There are few things hotter in tech right now than artificial intelligence. You’ll hear people with titles like “chief experience officer” and “thinkfluence concierge” talk about “neural networks” and “machine learning” and “natural language processing.” Alexa, Siri, Cortana … the Google Assistant?
The idea is, you can talk to your computer as if it were a person. Granted, a child with a relatively limited vocabulary and set of skills, but technically, that counts. Eventually, the idea goes, you can hold full conversations with an AI chat bot, asking it to answer complex questions and undertake complicated tasks.
But nearly two decades ago, our current era of AI overload started with two simple sentences.
“It looks like you’re writing a letter. Would you like help?”
In the mid-’90s, Microsoft was exploring how to make personal computing more approachable to consumers who might never encounter a PC in an enterprise environment. The personal computer, still trying to shake off the assumption that it was a tool for writing office memos and playing games, was not quite mainstream yet
Amid this environment, Microsoft tried to create a friendlier, approachable operating system — a debacle known as Microsoft Bob. Instead of using office metaphors for computer functions (folders, files, recycle bin), Bob arranged things in different rooms of a virtual house. Guiding you through the house was a cartoon character that offered help and advice.
Bob was unveiled in 1995. As one edition of Soft-Letter put it at the time:
In fact, virtually all interaction with the program takes place through these “friends of Bob,” who constantly move around the screen, making suggestions, juggling objects, and otherwise displaying symptoms of pseudo-spontaneous behavior. Each of these dozen characters, moreover, has a distinct and often obtrusive personality; Bob users interact with their PCs by communicating with a psychotic MTV rat, a coffee-drinking lizard, a submissive rabbit, a hostile parrot, and other oddly-behaved creatures.
As a program, Microsoft Bob was a notorious flop, and little-known. But the underlying tech, like the “Microsoft Actor” file format, was updated and repurposed for a new project. In November of the next year, Microsoft released the latest version of its suite of applications, Office 97. In doing so, they released a metallic menace upon the world. Bob’s son, Clippy, had been born.
Clippy was ostensibly designed to make writing easier. If you started a document with “Dear So-and-so,” Clippy figured out that you were writing a letter, and tried to offer help. The Assistant feature was meant to function, well, like an assistant: You could ask it to do something, and then it would perform the task … somehow. Automated assistants are a black box into which you place a request, and a response comes out the other end.
The Big Idea behind Clippy was to move away from the strict, regimented flow of computer tasks. In Clippy’s heyday, if you wanted to do something in Word — like, say, add footnotes — you had to know where the “Footnotes” menu item was, and how to navigate the preference window that popped up. But with a digital assistant, you could theoretically just dictate what you wanted — “add footnotes to this document” — and get a result.
If you had never used Microsoft Word before, this was nice! If you had used it before, Clippy was a nuisance. The problem, according to a postmortem written by Microsoft employee Chris Pratley, was that Clippy was “optimized for first use.”
What has happened is that the usability test showed that people who have never seen a feature before have trouble with it in the first hour of using it. So the designer makes the feature hold your hand through the process. That improves the results in the test, but ruins the feature for people who know what they are doing.
By the late ‘90s, enough people knew what they were doing in Microsoft Word that Clippy was less a helpful friend than a chipper redundancy. As one-time Microsoft employee James Fallows phrased it, “The next billion times you typed ‘Dear …’ and saw Clippy pop up, you wanted to scream.”
By 2001, Clippy was getting pink slips from Microsoft. An ad campaign for Office XP literally depicted the character as useless and obsolete, the most damning insult one can level at a piece of software. Clippy jokingly assumed that “XP” stood for “Ex-Paperclip.” By 2007, Clippy had been relegated to the dustbin of history, and Microsoft’s Assistant initiative was practically dead.
But a funny thing happened in the intervening years. In 2011, Apple announced the marquee feature of the latest edition of the iPhone 4s: Siri. Siri, which the company had recently acquired, let you talk to your phone as if it were a person, dispensing orders and asking questions. Zooey Deschanel famously queried, “Is that rain?”
Other chatty, helpful assistants soon followed. Microsoft introduced Cortana on its mobile and desktop operating systems, and Amazon discovered a bona fide must-have gadget with its Echo speaker, which uses a talking AI assistant called “Alexa.” Maybe most familiar to Microsoft Word users of the 1990s is the Google conversation AI assistant, which is at the heart of the operating system on Google’s new Pixel phone.
The current assistants are better than Clippy. One key difference is that this generation of assistants perform most of their number-crunching in the cloud, on remote servers far more powerful than a turn-of-the-century Gateway tower. They can fulfill more tasks and understand more requests. But they come from the same kernel of an idea: that your computer should be able to understand what you want, and do it automatically. What makes them an improvement is that they’re not intrusive. Clippy interrupted you, but Alexa and Siri only show up when you call. They analyze and sort your photos and email in the background. They don’t announce themselves.
But the idea is essentially the same. “We are just at the beginning of social interface,” Bill Gates told a crowd at the CES trade show in 1995. “The whole way you interact with the machine will be different, you’ll be able to talk to the machine, and it will use voice recognition, or so-called natural-language processing, to be able to understand what you do.” This sounded like an absurd thing to say when you consider that he was announcing the technology that would become the world’s most-hated paper clip. But if you’ve talked to your iPhone or your Amazon speaker or asked Google a question in plain English, you know he was right.
Eventually, we got to the point where AI assistants are a viable option for computing. But before all of these innovations came to pass, there was Clippy, a mascot whose fatal flaw was the hubris of announcing himself.