T.The last year has been marked by a slew of new artificial intelligence (AI) programs being released into the world since OpenAI, a Microsoft-backed research lab, launched ChatGPT in November 2022. Both Microsoft and Google announced the following products in March: Using AI to transform work, IBM’s CEO Arvind Krishna said his AI tools at the company could cut him 30-50% of repetitive paperwork. rice field.
Since taking over at Microsoft in 2014, Satya Nadella has focused on keeping the company relevant at a time when traditional software products were losing market dominance. . The company has invested heavily in its cloud computing platform Azure and AI, pouring at least $13 billion into its flagship lab, OpenAI. Since Nadella took over as CEO, Microsoft’s stock has risen nearly 10-fold, outperforming the S&P 500, which has only doubled in value over the same period.
Now, Nadella is using those investments to revitalize Microsoft’s traditional Office suite of products, including Word, Outlook and Excel, now called Microsoft 365. In his March, Microsoft launched “Copilot,” his AI tool that says it will get people out of work. We help with the tedious tasks of drafting emails and white papers, transcribing and summarizing meetings, and teaching how to interpret data in Excel. Copilot was initially released for small enterprise customers, but Microsoft is now rolling out the system to a larger customer group.
Improving office software like Microsoft Word might not sound so exciting given some of the headlines about AI and its potential uses, from virtual companions to fitness coaches, but this is us. may be one of the most impactful use cases for many. It is alive. Nadella says improving the way we work helps us thrive both as individuals and as a society.
TIME spoke with Nadella about Microsoft’s principles for AI, how this technology will transform work, and what safety measures should look like.
This interview has been condensed and edited for clarity.
You said AI will unleash a new wave of productivity and take the drudgery out of our daily routines. What specific things do you think will change in the workplace with the introduction of AI?
AI itself is very much present in our lives. But if anything, we’re moving from autopilot to copilot to help us do our job. You can empower humans by putting them at the center and creating this tool around them.
Not only does it increase productivity, it actually takes the drudgery out of the job. Thinking about all of us at work, how much time do we spend expressing ourselves and creating? This is what gives us real joy. How much time will you spend just adjusting? So I think we would all be better off if we could tip that balance in the direction of being more creative.
There are also concerns that this could lead to job losses. How does Microsoft take responsibility for addressing these concerns about turnover, and what is it doing in that area?
One of the things I’m most excited about is how. [AI] Democratize access to new skills. To give a concrete example, a developer using GitHub Copilot sees his productivity increase by 50-plus percent and stays in flow more. We have about 100 million professional developers, and he probably has 1 billion professional developers in the world. The total number of developers will increase significantly as the barriers to becoming a software developer are lowered. This doesn’t mean that good software developers will continue to be good software developers, but that it increases the ability of more people to enter the field.
Is it Microsoft’s responsibility to help displaced people acquire these new skills?
It absolutely is. In a way, it’s good for our business too. Our mission is to enable every individual and every organization on the planet to achieve more. So for me, this is a great way to create higher paying jobs, more empowering jobs, jobs that give people more meaning.
What are your biggest concerns about future AI adoption?
There is one thing that I really, really like about the way this dialogue is going. It’s not just optimism about technology. It’s about thinking about technology and its opportunities, but also about the responsibilities of the tech industry and the broader unintended consequences, and how to mitigate them long before they spread to society. . So I think that’s the right way to go in 2023. Having both of these conversations at the same time shows the level of maturity in both our industry and civil society.
That’s why perhaps the biggest investment we make, even when we think about AI, is not thinking about it responsibly. It puts humans at the center, not just in abstract principles, but in the engineering process and in the design choices we start with. It’s a design choice.
There is dialogue, then regulation. If you were the government, what would you do to ensure adequate regulation to protect the public from AI?
Already have. Think about what Microsoft did before there was generative AI and all these copilots. Consider what Microsoft has done with neural voices. While there are no laws yet, we ourselves have a lot of governance over who can use neural voices and how. I think there is a forum for dialogue, and I think there is also a forum that expects us to take responsibility as providers of this technology before regulation and regulation after that. But at the end of the day, I think we’re all going to be judged on just one thing: whether the benefits far outweigh the impact on society.
TIME reports that Microsoft is lobbying against proposals to regulate general-purpose AI in Europe. Why is Microsoft specifically involved in this debate?
I’m not particularly familiar with that particular comment about what we do or don’t do in Europe. But I think the fundamental thing is that in the end the European Union and its regulators do what is best for Europe and we participate in Europe within the framework of legislation and regulation. What regulators, governments and society really need to do is strike the right balance between the benefits of this technology and the unintended consequences they want to mitigate. We are willing to discuss it and try to prevent the former from happening and the latter from happening.
What unintended consequences do you think regulators should be very careful about mitigating?
For example, consider bias here and now. One important thing is to ensure that biased output does not harm the real world in unintended ways when using these technologies. You have to think about the origin of your data. What are we doing to de-bias these models? This is where Microsoft has done a lot of work, both in the pre-training stage and after the models have been deployed. .
Do you agree to the restrictions on the use of AI for military applications?
I think we’ve always said that we want to make sure that the very best Microsoft technology is available to the very institutions that protect our freedoms.
You may have seen the open letter to major AI labs asking them to stop training new AI for six months, but TIME has an op-ed urging labs to shut down AI entirely. It has been. Should I slow down and brake a little?
I think there are two things that are important for us to have a solid discussion about. The first, which is here and now, is what impact will AI have on the real world once it is introduced?
And the second part, which I think is also worth talking about, is how we make sure that any intelligent system we create is controlled and aligned with human values. That’s it.
We must rethink practical ways to take advantage of these solutions and mitigate their unintended consequences. Ultimately, however, these decisions are made by regulators and governments concerned.
What about this idea that the developers behind AI may not fully understand the results it produces?
I belong to the camp of thinking that one should not immediately relinquish responsibility for oneself. This is the most probabilistic complex system. There are many probabilistic complex systems that we work with. We use a number of other evaluation tests to characterize these probabilistic complex systems and ensure that they are safely deployed. So this isn’t the first time we’ve dealt with complexity in the real world.
What is another example of such a system?
biology; environment? We observe a lot and try to get empirical results. We try to address it in a way that we really benefit from what we do. I feel like I’m going to be quick to say that this is the final frontier and the only technology.
I think this is a really great set of technology. These show scaling effects. We need to have a firm grasp of what is happening, characterize them, and be able to deploy them safely. All I want from Microsoft is to try harder. Work hard to make sure the technology and its benefits far outweigh any unintended consequences.
For example, if we think about biology, it exists in the world we are exploring and trying to understand. AI, on the other hand, is of our own making, which is probably why it inspires so many fears. Given that AI is what we do, should we be a little more cautious than other systems like biology?
As Vannevar Bush writes in As We May Think, what was the true origin of the entire computer industry? The computer industry was meant to create tools for the human mind so we could do more and understand the natural world better, like climate and biology. . Therefore, I feel that at the core of enlightenment is creating technologies that enable us humans to increase our knowledge, do science, and help the human condition. So trying to say “now is the time to stop” doesn’t seem like the right approach.
There seems to be an urgent need to get the most out of AI. What drives that urgency? Is it the shareholders, the research community, or Microsoft executives?
According to my book, global economic growth is in a kind of stagnation. In fact, the last time economic growth could be attributed to information technology was when it appeared in productivity statistics when PCs became ubiquitous in the workplace.
So if anyone in the world is to have economic growth and really has a goal that it should be climate positive, then there must be trust in society around it. We need to build new technologies that achieve both of these goals. I think that’s why AI is so interesting.
That doesn’t mean AI won’t have unforeseen consequences, whether it’s forced displacement of the workforce, safe placement or bias, and we need to address them. But let’s not confuse it with the need for new technology to help the economic growth we enjoyed in the early 20th century. What if that kind of economic growth could happen? But this time, it’s much more equal than just the West Coast to the United States. [but] Everywhere in the world: small businesses, large corporations, public and private sectors. It’s a beautiful world that I aim for.
Are you arguing that if we decide to embrace economic growth, we should also decide to embrace AI?
When we think about economic growth, it depends on a return to the Enlightenment ideals of human happiness and prosperity. Economic growth has enabled more people in the world to enjoy a better standard of living. And for me, that is the goal, and in that context economic growth plays a role, and in that context technology plays a role.
Other must-reads from TIME
