Privacy over efficacy Alphabet’s AI powerhouse DeepMind has formed relationships with the UK’s National Health Service to access and analyze healthcare records. On the one hand, the idea of applying artificial intelligence to healthcare promises innovative new treatments and superior preventative care. If DeepMind’s AI engines can spot hitherto unknown patterns in patient data, it may be possible to spot cancer or heart disease or other health problems months or even years earlier. On the other hand, this is a huge can of worms in terms of privacy. In fact, this is all the worms in one place at the same time. Google/Alphabet has developed a hair-trigger sensitivity around privacy issues. So they are naturally going to great lengths to ensure that patient data is private, secure, and cannot be traced to individuals. But these exact protections might render AI useless. Artificial intelligence is only useful to healthcare if it is able to spot subtle connections. If DeepMind’s AI engines yield a result like “obesity = heart disease” then the medical community will react with a collective “no duh.” But subtle connections depend on the ability to access the full breadth of a person’s records while necessarily maintaining the awareness that they belong to a single individual. And that degree of knowledge makes the individual traceable. Keep in mind that there is still no proof that AI will actually spot any of these subtle connections. DeepMind is caught up in a catch-22. To access the data, they need to guarantee patient privacy. But to achieve actionable results, they may need to violate that privacy. It’s not surprising that DeepMind is working in the UK where privacy protections for healthcare information are weaker than the United States. Why does this matter? This is a particularly Googly problem. Google is the acknowledged master of processing and analyzing huge amounts of data. They have built their search and advertising dominance on their ability to analyze, cross-reference, and recommend. Privacy advocates are naturally going to be suspicious of a company that makes billions of dollars based on projecting needs and behavior from vast amounts of consumer data. Google is the victim of their own brilliance. In a nutshell: Healthcare is a tough nut for AI to crack because privacy protections limit the free analysis of large datasets. Read More The limits of voice Ben Evans has written an interesting blog post about using voice as a user interface. In essence, he argues that voice appears to be low friction but that using voice is not necessarily more efficient and requires a change in behavior users may not be prepared to make. Machine learning has transformed voice input since it alleviates the need to laboriously transcribe language rules and exceptions. With machine learning, audio input can be transcribed to text and text into a structured query. But all of this is exclusive to input. When it comes to output, voice is more problematic. We don’t have prescribed outputs (also known as answers) for most of these queries. A certain amount of the time, Alexa or Siri give the equivalent of “I don’t understand your question” – meaning that the question is understood, but that a prepared answer isn’t available. Why does this matter? I don’t talk to tellers at my bank. I don’t talk to reservationists at restaurants. I don’t talk to travel agents. My unwillingness to talk to these people isn’t because I am antisocial (or it’s not purely that.) My unwillingness is because a GUI is a superior medium for these types of transactions. Talking to a computer to accomplish these task, with all the attendant misunderstandings, sounds like a step backwards. Evans makes the point that voice input may be confronting its own uncanny valley; the better it gets the more we will notice its failures. But I think this goes beyond just the current incomplete nature of voice user interfaces. I think voice may not be the best medium for many inputs and outputs. I didn’t stop talking to bank tellers because they couldn’t answer my questions. I stopped talking to them because I found a more efficient way. In a nutshell: Voice input works better than ever. Voice output just isn’t there yet. And voice may not be the answer to everything. Read More Virtual reality – embodied medium An interview with Jessica Brillhart, principal VR filmmaker at Google contains some of the most insightful observations about virtual reality I have ever read. Brillhart argues that VR is an embodied medium, rather than a passive one. I have argued in the past that VR makes storytelling difficult because it does not allow you to focus the audience’s attention. Brillhart argues that storytelling is largely irrelevant because VR reflects the nuance of actual lived experience. The tropes and methods of a filmmaker have all the artifice of Kabuki theater compared to the reality of how we live our lives. Brillhart points out that users of VR are often defiant of our attempts to get them to focus on a single detail. She embraces defiance in her work, guiding users of VR more by the cadence of experience in a video game, rather than demanding that they look at one thing. Why does this matter? It’s no surprise that marketers are doing a terrible job of embracing virtual reality. Marketers have a common pretension of calling themselves “storytellers” even though their “stories” are depressingly simple and prescriptive. Having made hundreds of TV commercials in my career, I can attest to the fact that most marketers believe that without the “bite and smile” framed shot, viewers would be entirely unaware that their food was supposed to taste good. The 30 second TV spot encourages cliché because there simply isn’t enough time to develop complex narratives or motivations. “Stupid husband helped by amused, but loving wife” is repeated endlessly, not because marketers believe all husbands are stupid, but because the instantly-recognized cliché alleviates the need to tell a more complex story. An embodied medium like VR depends on the subtlety and differing perceptions of real life. Most marketers have spent their careers hooked on the quick fix of cliché. They have a lifetime of bad habits to unlearn. Watch the films of Terrence Malick or Robert Altman to see how talented directors attempt to capture real life on film. In a nutshell: Don’t make a commercial using VR. Make an experience. Read More