AI is about maximising human potential, says Google DeepMind COO: ‘If I could rebrand AI, I would’

From banking to healthcare and entertainment, artificial intelligence (AI) is rapidly transforming almost every aspect of life. At the same time, there are many who fear its impact on jobs, privacy and safety.

In a decade shaped by AI, how should we understand this powerful emerging technology?

That was the question I posed to Lila Ibrahim, the chief operating officer (COO) of Google DeepMind, a research laboratory based in London that builds Google’s next-generation AI system, when she was in Singapore in September.   

Ibrahim’s answer surprised me. Instead of talking tech, she told me a deeply personal story.

“I grew up as the eldest daughter to parents who had immigrated to the United States. Growing up, I was the dark kid in my school, the foreigner. English was my second language… My entire childhood, I was always an outsider,” said Ibrahim.

Her dad grew up in a Lebanese orphanage. His future should have been over, she said. Instead, he overcame a disadvantaged childhood “to design heart pacemakers that saved millions of lives because he had an engineering degree”.

“That’s what technology has always meant to me,” said the 55-year-old tech leader. “The reason I went into engineering is, I felt, how can I use technology to solve problems?”

WHAT IS AI, REALLY?

“Actually, if I could rebrand artificial intelligence, I would,” she laughed.

“I’d like to think of it as moving away from the name ‘artificial intelligence’. Sometimes, we get caught up with trying to define it, when instead we should say, how are we going to deliver the value?”

“If you are watching a video on YouTube and the next recommendation comes up, that’s artificial intelligence. If you have an Android device and the battery lasts longer, there’s artificial intelligence in there optimising the battery. 

“When things have value, we don’t call it artificial intelligence. It just is part of the experience,” she said.

“Even if you think of organising the world’s information – Google and search – that’s all about artificial intelligence, trying to unlock knowledge and make value for consumers and customers.”

WILL GENERATIVE AI REPLACE HUMANS?

In December 2023, Google took its AI technology to the next level. It launched Gemini, which the company says is the world’s first multimodal AI system able to analyse and generate text, images, video, audio and code.

Its more advanced features require an AI premium plan. For example, “I can put in a picture, a video, audio… an entire book like War And Peace, and then interact with the AI in a way that is like a conversation,” said Ibrahim.

For instance, AI can help to analyse and summarise the huge chunks of information from various media and answer any specific questions you may have.

Such AI advancements will no doubt change jobs, but it will not replace humans, she said.

Ibrahim drew a parallel to the time when computers were first brought into the classroom. “What we found was students wanted to come to class. Teachers were able to facilitate the class in a different direction. People could move at their own pace. So everybody got a different path to learning,” she said.

AI is very similar – it is about “getting the right information at the right time to help people make better decisions in more productive and efficient ways”.

“Maybe you would have gotten to that same thing but taken many different steps to get there. You also had to have the right network and resources to do it,” she said.

By giving you some of that input very quickly, AI supports self-exploration so that people can build upon the input, explained Ibrahim.

That is how AI can help you “increase your own potential”, she said.

HOW DOES AI ADVANCE SCIENCE?

“AI is so much more than generative AI,” said Ibrahim.

“When DeepMind was founded, the whole concept was, can we help machines learn in a way that they can uncover mysteries of the universe that humans haven’t discovered yet?” she said. DeepMind was founded in 2010 and acquired by Google in 2014.

A lot of Google DeepMind’s efforts are focused on science. In 2020, it launched AlphaFold2, a breakthrough in protein structure prediction used by millions of researchers globally in areas such as malaria vaccines, and Parkinson’s and Alzheimer’s therapeutics.

Proteins first exist as linear chains before folding into three-dimensional structures, necessary for them to perform their biological function, such as providing structural support as collagen, transporting oxygen as haemoglobin, or enabling muscle movement.

“Proteins are the basic building block of life,” said Ibrahim. “If we can predict how [they] might fold, we can understand the function. If it misfolds, you can understand what happened [such as in the case of neurodegenerative] diseases.”

In May this year, a new version, AlphaFold3, was launched. Beyond predicting the three-dimensional structure of proteins, AlphaFold3 also predicts the interaction of proteins, DNA, RNA and other small molecules to further scientific research.

In addition, Ibrahim’s team also launched a state-of-the-art weather forecasting AI model, GraphCast, in November 2023. Trained on decades of historical weather data, it provides faster, more accurate weather forecasting than traditional models. It is also cheaper – research indicates it can be up to 1,000 times more energy-efficient.

In light of more extreme weather events today, this enables the world to be more prepared, Ibrahim said.

WILL AI WIDEN OR NARROW THE GAP?

Providing more equitable access is something close to Ibrahim’s heart.

In addition to describing her father’s disadvantaged childhood, Ibrahim told CNA Women that her sister has cerebral palsy.

“She’s been in a wheelchair her whole life. The typical way of working in the 1970s is that she would have to go to a special school. My parents fought the state that they lived in [in the US] to make sure that my sister got access to a quality education,” she said.

“My sister ended up getting two master’s degrees, and is now a senior policy advisor,” she added, with steady pride.

It is an example of “where human potential was taken to a different level than what maybe society expected at the time”.

This belief in human potential inspires her to continue to democratise AI. For instance, AlphaFold is freely available for non-commercial usage such as academic research.

“In the Global South, there are a lot of diseases which pharmaceutical companies have not invested in because there’s not a good commercial outcome,” she said.

The Global South refers to countries mainly in the southern hemisphere, such as in Africa, Latin America and Asia that are less economically developed. Neglected diseases that AlphaFold has been used in research include dengue fever and other parasitic diseases.

Additionally, Ibrahim’s team developed LearnLM, models for learning and education based on pedagogical best practices such as stimulating curiosity and adapting to the learners’ goals and needs.

HOW CAN PEOPLE BENEFIT FROM AI?

“Such transformational technology requires exceptional care,” said Ibrahim.

Having witnessed major technology transitions such as the internet, mobile phones and social media, one of her takeaways is: “We didn’t know how people would be using it… but we had time to learn,” she said.

“Now, we’re in the AI era, and things are happening very quickly. When a model is released, it’s available to billions of people on their phones,” she said, stressing that it is now all the more crucial to minimise mistakes.

“We have to be thinking, how do we make sure people know how to use it responsibly, what to expect, what not to expect?” she said.

Google DeepMind’s work around responsible AI includes ethics research, frontier AI safety research and socio-technical research to minimise misinformation, misuse, and ensure alignment with cultural values in local communities.

They have hosted roundtables with people with disabilities, teachers, artists and other groups within communities, integrating their needs and concerns into the development of the technology.

In a future world that is AI-enabled and AI-powered, what sort of mindset shift is necessary to thrive?

“My recommendation would be to get curious. When you put a prompt [into Gemini] maybe ask three more questions. Don’t get lazy. Actually, get more curious,” Ibrahim suggested.

As a mother of 14-year-old twins, she has seen her own daughters harness the power of AI for their unique needs.

“I have one daughter who is dyslexic. She’s smart, but she learns in a different way. Right now, she’s able to have more of a conversation with AI. And instead of typing or writing, she can take a picture of a math problem. It doesn’t give her the answer. It helps her kind of step through it.

“My other daughter is into poetry. She puts something in, she gets some ideas, and then she writes something,” she said. 

“It’s been very interesting to see how this generation is growing up. They’re not questioning their creativity. In fact, they see [AI] as almost turbocharging their creativity,” Ibrahim reflected.