For Mira Murati, OpenAI’s chief technology officer, the evening of November 29 was no different to any other. She got back home from the start-up’s San Francisco offices that evening, after her team wrapped up the release of an experimental product: ChatGPT.
ChatGPT was lined up as a research demonstration. That plan changed dramatically when the chatbot hit 1mn users within five days of launch. Murati, who was at least an order of magnitude off in terms of her expectations, realised that the company was at the forefront of a race to commercialise generative AI — artificial intelligence systems that can quickly output humanlike text, images and content.
“This contact with reality was incredibly important to figure out what really to prioritise, and where to go next,” said Murati, who heads up the technology team for the 375-person company led by chief executive Sam Altman. “We definitely had an important shift when we decided to build product and deploy the technology.”
Now, Murati is leading OpenAI’s efforts to place ChatGPT as a standalone product, seeking ways to leverage its popularity with tens of millions of consumers, after the chatbot’s extraordinary rise. But she insists that despite new financial imperatives, the company’s main ambition is unchanged.
“Our mission is to get to artificial general intelligence and figure out how to deploy that safely,” she said, referring to a future software that could perform a range of cognitive tasks at human level. “And so we’re always very careful not to lose sight of that.”
Meanwhile, the company is reshaping the tech industry. The launch of ChatGPT, which has about 100mn users a month according to SimilarWeb, led tech giants such as Google to redraw their AI strategy and rush out rival chatbots. Businesses around the world have begun to experiment with the technology in the belief it could transform industries from media, finance, law and professional services.
Soon after the release of ChatGPT, Microsoft agreed a “multiyear, multibillion-dollar” investment in OpenAI estimated to be worth $10bn. The Seattle-based group is also incorporating the underlying technology across its Office suite of productivity apps, a lucrative base of hundreds of millions of large enterprise customers.
The chatbot helped to clarify for Murati and OpenAI’s senior leadership the clear business value of generative AI. “Our strategy on deployment now has expanded to . . . the ChatGPT platform where we have a direct relationship with the user,” she said. “We can ask them about their feedback, their preferences, and use that feedback . . . to make the code better,” she added.
A mechanical engineer who previously worked at augmented reality start-up Magic Leap, and electric-car maker Tesla, Murati joined the company in 2018 while it was still a non-profit research lab. She now oversees the mass-market distribution of OpenAI’s host of products, including AI image-generator Dall-E, AI code generator Codex and ChatGPT.
Since she joined, her role has evolved into testing OpenAI’s technology in the real world, so industry partners can build AI versions of products in everything from education, to financial services, law and healthcare.
Murati’s team has begun work towards offering a ChatGPT business subscription, which will allow clients to customise versions of ChatGPT for specific purposes. They have also launched so-called Plugins, which allows users to stay inside ChatGPT while performing tasks like browsing the web, buying groceries and booking restaurant tables through the likes of Instacart and OpenTable.
Today OpenAI, like many software companies in Silicon Valley, touts itself as a “platform” company, with two types of offerings: one is their API or application programming interface, which allows third parties to integrate OpenAI’s software into their products for a fee, and the other is ChatGPT. Microsoft, its largest shareholder, will be a third, independent revenue channel.
“[ChatGPT] allows us to reach people directly and gather feedback and make our models more aligned, more helpful,” Murati said. “And [the API] is a platform that allows other people to build on top of our models.”
Customers of OpenAI’s API include a range of businesses such as education start-up Khan Academy, social media company Snap, and asset managers such as Morgan Stanley Wealth Management. Each pays to tailor the tool to their individual needs. OpenAI, which reportedly made $28mn in 2022, has projected it would make $200mn in 2023.
The company needs to increase revenues in order to fund the eye-watering costs of computing power required to train and run large AI models, with Altman recently describing the company as on track to be “the most capital-intensive start-up in Silicon Valley history”.
Estimates put ChatGPT’s running costs, assuming 10mn monthly users, at $1mn per day. Microsoft chief Satya Nadella claimed that Microsoft had built a supercomputer to handle OpenAI’s work, and that it could now handle some AI calculations at half the cost of its rivals.
The potential risks of AI software, especially as it becomes more powerful, are top of mind for policymakers and the business and research communities.
Among these are challenges like the spread of AI-generated manipulation and misinformation. In the longer term, there are concerns about existential risks from AI tech that may behave in unethical ways, without a sense of human morality.
Last month, 350 industry insiders, including Altman said that mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.
Murati said she supported the use of technical “guardrails” to reduce, erase or even prevent some of these dangers.
“Progress is happening very fast, obviously,” she said. “But today, we use these systems more like assistive tools. We don’t rely on them blindly, or entirely. They’re more like tools that enhance our productivity [and] creativity.”