In Silicon Valley, We Keep Seeing the WeChat Official Accounts of Four Years Ago. What's Going On? | What Silicon Valley Is Anxious About❶

In the end, this is about building artificial intelligence — the tech industry's search for its next platform.

In early June, at a meetup for user-experience designers in San Francisco, one segment gave small groups ten minutes to brainstorm the design of a new app.

My group — me and four other designers — was handed the brief "an app for finding a gym." One of the designers insisted on putting a "chat interface" in the most prominent spot the moment the app opens, so users could type in what they wanted.

Chinese users will find this familiar. It is exactly what WeChat did when it launched official accounts in 2012.

The other three designers disagreed with him — not over why finding a gym needs a chatbot, but over exactly where to put the feature.

When I asked, "Do you all think chatbots are the future of human-computer interaction?" all four gave the same answer: definitely.

These designers are not outliers in the American tech industry.

In San Francisco and the surrounding Bay Area, there are several chatbot events every week. Not long after that designers' meetup, I went to a chatbot forum in Palo Alto, in the heart of Silicon Valley. The moderator opened like this:

"Quick poll. If you think chatbots will replace apps within five years, raise your hand."

Ilya Gelfenbeyn was one of the people who raised his hand. The founder and CEO of the startup API.ai, he sells developers a bot interface that lets them build bots with different functions quickly. His optimism is not unrelated to his own recent surge in business.

"Developer numbers have jumped in the past two months — 500,000 developers are now using our service to build chatbots," Gelfenbeyn told QDaily.

It looks much like WeChat's situation in China. Today most of the chatbots in the US that anyone actually uses live on the country's most popular chat tool, Facebook Messenger.

In April, Mark Zuckerberg announced that Facebook Messenger would open up, letting developers bring their own chatbots onto a chat tool that more than 900 million people worldwide already use every month. Google followed last month.

Funding for startups in this field keeps getting better, too. Bots that handle customer service and shopping advice for companies, bots that check the weather or the news… it seems every function and every app can have a layer of chat interface added on top.

It Feels Like WeChat in 2011 All Over Again

All of it is exciting, except the products. They have a familiar feel: WeChat official accounts.

Here, for example, is CNN's news chatbot. Every day it sends you some of the latest news, and you can choose to read the full story or just a condensed version.

As with a WeChat official account, you can also talk to it directly and ask about a particular story.

It is not just familiarity. Use these chatbots for a while and you find they are as primitive as WeChat official accounts, no more capable. Take the personal assistant Jarvis:

In theory Jarvis can set reminders for you. But it hems your language in tightly: follow its template exactly and simply say "in an hour" instead of a specific time, and it has no idea what you are talking about.

Other bots are much the same. The ones that recommend restaurants only send back accurate results if every word of your sentence matches the template.

Their current level is roughly that of keyword auto-replies on a WeChat official account.

Tencent launched official accounts in July 2012, and back then accounts were not only for publishing articles. It was popular to use the keyword auto-reply function to turn your account into a "customer service rep" or "bot" answering a long list of keywords.

Tencent has since built corporate customer service into official accounts, while ordinary media accounts went around the loop and came back to pushing out articles.

And if you think Siri is not smart enough, the bots on Facebook Messenger right now have not earned the right to be called dumb.

Going by the enthusiasm of the entrepreneurs, though, the end result should be this: you type a few words into a chat app, the way you would talk to a person, and your breakfast is ordered and your flight rebooked.

To get there, the first step is understanding what the user is saying. That is not easy.

Behind This Is a Push to Fold AI Into Everything

"If the difficulty of AlphaGo winning a world championship is 1, then the difficulty of AI fully understanding human natural language is 10."

That is how Sanjit Dang, an investment director at Intel Capital, described the gap between the ideal and the reality to QDaily.

Huge as the gap is, investors like Dang remain enthusiastic. "Chatbots have been around a long time," he said. "What we value now is the artificial intelligence driving them."

Out from under Facebook's constraints, chatbots get somewhat more powerful.

X.AI is one of the most written-about startups here, and it lives inside email.

You have to connect your existing calendar and a few places you go often. After that, if you want to fix a time and place for a work meeting or a get-together over email, all you do is cc the virtual assistant Amy. Like this:

Besides X.AI, plenty of similar personal assistants in vertical fields are leading with artificial intelligence — the smart fitness coach Shae, for instance, and Maluuba, a startup building semantic recognition for AI assistants.

In the first half of this year, the number of AI startups was six times what it was in 2011. In the first quarter alone, 143 AI-related companies raised money:

Investment data from CB Insights shows AI startups took in more money than ever in the latest quarter. Year by year, the trend is clearer still:

However Bad Today's Products Are, the Tools That Let Startups Fail Cheaply Are Already Here

Startups are pouring in because the technology underneath has matured.

Last November, Google open-sourced its machine-learning tool, TensorFlow.

Put simply, a developer who uses the tool gets the results of Google's machine-learning work — reading text and images, understanding what a user says. Entrepreneurs no longer have to learn how to recognize speech from scratch; they can concentrate on what to do with the information once they have it.

Seven months after release, TensorFlow had been starred by developers more than 25,000 times on the code-hosting platform GitHub, making it the most popular machine-learning project there.

In Silicon Valley, TensorFlow is also treated as one of the coolest products going; you will have no trouble running into entrepreneurs and engineers with TensorFlow stickers on their laptops.

And in a place that has no shortage of tech events, TensorFlow meetups are hard to get into — often booked out weeks in advance.

Google is not alone. Microsoft has open-sourced its own AI platform, and the guess-your-age-from-a-photo app that went viral for a while was one thing built on that kind of platform.

Bot Libre, an open-source bot platform founded three years ago, has seen a clear rise over the past three months in developers using its product to build bots, founder James Sutherland told us.

Most of them put the bots on their own company pages or on social networks, to take over customer service and inquiries.

If TensorFlow is the technical foundation, then Facebook opening Messenger to chatbots in April is the doorway that gets AI services to users faster — a shot at 900 million of them.

AI Is the Next Future Silicon Valley Is Looking For

It is no coincidence that chatbots today look like America's WeChat official accounts.

WeChat began as nothing more than an instant-messaging tool, then from 2012 started absorbing all kinds of functions and became a super-portal for services. It sucked up most of the time Chinese users spend on their phones, and along the way helped Tencent challenge Baidu's digital advertising business.

All of this caught Silicon Valley's attention.

After WeChat's international push failed in 2014, most parts of the world lost interest in it.

The one exception is the United States:

Important figures in Silicon Valley's startup world, among them Y Combinator partner Sam Altman and Snapchat founder Evan Spiegel, have talked publicly about WeChat's model.

Ted Livingston, founder of Kik Messenger — another major chat tool, and the earliest inspiration for the WeChat app — has said outright that he wants to build the "WeChat of the West."

The company playing WeChat's part in the US is Facebook. WhatsApp and Facebook Messenger, both in its hands, reach well over a billion users. Last year it built Uber's ride-hailing service into the app.

Letting chatbots move in is a big step toward Facebook building a WeChat-style platform.

Small and mid-sized developers have been waiting for that step, because starting an app from zero today and getting people to download it is simply too hard.

By one estimate, there are close to 600 million iPhone users. But Apple's store can only display so many apps, and 2 million apps are fighting over that limited space. A new app, especially one from a startup, has little hope of being noticed.

Facebook Messenger, by comparison, has 900 million users and only a few hundred chatbot services so far. The odds of being found are much better.

The Question Is Whether "Conversation" Can Really Become the Next Gateway

America's biggest tech companies have all seen the value of conversational services. Facebook and Google have both launched chatbot platforms on their own chat tools. Amazon built its own conversational bot, Alexa. Microsoft has Xiaoice, and also made an American version of it and put it on Twitter.

Even Apple, relatively less aggressive about artificial intelligence, announced a set of attempts last week to fold AI into iOS, and opened iMessage to third-party services.

What remains in doubt is whether "chatting" will become the way we interact with machines in the future.

Next to a button on a screen, chat is far too unreliable. It will not be 100% accurate for a long time yet, while pressing a button on a screen always gets you a definite result.

For everyday services like reading the news or hailing a car, the button on the screen is still the most convenient thing there is.

As the chatbot forum in Palo Alto was ending, the moderator put it to the room again.

"I'll ask once more: if you think chatbots will replace apps within five years, raise your hand."

Barely anyone did.

"If you think most apps will have artificial intelligence built into them within five years, raise your hand."

Most of the room raised their hands.

Editor's note: the PC, the internet, the smartphone… the spread of each foundational technology has come with years of rapid growth. For the tech industry, that growth has now run for forty years.

It may be about to pause: the growth the smartphone brought has already slowed, and since ride-hailing and food delivery we have not seen another breakthrough on the scale of social networking or e-commerce, the kind that moves hundreds of billions in wealth.

What new technology will clear the way for the whole industry's next stretch of growth? And will it arrive in time?

These questions will affect everyone, from giants like Apple and Google, to venture capital firms and startups, to every user of a tech product — basically everyone.

Over the past forty years, most of the technologies that laid the ground for the tech industry's lasting prosperity came out of Silicon Valley. We will run a series of reports from Silicon Valley, looking at this change from different angles.