Home / Technology / State of the Art: A High-Stakes Bet: Turning Google Assistant Into a ‘Star Trek’ Computer

State of the Art: A High-Stakes Bet: Turning Google Assistant Into a ‘Star Trek’ Computer

That’s the threat. But the Assistant also presents Google with a delicious opportunity. The “Star Trek” computer is no metaphor. The company believes that machine learning has advanced to the point that it is now possible to build a predictive, all-knowing, superhelpful and conversational assistant of the sort that Captain Kirk relied on to navigate the stars.

Photo

Credit Stuart Goldenberg

The Assistant, in Google’s most far-out vision, would always be around, wherever you are, on whatever device you use, to handle just about any informational task.

Consider this common situation: Today, to book a trip, you usually have to load up several travel sites, consult your calendar and coordinate with your family and your colleagues. If the Assistant works as well as Google hopes, all you might have to do is say, “O.K., Google, I need to go to Hong Kong next week. Take care of it.”

Based on your interactions with it over the years, Google would know your habits, your preferences and your budget. It would know your friends, family and your colleagues. With access to so much data, and with the computational power to interpret all of it, the Assistant most likely could handle the entire task; if it couldn’t, it would simply ask you to fill in the gaps, the way a human assistant might.

Computers have made a lot of everyday tasks far easier to accomplish, yet they still require a sometimes annoying level of human involvement to get the most out of them. The Assistant’s long-term aim is to eliminate all this busywork.

If it succeeds, it would be the ultimate expression of what Larry Page, Google’s co-founder, once described as the perfect search engine: a machine that “understands exactly what you mean and gives you back exactly what you want.”

At this point, a few readers may be recoiling at the potential invasion of autonomy and privacy that such a machine would necessitate.

The Assistant would involve giving ourselves over to machines more fully. We would trust them not just with our information but increasingly with our decisions. Many people are already freaked out by what Google, Facebook and other tech companies know about us. Would we be willing to hand over even more power to computers?

Those are important questions, but they are also well down the road. For now, the more pressing question for the Assistant is: Will it even work?

Photo

Sundar Pichai, Google’s chief executive, calls the development of the assistant “a seminal moment” for the company. Credit Justin Sullivan/Getty Images

Google has technological advantages that suggest it could build a more capable digital assistant than others have accomplished. Many of the innovations that it has built into its search engine — including its knowledge graph database of more than a billion people, places and things, and the 17 years it has spent trying to understand the meaning of web queries — will form the Assistant’s brain.

Google has also been one of the leaders in machine learning, the process that allows computers to discover facts about the world without being explicitly programmed. Machine learning is at the heart of a number of recent advances, including Google Photos’ uncanny capacity to search through your images for arbitrary terms (photos of people hugging, for instance).

“We are in the process of transforming into a machine-learning company,” Jeff Dean, who is in charge of Google Brain, the company’s artificial intelligence project, told me this year. For each problem Google solves this way, it gets better at solving other problems. “It’s a boulder going downhill gathering more momentum as it goes,” Mr. Dean said.

If you use the Assistant today, you’ll see some of these advances. As my colleague Brian X. Chen explained last week, if your friend sends you a picture of his dog on Allo, Google Assistant will not only recognize that it’s a dog, but it will also tell you the breed.

That’s an amazing technological feat. But as Brian pointed out, it’s also pretty useless. Why does your friend care if you know his dog’s a Shih Tzu?

This gets to a deeper difficulty. The search company might have the technical capacity to create the smartest assistant around, but it’s not at all clear that it has the prowess to create the friendliest, most charming or most useful assistant. Google needs to nail not just Assistant’s smarts, but also its personality — a new skill for Google, and one that its past forays into social software (Google Plus, anyone?) don’t speak highly of.

Then there is the mismatch between Google’s ambitions and Assistant’s current reality. Danny Sullivan, the founding editor of Search Engine Land, told me that so far, he hadn’t noticed the Assistant helping him in any major way.

“When I was trying to book a movie, it didn’t really narrow things down for me,” he said. “And there were some times it was wrong. I asked it to show me my upcoming trip, and it didn’t get that.”

Of course, it’s still early. Mr. Sullivan has high hopes for the Assistant. It would be premature to look at the technology today and get discouraged about its future, especially since Google sees this as a multiyear, perhaps even decade-long project. And especially if Google’s future depends on getting this right.

Continue reading the main story


NYT > Technology

Leave a Reply

Your email address will not be published. Required fields are marked *

*