Mira Murati was OpenAI’s Chief Technology Officer on the evening of 29 November. It was just like any other night. After her team had completed the testing of ChatGPT, she returned home that evening from the San Francisco offices of the start-up.
ChatGPT had been planned as a demonstration for research. This plan changed drastically when the chatbot reached 1mn users in just five days after launch. Murati, who had expectations that were at least a magnitude off, realized that her company was leading the race to commercialise artificial AI systems which can output text, images, and content that is human-like.
This contact with reality helped Murati to determine what to prioritize and where to move next. Murati is the head of the technology team at the 375-person firm led by CEO Sam Altman. “We had a major shift in our thinking when we decided to develop the product and deploy technology.”
After the chatbot’s incredible rise, Murati now leads OpenAI’s effort to position ChatGPT in a standalone market, looking for ways to leverage this popularity among tens and millions of consumers. She insists that the main goal of the company remains unchanged, despite its new financial imperatives.
She said that “our mission is to achieve artificial general intelligence, and to figure out how we can deploy it safely.” She was referring to future software that would be able to perform cognitive tasks on a human level. We’re very careful to never lose sight of this.
The company is also reshaping technology. ChatGPT’s launch, which according to SimilarWeb has around 100mn monthly users, led Google and other tech giants to re-evaluate their AI strategy, and to rush to release rival chatbots. Businesses have been experimenting with the technology around the globe in hopes that it will transform various industries, including media, finance and law.
Microsoft announced a $10bn investment in OpenAI shortly after ChatGPT was released. Microsoft, which is based in Seattle, has also integrated the technology into its Office productivity suite. This includes hundreds of millions large enterprise clients.
The chatbot clarified for Murati’s and OpenAI senior leadership, the clear business value that generative AI brings. “Our deployment strategy has now expanded to include. . . “We have a direct connection with users on the ChatGPT platform,” she explained. “We can use their feedback and preferences to improve the service.” . . “We want to improve the code,” she said.
Murati, a mechanical engineer, previously worked for Magic Leap and Tesla. She joined the company as a non-profit in 2018, when it was still an unincorporated research lab. She oversees the distribution of OpenAI products such as AI image-generator Dall-E and AI code generator Codex.
Since joining, her role has evolved to include testing OpenAI technology in the real-world so that industry partners can develop AI versions of products for everything from education, financial services, legal and healthcare.
Murati’s team is working on a ChatGPT subscription for businesses, which allows clients to customize versions of ChatGPT according to their needs. They also launched Plugins that allow users to remain inside ChatGPT when performing tasks such as browsing the internet, purchasing groceries, and booking restaurant table through Instacart or OpenTable.
OpenAI, as many Silicon Valley software companies, is now promoting itself as a platform company. They have two offerings: the API, or application programming interface. This allows third-party software developers to integrate OpenAI software into their own products in exchange for a small fee. The other offering is ChatGPT. Microsoft, the company’s largest shareholder, is a third revenue stream.
Murati explained that “[ChatGPT] enables us to reach out directly to people and collect feedback, allowing us align our models and make them more useful.” “And [the API] provides a platform for other people to build upon our models.”
OpenAI’s customers include businesses like Morgan Stanley Wealth Management, Morgan Stanley Wealth Management, Snap, a social media company, and Khan Academy. Each pays for the tool to be tailored to their needs. OpenAI, that reportedly earned $28mn by 2022, projected to make $200mn by 2023.
Altman, who recently described the company as “the most capital intensive start-up company in Silicon Valley’s history”, said that the company must increase its revenues to pay for the staggering costs of computing power needed to train and run large AI model.
ChatGPT costs are estimated at $1mn a day, assuming 10mn users per month. Microsoft’s Satya Nadella said that Microsoft built a supercomputer that could handle OpenAI’s work and that Microsoft was now able to perform some AI calculations for half the price of its competitors.
Policymakers, business leaders, and researchers are concerned about the potential risks of AI, particularly as software becomes more powerful.
These include challenges such as the spread of AI generated misinformation and manipulation. Longer term, AI technology that behaves unethically and without human morality could pose existential threats.
Altman was one of 350 industry insiders who said last month that mitigating AI’s risk of extinction should be a priority for the world, alongside other risks at a societal level, such as pandemics or nuclear war.
Murati said that she supports the use of “guardrails”, which are technical devices, to reduce or eliminate some of these hazards.
She said, “It is obvious that progress is occurring very quickly.” But today, these systems are used more as assistive tools. We do not rely on these systems blindly or exclusively. “They’re like tools to enhance our creativity and productivity.”