Why OpenAI Should Be A Platform

(original post at Searchblog)

Do generative AI innovations like OpenAI's ChatGPT and Google's LaMDA represent a new and foundational technology platform like Microsoft Windows, Apple iOS or the Internet? Or are they just fun and/or useful new products that millions will eventually use, like Google Docs or Instagram? I think the answer can and should be "both" - but to get there, the Valley is going to have to forego the walled garden destination model it's employed these past 15 or so years.

The question of OpenAI's ultimate business model has dominated nearly every conversation I've had this week, whether it's with reporters from the Economist and the Journal, senior executives at large-scale public companies, or CEOs of ad-tech and data startups. Everyone wants to know: What's the impact of generative AI on the technology industry? Will OpenAI be the next Google or Apple? Who wins, and who will lose?

It's been a minute since Silicon Valley has captured the world's attention at this level - I can think of just two examples over the past decade. The first crested in 2018, and the narrative was an ugly one: the Cambridge Analytica story catalyzed a five-year decline in Silicon Valley's popularity. The other has to be the rise of TikTok, and that story been similarly snake-bit:?Here was an inscrutable and addictive innovation driven by black-box AI and potentially hostile political motives. Not exactly Microsoft in the 1980s, Apple in the 1990s, Google in the 2000s, or Facebook before its Cambridge Analytica-driven fall from grace.

The Valley loves the story of a singular success - a mission-driven, entrepreneurial startup that redefines the industry and creates unimaginable wealth. Enter OpenAI. Its generative AI has engendered a largely positive story line. We'll note all those hard won caveats, but the truth is, it's been a very long time since we've had a potentially world-beating technology narrative on our hands. How might it play out?

OpenAI's recent moves - supporting a nuanced range of developer APIs while also?launching a paid consumer-facing service - makes it seem like the company wants to enter two markets simultaneously. The first positions the company as a platform service - let a million applications blossom! But the second positions OpenAI as a consumer portal along the lines of Facebook or Google: Come one, come all to our super awesome new service!

Not surprisingly, it's the second, more familiar consumer model that has captured the lion's share of coverage these past few days. Breathless pieces use well-worn methods of quantifying the success of Internet services: Here's Reuters reporting that ChatGPT has rocketed past 100 million users faster than any consumer application in history.?Here's respected analyst Azeem Azhar calculating ChatGPT's revenue potential based on those numbers: To The Moon! And here's VentureBeat asking if "anyone catch up and compete with OpenAI and ChatGPT?"

I'll leave that last question for future posts (erm, cough cough Google cough cough), but I want to step back and suggest that while it'd be sexy for OpenAI to become a consumer destination along the lines of Facebook or Google, it would also be a monumental mistake. There's a far larger opportunity - for OpenAI's shareholders, partners, and the public at large, if it were to single-mindedly pursue the platform opportunity instead.

When I imagine how we'll all most likely engage with generative AI in three to five years, I imagine it will be through a constellation of applications. ChatGPT might power my customer service experience with American Express or United online. It might also help me figure out a complicated itinerary at Expedia or on news services like Tablet. I'd likely find the same technology integrated with my Office or Google suites - who doesn't need better words and images in presentations and work memos? Make a list of everything you do online now - filing your insurance claims, managing your finances, wasting your time on gaming or social media sites, buying lord knows what, and yes, even searching for information - there's almost nothing we do online that I can't imagine being impacted by generative AI technology in some significant way.

And that's just for the applications we already use. The most exciting element of new platforms are their fundamentally generative nature - they enable unexpected and delightful new services and experiences. Who could have predicted Uber, Airbnb, or TikTok when smartphones gained platform status 15 years ago?

Bill Gates famously defined platforms as creating far more economic value for its users than for the platform itself. While I understand why OpenAI might wish to leverage its newfound status as a Silicon Valley rock star into becoming a world-beating destination like Google or Amazon, I hope the company will instead play the long game, and look to the platform models of Microsoft, Apple, or the open Internet.

There are reasons to suspect this is exactly what OpenAI will do. If the company was serious about consolidating its business around a destination model, it would be hard at work on search - and two months ago, that's exactly what most Valley observers assumed would do next. Instead, Microsoft announced it would integrate OpenAI's technology into its own search engine. ?I'm pretty sure that deal, with its attendant ten-plus billion dollar investment from Microsoft, doesn't contemplate OpenAI launching a Bing competitor. Put another way, that deal positions OpenAI as a platform, not a destination. Let a million applications blossom - that's good for the world, good for OpenAI, and good for the inevitable competition that to come, whether it be from Google, Meta, or the next world-beating startup no one's yet heard of.

--

You can follow whatever I'm doing next by signing up for my site newsletter here. Thanks for reading.