Next Up For AI Chatbots: It’s All About The APIs
The magical demos displayed the capability. Now the platform wars begin.
This is an excerpt from my weekly newsletter, Big Technology, published on Substack. Check it out here and sign up to recieve it each Thursday.
Snapchat is building a bot into its app, called “My AI,” that will look a lot like OpenAI’s ChatGPT.
At first, it might seem like Snap and OpenAI are competing. But Snapchat’s bot is built on top of OpenAI’s technology. And that, of course, was always the point.
Headline-grabbing bots like ChatGPT and Bing Chat were never the end product themselves. They were always a method for OpenAI and Microsoft to showcase what they could do for other companies. And when other companies bought in, they’d pay to build on top of their tech stacks.
Snapchat’s My AI bot is just one of many applications we’ll see come out of this moment. Even Coca-Cola — as you’ll see in this week’s newsletter — is thinking about how to put this technology to use.
In other words, the demos worked.
Speaking of chatbots, I just hosted one of our most fascinating discussions Big Technology Podcast to date. Blake Lemoine and Gary Marcus joined to debate the nature of chatbots, and ended up agreeing on some fundamental concerns.
Lemoine, you’ll remember, is the ex-Google engineer who concluded the company’s LaMDA chatbot was sentient. Gary Marcus is an academic, author, and outspoken AI critic.
You can listen to the conversation on Apple Podcasts, Spotify, or your podcast app of choice.
And now, this week’s Big Story:
Next Up For AI Chatbots: It’s All About The APIs
The chatbots did their job. They inspired awe, mockery, and even some fear. Most importantly, they drew attention. Front-page headlines, cover stories, and word of mouth caused millions to try them, leading businesses and developers to ask how they could put the technology to use.
The APIs, of course, were always the point. ChatGPT and Bing’s chatbot were never the end product. They were demos meant to sell other companies on tools they could use to build their own. And it worked. Now, the war to build the leading generative AI platform is underway.
“For OpenAI, the vast majority of the money they will ever make will come from developers,” Ben Parr, president of Octane AI, told me via phone Thursday. “ChatGPT is just the entry road into everything else.”
Even before this wave of AI chatbots reached the public, the companies behind them prepared APIs for developers. When ChatGPT gained momentum in January, OpenAI president Greg Brockman teased an API “coming soon.” That same week, Microsoft made OpenAI models available through Azure. On the day Google introduced its BARD chatbot, CEO Sundar Pichai promised to make some of the underlying technology available by March. And this week, just a bit late, Amazon announced it would partner with Hugging Face to make a generative language tool available through AWS.
“Everybody who develops software is either alerted, or shocked into alert, or actively working on something that is like ChatGPT to be integrated into their application, or integrated into their service,” NVIDIA CEO Jensen Huang said during his company’s earnings call Wednesday. NVIDIA provides the chips the tech runs on, so it stands to benefit too. Its stock jumped 14% Thursday.
Finding broad, useful applications for generative AI will be challenging, but some obvious early applications stand out. Customer service departments, for instance, could use chatbots that can hold a conversation. Gaming companies could build intelligent characters and make NPCs a thing of the past. And marketers could attempt to use generative language models to forge deeper bonds with customers.
This is all moving fast. On Tuesday, OpenAI announced it had partnered with Bain to help clients build on its API. Zack Kass, OpenAI’s chief customer officer, said in a launch video that OpenAI couldn’t keep up with the interest in its technology. “We are inundated at this point with enterprise demand that we sort of waited for, for a long time, and here it is,” he said. “Now we just need to figure out how we field it.”
Later in the video, Coca-Cola executives said they planned to use the tech in their marketing efforts “to deliver creative content at speed.” Coca-Cola CEO James Quincey also mentioned he believed the tech would change knowledge work, but without going into specifics.
Coca-Cola is a fitting launch partner. In a recent presentation, investor Chamath Palihapitiya mentioned that Coke succeeded thanks to another invention: refrigeration. Coca-Cola made more money than the people that invented the refrigerator, he said, and that could happen here too. “If AI/LLMs are the refrigeration,” he asked. “Who will be the next Coca-Cola?”
The companies that enable successful AI applications — the refrigeration, in Palihapitiya’s analogy — still stand to benefit tremendously though. And so those developing the underlying technology are doing what they can to help launch the next big thing on their platform, and perhaps take a chunk of it too. OpenAI, for instance, has a $100 million startup fund meant to work with AI companies in health care, climate, education, and elsewhere. “Look at some of the companies that OpenAI’s invested in,” said Parr. “There are real use cases.”
The APIs, amid the commotion, are what matter. They’re why Microsoft was willing to release an unproven chatbot into Bing, even when it knew it was a bit crazy. And why the company didn’t seem to mind when the bot’s flaws exploded into public view. It was never about Bing or ChatGPT, but about the potential future they previewed. And now, given the demos’ success, the race to enable that future is underway.
