The importance of user onramps for emerging technologies; what the metaverse, blockchain, and other new tech can learn from OpenAI
On November 18, 2021, OpenAI dropped the waitlist to its large language model platform GPT-3. After more than a year of testing in private beta, the breakthrough technology was finally widely available to all via a public API. Over the next year, GPT-3 would be integrated into hundreds of products from leading companies including Microsoft, Zoom, IBM, Intel, Salesforce, and more. How did the world respond to this GPT-3 boom? It actually didn’t.
According to Google Trends, searches for “GPT-3” were flat for all of 2021 and the first half of 2022. So even after GPT-3 was open to the public, even after any person or company could access the natural language processing system and leverage the transformative AI platform, there was no increase in interest level in the technology according to Google. Either no one noticed, or no one cared.
But the second half of 2022 brought about a completely different story. On November 30, 2022, OpenAI launched ChatGPT, a new product built on GPT-3 that added enhanced language models and most importantly added a new interface: a chat box. The impact of that release was immediately noticed.
Interest in GPT-3 (again as measured by Google Trends) skyrocketed as millions of people began trying out the technology through ChatGPT. In early February, it was reported that ChatGPT exceeded 100 million users just two months after its initial launch, making it arguably the fastest growing consumer product ever. Fueled by the fevered excitement for ChatGPT, venture capital investments in AI startups has boomed (despite overall startup fundraising falling 63% YoY in Q4 2022) with the most notable example being OpenAI’s own $10 billion funding round in January. There is an AI frenzy happening sparked on GPT-3.
Again, ChatGPT is built on GPT-3, which opened to the public in 2021. The large language model that has captured the world’s attention and imagination was widely available for a year before ChatGPT’s launch. But the only way to experience GPT-3 directly back then was through an API. In other words, the user onramp to GPT-3 was previously steep and hard to access because it required programming. Then with ChatGPT, OpenAI created a shallow user onramp, an easy way to experience GPT-3 for yourself. People could now access GPT-3 through a text field instead of an API call. Instead of writing code, users just had to… write.
For technology to change the world, people have to be able to actually use it. ChatGPT did just that for GPT-3. Before ChatGPT launched, GPT-3 was promising technology. But after ChatGPT, GPT-3 became revolutionary technology. Before ChatGPT, OpenAi.com was a nice company website. But after ChatGPT, it’s now one of the Top 30 most popular websites in the entire world. And all because of the shallow user onramp of ChatGPT.
Steep onramps for the Metaverse and Blockchain
On March 23, 2023, Meta published a blog post pushing back against European mobile network operators’ efforts to require additional fees from content providers. Embedded within their argument against these fees was this interesting quote: “metaverse adoption for the foreseeable future will continue to be driven predominantly through Virtual Reality”.
When Facebook announced it was changing its name to Meta to show its commitment to the emerging metaverse (coincidentally just a few weeks before the GPT-3 public release), the company described the metaverse as allowing users to “share immersive experiences with other people even when you can’t be together”. Virtual Reality was briefly mentioned and is certainly a component of the metaverse, but it was not the entirety of the metaverse. Mark Zuckerberg even wrote about using “phones and computers to jump in from existing platforms” when describing his metaverse ambitions that were broader than just VR.
Over the subsequent 18 months that followed, that position changed. Meta would consistently link the metaverse to Virtual Reality in every statement, announcement, and interview until it became defined by VR. To experience the metaverse, you had to do so with VR. VR was now the user onramp to the metaverse.
But that’s a problem because VR is also an incredibly steep user onramp. VR requires custom hardware that covers your eyes and may make you nauseous and dizzy. So to try out the metaverse, Meta is asking users to wear a device resembling bulky glasses when people spend billions of dollars on contact lenses and LASIK surgery each year to get rid of glasses. It’s not surprising that more than half of all Meta’s Quest VR headsets sold end up being unused after 6 months.
Emerging blockchains like Bitcoin, Ethereum, Solana and more also suffer from the same challenge of a very steep user onramp. The technology is fundamentally rooted in cryptography, which becomes painfully and confusingly apparent when trying to get started with blockchains. The onramp to blockchains is the crypto wallet, and setting up the wallet involves having to deal with cryptography concepts like managing private keys or passphrases and digitally signing transactions. And on top of that complexity, any mistake you make with your crypto wallet comes with the risk of losing all of your cryptocurrency assets. So not only is the blockchain user onramp steep, a slip up on that onramp can be catastrophic.
None of this is a debate about how compelling, impactful, or important metaverse and blockchain technologies are. The metaverse may be truly revolutionary and the blockchain truly transformative. But no one will know if the steep user onramp of VR for the metaverse and crypto wallets for the blockchain continue to prevent people from accessing those technologies in the first place.
For any emerging, bleeding edge technology to thrive, it needs a shallow user onramp to make it accessible. GPT-3 and large language models found its shallow onramp through chat. What will be the shallow onramp for the metaverse, blockchains, and other “next Internet” technologies to reach mass audiences?
Maybe GPT-3 has the answer, and because of ChatGPT, it’s really easy to ask.