James Manyika from Google: AI Will Not Destroy Jobs

Optimism in the Face of Automation: A Nuanced View of AI and Employment

Recently, I had the opportunity to discuss with Aaron Levie, CEO of Box, who presented an interesting argument: jobs are more complex to automate than AI companies suggest. According to Levie, the human aspects of work, such as judgment and context, will not be replaced anytime soon. This perspective left me optimistic about the future of employment.

This week, I wanted to confront this optimism with the views of James Manyika, Senior Vice President at Google and Alphabet. Manyika leads Google's research operations and labs, as well as a team dedicated to studying the societal impacts of technologies, including AI. His role involves assessing the consequences of AI systems like Gemini and shaping Google's strategy accordingly.

It is crucial to note that seven out of ten Americans now oppose the construction of data centers in their communities. The message conveyed by the tech industry, suggesting that AI might first replace jobs before threatening humanity, has failed to convince the public. In this context, Manyika adopts a more optimistic approach. He believes that jobs are harder to automate than Silicon Valley claims, and that the process will be slower than the most pessimistic forecasts. This opinion is shared by some of his colleagues at Google, such as Demis Hassabis, CEO of DeepMind, who recently warned against the replacement of software developers by AI tools.

Unlike many tech leaders, Manyika has formed his opinions over a career outside Silicon Valley. A former executive at McKinsey, he co-authored a report on the potential effects of automation on employment nearly a decade ago. He also co-chaired the UN Secretary-General's High-Level Advisory Group on AI and served as Vice Chair of the National AI Advisory Committee under President Biden.

In the face of alarmist predictions from some competing companies, like Microsoft and Anthropic, which foresee a massive disappearance of white-collar jobs, Manyika remains skeptical. "Some of these predictions were made two years ago, claiming that 50% of jobs would be eliminated," he explains. "Two years have passed, and those predictions have not come true."

Highlights from Our Conversation

The highlights of our conversation are below, edited for clarity and conciseness. We also hope you will listen to the full conversation wherever you get your podcasts — just search for Platformer — or watch it on YouTube at youtube.com/caseynewton.

Casey Newton: You earned your PhD in AI and robotics at Oxford decades before AI became the biggest news topic in the world. What did you believe or see at the time that most people missed?

James Manyika: I'll take you back even further. I did my undergraduate degree at the University of Zimbabwe, and my undergraduate thesis was actually the first paper I ever published. Guess what it was about? AI — training and modeling an artificial neural network. There was a visiting postdoc from Canada who had worked with Geoff Hinton's team in Montreal-Toronto, and he suggested I build a neural network for my undergraduate thesis. That was the very first thing I published, in 1993.

Newton: Long before people like me spent every waking hour reading and writing about this. What captured your interest?

Manyika: Two things. I grew up with Star Trek, so the idea of AI fascinated me. I watched 2001: A Space Odyssey. But I was also intrigued by the idea that it would be possible to build systems capable of performing advanced cognitive tasks. So, when I went to Oxford, I pursued a PhD in AI and robotics to continue exploring that.

Newton: Since then, you've spent a good part of your career trying to measure how technologies change economies. You spent a long time at McKinsey, where you wrote a paper titled "Jobs Lost, Jobs Gained" nearly a decade ago. Now that you're at Google, where you can see what happens when these tools actually land in the workplace. When you look at the debate we are currently having about the future of AI and jobs, where do you stand?

Manyika: This is such an exciting moment. Technology and its capabilities are expanding at an incredible pace. But when you try to translate that into what it might mean for work and jobs, I have a very mixed view. It's pretty much what that paper said ten years ago, which I still think is correct: there will be jobs that decline, there will be jobs that grow, and most importantly — a third aspect — many more jobs will change.

Whether you look at the global economy, at the sector level, or by profession, you get a different mix of these three things. But all three will happen. Research hasn't changed much. The debate people have is: what is the mix of these three things? Rather than asking whether these three things will happen.

Newton: Let me name a dynamic that might concern some listeners. You are now employed by one of the biggest beneficiaries of the current wave of AI. How do you distinguish between the labor economist in your head and the Senior Vice President at Google?

Manyika: I hear both things. Less the Senior Vice President at Google — more the AI researcher and computer scientist in me is extraordinarily excited about the pace of technology. That part of me thinks: "Oh my God, this is going to be extraordinary, and it's going to happen very, very quickly."

The labor economist part of me says: "Wait a second — these things don't unfold as quickly in the economy, and the dynamics are more mixed." So, I hear almost two speeds unfolding here. I often think that as AI researchers, our community tends to exaggerate what is happening in labor markets based on what we see at the technological frontiers. These are two very different conversations.

Newton: At the McKinsey Global Institute, you found that about 50% of tasks could be automated through AI, but only about 10% of professions would be fully automatable. A few generations of AI later, does that ratio still seem correct to you?

Manyika: All the elements have evolved. At the task level, many more tasks are now possible to automate — that picture has evolved quite rapidly. But if you look at the composition of professions — the Bureau of Labor Statistics tracks somewhere between 850 and 1000 actual professions — and you ask how many existing professions have the majority, let's call it 90%, of their constituent tasks automatable, that number is still below 10%. Most researchers would still say that.

How many tasks seem difficult to automate? Partly because AI can't yet perform them, or due to coupled tasks where the weak link slows down the combination. If you take two tasks and can automate one of them, but they need to be performed together, you will only go as fast as the weakest link. Most jobs have these couplings that make complete automation very difficult.

Another thing that has evolved is the duration of tasks. If you had asked in 2017, among the tasks possible to automate, some were very short — 30 seconds or a minute is about the maximum duration you could reliably predict a task being performed in an automated way. Now, we can perform some of these tasks for over four hours. The duration of tasks with reasonably predictable completion has made huge strides.

Newton: So, what I hear is that if you measured the tasks that are automatable now, that number is trending much higher than 50%. But at the same time, the number of jobs you could fully automate remains stubbornly in the same place as it was ten years ago. What is your best explanation for this gap?

Manyika: Part of the gap is that we now understand more fully that entire jobs have a much more complex mix of tasks, and this idea of weak links or coupled tasks is very important in most professions. If you look across the entire economy, for most complex tasks, we cannot automate most of them. So, the question of which entire jobs you can automate to more than 90% is still a relatively small number.

Most of the debate among labor economists is whether in the next decade that number is more like 2 or 3% or more like 9 or 10%. I don't think anyone who has looked at the automation of entire jobs would say it's 50% or one of those extraordinarily high numbers.

That's why I come back to the idea that three things will happen. Yes, there will be job declines. But there will also be jobs that grow — this is a function of existing jobs that increase in demand because technology often changes the landscape of demand, and new jobs are created. We forget that David Autor and others have shown that if you go back to 1945 and compare it to today, something like more than 60% of the jobs we have in the economy today did not exist back then, and many were introduced as a result of technological changes that created the category.

But the most important effect is that the very nature of work is changing.