Wednesday, February 7, 2018

Digital Assistants or AI Everywhere

One of the major topics of conversation coming out of this year's CES was that digital assistants seemed to be everywhere. Digital assistants are an example of artificial intelligence (AI) and are found in products like Amazon's Alexa, Google's Google Assistant, Apple's Siri and Microsoft's Cortana. The products most associated with these AI are Amazon Echo, Google Home, and just recently available for sale, Apple HomePod. They range in price from slightly under $50 for Amazon Echo Dot and Google Home Mini to $349 for Apple HomePod. There are also third party manufacturers creating devices for use with these AI, e.g., Harman Kardon makes a speaker for Cortana. Sonos makes a speaker that works with Alexa and promises to integrate eventually with Google Assistant.

Google Assistant, Siri and Cortana have all been around for several years and accessible via smartphone, tablet and/or computer. So, we've had a little while to get used to the pros and cons of interacting with voice activated digital assistants. But their move into the home started with Amazon's Echo, essentially a smart speaker, a couple of years ago.

Smart speakers coordinate with music streaming services so you can ask them to play a wide variety of music. The AI can look up information on the web so you can ask for weather or news updates. You can ask also ask for event information like movies and times near you. You can even ask for calculations and conversions, e.g., teaspoons to ounces for cooking, or Fahrenheit to Celsius. Smart speakers are often tied in with the Internet of Things and can work with various gadgets and appliances. As they tend to take answers off the top without scrutiny, they may not always be the best sources for reliable unbiased information. Alexa tends to over rely on Wikipedia, for example. So when people change the Wikipedia entries, answers can change. A recent article in Fortune: 'Who Is Jesus?' Google Home Couldn't Answer and People Weren't Happy  helps illustrate some of the potential problems in AIs remaining unbiased in a polarized information world. Amazon Echo, in particular, also has a large variety of third party skills available so it can be used for an ever growing number of games and various tasks.

Digital assistants are offering a lot of possibilities to people with visual and physical disabilities. They may prove particularly useful to our aging population. They offer quick information in response to verbal requests and can be asked repeatedly for the same information without ever showing signs of tiring or exasperation. As they can also be tied in to IoT hubs and devices, they can be used for routine tasks like turning on and off lights, locking doors, changing channels on TV, setting thermostats, as well as monitoring cameras and doors for security. People can use them to order groceries or other supplies as well as provide reminders of appointments or times to take medications. They can be linked to family members' devices for quick calling features. In short, they might make it easier for people to stay in their homes and for families to feel comfortable about it. There are many articles such as this one written on the topic: Amazon Echo for Dementia: Technology for Seniors

One of the downsides is privacy concerns over always on listening devices in the home. There are indications when they're actively listening. The Echo lights up blue as in the photo.
However, one does notice that they seem to turn on periodically for no apparent reason. This is particularly true if they're near a radio or TV. People seem to say the A-word setting off the Amazon Echo at an alarming frequency. There are other optional wake words for Amazon's product. You can change it in settings to: Amazon, Echo or Computer. But I doubt any of these will solve the problem especially if you watch or listen to tech podcasts like I do. It will still turn on at odd times. The phrase, "Hey Google" is probably a lot less likely to be uttered by accident. In the case of the Echo, you can go into the settings and see when it was activated and what were the triggers and responses. You can also delete any or all entries from your history. But many are uncomfortable that information is being collected and stored. And law enforcement has at least once that we know of asked for it. So, if you have privacy concerns, you might want to think twice about bringing listening devices into your home. But do keep in mind that most phones, particularly newer Android phones are also always on listening devices. They're just not quite so open about it.

There are some open source solutions.
Mycroft Mark 1 pictured here. As it's open source, it's designed to be modified and customized by users. But it will also provide basic AI functions like answering questions, controlling IoT devices, playing music, etc. From reading some of the information on their website, I think that the key difference they offer is that your voice information is aggregated so that it cannot be tracked back directly to you. This is different from Amazon which connects your queries to your Amazon account as one of the services offered is shopping via your Echo device. Google also offers shopping capabilities. As their main business is advertising, so you can assume that Google Assistant information goes into their aggregate database about you and your interests, all the better to direct ads at you. Apple offers a more privacy oriented service for those who are willing to shell out premium dollars. 

The allure of open source is its customization. For those who are so inclined, one can use the Mycroft software on a raspberry pi and enable a number of different devices and uses. The downside is that it might be a bit trickier for non-techies to set up and troubleshoot and it may not work with certain services. Digital assistants seem to be following the services silo model. Amazon products work best with Amazon services: Echo with Amazon Prime music and video, Fire TV, Google Home with Google Play and YouTube, Apple HomePod with Apple Music and iTunes. They all purport to work with major third party services like Spotify. But if you're tied to a particular service, it would be a good idea to make sure that it is supported by the device of your choice.

So, what does all of this mean for libraries? I think it will continue to be important for librarians to help educate the public about new products and how they can be used or possibly misused. It also seems very important to make people aware of the privacy implications of many of these devices and services. People should understand there's a trade off between convenience and giving up personal information and data so that they can make better informed decisions. 

But there are wider implications as well. What about the increasing reliance on voice control? It is convenient but it does limit the kind of information that is routinely accessible. There is no and evaluation of sources and perhaps not even a citation. But I have heard Alexa say a piece of information came from Wikipedia. There's also very little opportunity to drill deeper or question further. Part of the assistance offered might be to help people set up their own custom news briefings so that they can get news of interest to them from reliable sources.

Once again, librarians also have to ask how they can get their services on some of these devices. One of the features people love is the availability of audiobooks. Unfortunately, as far as I know, this only applies to purchased audiobooks. Wouldn't it be great if the libraries' offerings were available as well? And how about ordering library materials via digital assistants? Shouldn't our ILS be searchable as well? It goes back to the notion of meeting our patrons where they are. If they are spending more and more time with digital assistants, shouldn't we be there as well?