Wednesday, December 7, 2016

Artificial Intelligence: Helping Lawyers More Than You Know

, Legaltech News

Here are four ways AI is already helping you be efficient in your everyday activities.

Credit: PHOTOCREO Michal Bednarek/Shutterstock.com

Ask almost anyone to define artificial intelligence, and you'll likely get an answer that sounds like something out of a sci-fi movie. A few real-life examples might come up, with self-driving cars, IBM's Watson winning "Jeopardy!," and so on. But, for many, AI still evokes a sense of futuristic fantasy.


Defining AI is hard. A common (if vaguely circular) definition is the replication of human thought processes—language, learning, decision-making, and so on—by computers. In part because it's hard to define AI in general, it's also hard to say what particular technologies are or aren't AI. Optical character recognition seemed like AI until it made its way into almost every ATM in order to read deposited checks (and it doesn't help that "optical character recognition" isn't quite as awe-inspiring as "a computer that can read").

In practical terms, AI isn't really one thing. It's natural language processing, both spoken (Siri, Alexa, Cortana) and written (chatbots). It's image processing, both still (facial recognition, automatic categorization of photos) and moving (self-driving cars). It's good old-fashioned data analytics. There's a significant hardware component (server farms). And it's more things besides. But the end result is humanlike output: software that seems to recognize and remember things.

Practicalities aside, in everyday usage, AI is often applied to everyday technology: any software that serves a useful purpose and seems intuitive seems to get shoehorned under the AI umbrella. The colloquial definition of AI is very loose, and in part because of this looseness, aspects of AI have already made their way into technology people use every day, often in ways so familiar that they go unnoticed.

Here are a few:

Setting Good Defaults: Most people don't customize software, for good reason: customizing takes time that could otherwise be used to do real work. Instead, people stick with defaults. But where things get interesting is when defaults change depending on what you do.

This happens almost every time you use an autocomplete menu (it's just easy to miss). Right now, going into the holidays, when you enter "artificial" in the search field at the top of your web browser of choice, "artificial Christmas trees" is the top hit (the default) in the autocomplete menu. Search for "artificial intelligence" instead, and then search for "artificial" again. Now "artificial intelligence" is the top hit (a new default). It's blink-and-you-miss-it, but when this happens, you might briefly get a vague sense of a computer seeming to remember what you've done and using this information in a helpful way, which is AI by almost any definition.

Drag and Drop: In real life, the effect of moving an item from here to there is as straightforward as can be. On a computer, moving an item (using drag and drop) can be interpreted any number of ways. For example, drag a picture from File Explorer to an Office app: in Word and PowerPoint, the picture appears in your document (just as you'd expect); in Excel, nothing happens. Or drag a Word document from File Explorer to an Office app: in Word and PowerPoint, it appears as an icon or as a document that you can edit; in Excel, an error message appears.

This is typically the kind of thing you notice only when it's wrong (I am looking at you, Excel). When it's right, you might briefly get a vague sense of a computer creating a desired outcome based on evidence of your intent, also AI by almost any definition. (And, as an added bonus, you might come away with the illusion of being in control.)

Language Generation: When you're using a computer, certainly when you're using a computer to do work, you're generally working with text. Just imagine using a computer with no text.

Courtesy graphic

You can't get very far in most apps without text. Text is important, and the closer that the text a computer shows you resembles the text a human would show you, the better. For example, you would probably notice (that is, be distracted) if every piece of time data in every app you used looked something like 2015 10 15 07:28:42. So apps don't do this. Instead, apps use familiar words and phrases: just now, yesterday, last week, Thursday, Oct. 15, 10 minutes (plural), one minute (singular), and more. This is typically the kind of thing you notice only when it's wrong (one minutes). When it's right, you might briefly get a vague sense of a computer seeming to give you information as a human would.

User Interface Design: How an app presents information involves aspects of AI, or at least offers the benefits of AI. For example, in any modern email app, your messages are presented not as reams of inscrutable HTML, but in ways that make sense to humans: sender, subject, message text, and so on are readily apparent; new messages appear first; unread messages are highlighted; pictures appear in context; related messages are grouped together; and so on. To be sure, no one who works on natural language processing and self-driving cars and the like will consider this AI: this is user interface design. But if one of the goals of AI is to present an overwhelming amount of information in a way that makes connections obvious, makes distinctions obvious, and surfaces what's relevant, and user interface design strategies accomplish this, what's the difference?

The thought of AI robots replacing humans at the office might be a bit scary, but in reality, we encounter AI in some form every day. Far from putting lawyers out of jobs, AI—at least by its loose, mainstream definition—has helped make just about every human activity that involves a computer (including the practice of law) more efficient and productive. The more your computer seems to think like a human, the more quickly and effectively you'll be able to serve your clients, and the happier they'll be with the bill.


No comments:

Post a Comment