After decades of depictions in sci-fi, the era of consumer-level AI finally seems to be upon us. Siri is well-ensconced in our phones, while voice-activated, personal assistants like Amazon Echo, Google Assistant, and Apple’s just-announced HomePod, are all vying to become our household servants (or what Walt Mossberg recently dubbed “information appliances”).

While HomePod takes a small step in the right direction, all these devices are still fundamentally designed around some highly flawed assumptions. Only when a top tech company recognizes this, and is willing to make some key course corrections, will we see consumer AI that’s worthy of the name.

Here’s what I mean:

Fun: Computing’s unacknowledged killer app

It’s half-jokingly said that porn made the consumer Internet possible, but it’s more accurate to say fun did. Even before web browsers, desktop PC sales were largely powered by the popularity of computer games. In the early ’90s, it’s estimated that a single game, Doom, was downloaded 15-20 million times — at a time when PCs were selling 37 million units a year. Online chatrooms, comedy videos, music files, image memes, and other varieties of fun soon followed. We saw a similar pattern when the iPhone launched decade ago, with games and other kinds of interactive entertainment immediately becoming the most-used apps, which remains the case to this day. To succeed as information appliances, PCs and smartphones first had to be embraced as entertainment appliances.

With that premise in mind, consider the current market of AI assistants: HomePod, Alexa, Google Home, and the rest are not likely to succeed on a mass scale because they’re missing the fun content.

Another problem is marketing: Just framing these devices as “personal assistants” that can answer questions and perform useful jobs, such as ordering ride services, or in HomePod’s case, intelligently select music on request, doesn’t convey to consumers a substantial differentiator from what they can already get within seconds though Google search or another app.

Which brings me to an even deeper problem:

Voice is no one’s choice (except Robert Downey Jr.’s)

Marketing voice command as a differentiator assumes that consumers actually want a product like Jarvis from the Iron Man films. Simply because a technology is featured in a fun movie doesn’t make it fun. In fact, the real reason sci-fi TV and movies have depicted talking computers for so long is a filmmaking conceit — it’s a good way to dramatically add computer interaction into the plot without boring shots of the hero typing on a keyboard.

It’s a mystery to me why so many tech giants have pushed voice as a home-based interface with little evidence consumers are actually interested in it. According to one study last year, the vast majority of iPhone users rarely use Siri, with most of us only turning to her for navigation assistance while driving. By contrast, there are numerous online videos depicting consumers using Siri for what she was not designed for — trying to flirt with her, asking her to tell jokes or respond to insults, and so on. Users turn to Siri, in other words, not for assistance but for entertainment.

So consumers who aren’t interested in talking out loud to a computer — that would be most of us — are likely to get confused as to why they would need a personal assistant at all. Especially when they already have a smartphone in arm’s reach. And their smartphones provide nearly everything they could want for digital entertainment.

Wanted: One fun-loving tech giant

This is why I think the real way forward for intelligent assistants belongs to a tech giant capable of creating a new kind of entertainment appliance. By marketing HomePod as a music speaker, Apple has a better approach than its competitors. But music is just one of many kinds of content that entertains us.

What’s needed is a device that helps us enjoy everything digital and interactive in our homes. It would use AI to seamlessly connect our smartphones and all the content we’re already using on them to the legacy entertainment appliances that still dominate our living rooms: televisions, stereos, and home theater systems. A company with the size and range of devices of a Samsung could build this — but only, perhaps, if it could somehow inject its corporate culture with the sense of fun that made a device like Nintendo’s Wii such an early crossover hit. From a marketing standpoint, the AI or personal assistant aspect would scarcely even need to be mentioned because consumers are not accessing it for AI or assistance per se, but for fun.

But that’s where it gets interesting. Users will come to such a device for fun, but machine learning integrated within the device would rapidly understand what kind of content its owners enjoy consuming and then gradually understand all their key behaviors.

And only then, once we’re all readily entertained, could it become a device that also helps us with our daily chores and needs. And, at last, it would become the intelligent assistant and information appliance we’ve always hoped computers would one day be.

Balaji Krishnan is a Cupertino-based entrepreneur and engineer. Currently founder and CEO of DabKick, a live media watching experience, he previously founded Cruxle, an AI-driven recommendation engine, and Snapstick, a mobile-to-TV streaming technology acquired by Rovi.