that implantation of chips in soldiers’ brains has not happened yet.
• #9: It’s also hard to know how we should feel about AI when big names such as these predict dire or scary results.
• #10: According to many, this is what is about to happen. Oh shit is right. We think we have seen amazing increases to the speed of computing and general technology progress over the last 20 years? Many experts say we haven’t seen anything yet.
• #11: Now what? What can we do to prepare?
• #12: I believe it is critical, as ideally with all technology - to design and create from a human centered perspective. It is often so tempting to use technology as a bright shiny solution. But we first need to understand the problem or opportunity. As Steve Jobs said, we have to start with the customer experience and work back towards the technology.
• #13: One important way to be customer centered is to focus on helping humans. For example, with AI we can help those who have sight challenges to ‘see’ and understand the world around them, including recognizing loved ones before they even speak.
• #14: And remember that we need to keep training machine learning driven AI systems. An example is this Seeing AI tool that is still trying to figure out my correct age. You can download it for free from the app store and try it yourself.
• #15: And the humans that we help should ideally be all consumers who could benefit from AI.
• #16: This means we have to think beyond some of the most talked about potential impacts of AI. Right now there is a lot of fear that one of the first huge impacts of more readily available and advanced AI will be automation that takes away jobs. But there are a number of experts who believe that jobs will change and grow for the better.
• #17: Thinking beyond automation we can come up with ways that AI can augment our human capabilities, such as this hospital robot who can deliver lunches to specific rooms so that the nurses don’t have to run around and potentially limit their ability to help their patients.
• #18: AI, whether in robot or some other form, can become one of our most important collaborators. Here at a company in California, robotic painters are working alongside human workers, sanding and painting cabinets. Despite early fears from employees, the human workers have grown to embrace the "cobots" -- collaborative robots who are helping them get the job done.
• #19: AI can amplify our human skills and intelligence, serve as a tool that makes us more successful and effective. There are AI platforms available today that connects surf rescue first responders with multiple live data views from drones, using video analysis to help responders better tell if the murky shape in the water is a shark or a piece of driftwood. They can detect someone who has got into difficulties while swimming.
• #20: In fact, there is an increasing movement to change the name of AI to EI - not artificial intelligence but extended intelligence. This is another example from the Microsoft App Seeing AI where AI can translate handwritten text to an audio version.
• #21: Now that we’ve discussed where AI might help us humans, you hopefully have a better sense of why designing human centered AI experiences is important. We’ll spend the rest of the talk about how you can do that.
• #22: I’ll share 3 principles for creating AI experiences that go beyond the buzz and provide real value for the consumers of those experiences.
• #23: The first is a big one, and directly ties to putting humans first vs. the technology. Go beyond gimmicks. AI is full of gimmicks right now. And gimmicks can be fun for a one off marketing stunt, but they generally aren’t customer centered. Take this example of the robot magic restaurant. The robots are really only there to enable selfies - they don’t yet have the sophistication to serve anyone food or to put the waiters out of a job.
• #24: To ensure that you are going beyond a gimmick and that you aren’t trying to solve a problem that doesn’t exist, start by focusing on what the value to the business or organization might be. There are lots of areas where AI can add direct bottom line value - just be sure that you map out those KPIs and then really think about what the right solution for growth will be. It may be something that is AI enabled - it may not be.
• #25: Also consider value to employees. While I’ve made a little fun of robots thus far in this presentation, these two in store robots are allowing sales associates to spend more time at the customer facing work. It’s a great example of potentially more creative and valuable - therefore higher paid - work growing because AI is helping with some of the grunt work.
• #26: Of course, adding value for end consumers is my personal passion. The beauty brand Sephora has done amazing innovative digital and AI work that both gets them extensive positive press AND serves customers. I believe this kind of customer facing AI innovation is currently under utilized and that there is a ton of opportunity for your organizations to think about creative ways you can leverage AI for consumers to drive revenue and increase satisfaction vs. just focusing on the value of AI for productivity or managing big data.
• #27: And let’s not forget the potential for AI to provide value to citizens and communities overall. Initiative like smart cities, AI enabling legal systems to work faster are amazing opportunities. This example is good for a bit of a laugh, but actually has been generally positively received. We have all experienced the trauma of finding a restroom when in need and the toilet paper is missing.
• #28: So how can we design our AI experiences to avoid gimmicks? An excellent tool to keep everyone centered and focused on the right goals is to develop experience pillars. These can be used for all of our customer experiences, or you could build off of any existing customer experience vision and strategy and create AI specific pillars. This is one general example of AI Experience Pillars that could work for a number of different types of industries and organizations - B2C or B2B. Feel free to borrow these and tailor to your own company.
• #29: And then make sure you are measuring the right things. The business KPIs on the left are from a recent Forbes survey of 300-plus executives, but don’t neglect customer goals. You will be most successful if you take the time to develop specific, measurable data points that put the customer in the center, such as whether you are providing help or service exactly where, when and how it’s needed.
• #31: Next, it is valuable to focus on services AI can provide to humans, vs. just thinking about the products and technology itself. AI is already providing services that can benefit our daily lives, and keeping a service mindset in the forefront is central to ensuring that your AI experiences stay customer centered and are truly adding value to your customers, whoever they may be.
• #32: An obvious example of AI and customer service benefit is the chatbot. When done well, they save humans time, can provide us proactive direction and can automate tasks that we may not want to handle ourselves such as paying bills, managing our savings or look up recipes.
• #33: A likely even more valuable service that chat bots can provide is easing interactions with what can be a scary and fraught with emotion conversation – health. I know I would be much more likely to tell Melody than a human doctor that I smoked a couple cigarettes last year and that I sometimes like more than a glass of wine a night.
• #34: And consider AI services that can not only help our productivity, but can help ensure we are making fair decisions without bias. HireVue and similar AI interviewing software not only helps manage the initial candidate screen, but has been shown to help increase diversity in recruiting and hiring.
• #35: Get up and act out your life, your needs. where could AI benefit your day to day?
• #38: We can’t have a set of pillars for designing AI without talking about ethics. This is obviously a critical topic that is still very much in its infancy. I do think that at minimum, if we follow our customer centered principles, that will keep us a step ahead from an ethical standpoint. I’m very pleased that such large and innovative technology companies are actually partnering to help establish ethical guidelines.
• #39: But AI technology is already starting to raise ethical concerns. With China arguably being the current leader in AI development and innovation, they are already seeing existing tools that would give many of us cause for concern. With millions of cameras and billions of lines of code, China is arguably building a high-tech authoritarian future.
• #40: And AI judges us. A number of very concerning studies like the one quoted here from Joy Buolamwini from MIT show significant bias from AI systems, particular from a gender and racial perspective. GENDER SHADES When I was a college student using A.I.-powered facial detection software for a coding project, the robot I programmed couldn’t detect my dark-skinned face. I had to borrow my white roommate’s face to finish the assignment. Later, working on another project as a graduate student at the M.I.T. Media Lab, I resorted to wearing a white mask to have my presence recognized
• #41: The thing is, AI learns from us. Including our potential biases and prejudices. We are the ones building AI systems, we are the ones teaching AI systems how to get better. Often our own unconscious biases can easily be designed into any design or strategy we create. Biases impacting AI systems is even more threatening, because of how the AI programs learn from us.
• #43: How do we solve potential bias from being built into our human instigated creations, interactions, marketing or experiences? The best starting point is to be inclusive.
• #44: AI is not a separate thing from us. In order for AI to work, we have to program it.