Why AI now?


The current level of energy around the topic of artificial intelligence (AI) demands an answer to one simple question, says Peter Coffee, and that is, “What’s different this time?”

Peter Coffee
Peter Coffee

The current level of energy around the topic of artificial intelligence (AI) demands an answer to one simple question: What’s different this time?

For literally half a century, AI seemed to be an ever-receding destination that was always “20 years from now.” In the 1960s and 1970s, optimism and expectations for human-level computer performance were high — and thought to be achievable by the mid-1980s — but there were disappointing shortfalls and financial returns were unimpressive. As reported by The Wall Street Journal in 1990:

After swallowing up hundreds of millions of dollars in venture capital, hundreds of AI start-ups have yielded only a few profitable public companies.

Disgruntled investors and self-deprecating researchers redefined “AI” to mean “almost implemented.” Some call that period of disappointment “the first AI winter.”

In just the past few years, the balance has suddenly tipped the other way and experts are routinely being confounded by progress more rapid than their well-informed expectations. For example, I’ve heard Andrew McAfee of MIT (at Zuora’s Subscribed conference in 2014) admit that his first ride in a Google self-driving car came many years before he expected driving a car to leave the list of “things only people will be trusted to do.”

Crucially, McAfee added that “within about ten minutes, you’ve stopped thinking it’s remarkable.” Not only is AI suddenly happening, but we even seem to be quite ready to stop being impressed by it.

What’s changed?

The most significant developments over the last two decades that are taking AI from concept to reality include:

  • The massive proliferation of connectivity and mobility.
  • The fact that systems today are simply capable of knowing more without requiring a person to teach them.
  • The large amount of data available combined with the low cost of extremely powerful cloud computing infrastructure.

For example, we had meaningful AI in the early 1980s when Digital Equipment Corp. built a practical and effective “expert system” called XCON to assure complete and correct configuration of custom-ordered minicomputer systems. Within a short time, the difference between an amazing version 1.0 and a sustainable capability had become painfully clear.  Back in 1987, researchers reported:

Over 7 years, XCON has grown to 6,200 rules, of which approximately 50% change every year. While the performance of XCON is satisfactory, it is increasingly becoming more difficult to change.

I consider this the canonical example of the failure of scalability and continuous improvement in the previous epochs of AI — resource starvations that led to what’s been termed the “second AI winter.”

What’s next for AI?

AI now exists, under various names for specific skills, in a rapidly expanding range of products and services that people use every day. Machine learning powers personalized content recommendations in Amazon and Netflix. Apple’s Siri and Amazon’s Alexa use natural language processing technology to understand voice commands. Salesforce Inbox relationship intelligence turns email into an opportunity cornucopia.

Increasingly autonomous cars from makers like Tesla and Google use AI for collision avoidance and traffic congestion logic. All of these products are now coming of age in a world quite different from the climate of past “AI winters” – a connected world where events arrive in a steady stream, and people can coach their machine assistants on what to do next.

Accelerating AI progress will unleash new levels of productivity, augment our personal and professional lives — and certainly intensify long-debated concerns about the relationship between man and machine. Businesses will be able to apply machine learning algorithms across billions of signals, suggesting which customers are most likely to purchase a particular product; automatically escalating and routing customer service calls to the most appropriate agent.

“Assistant intelligence” might be the next re-translation of AI, as algorithms plan events and vacations with acumen approaching that of a human assistant — but with never a missed detail or departure from best practice. Next-generation “analytics intelligence,” yet another AI, will surface and monitor sentiments around a brand by analyzing a planet’s worth of signals from social networks and other data sources.

Relevant intelligence, constantly fuelled by connected sources of real-time knowledge, will transform our expectations of how things get done. Do not wait for this to be obvious to everyone. Start looking for ways to make it useful now.

Image credit - Salesforce/Freeimages.com