AI breakthroughs are bringing hope to cancer research and treatment
Editor’s note: Ruth Porat recently spoke at the American Society of Clinical Oncology’s Annual Meeting in Chicago. Below is an edited transcription of her remarks.
We are living in an extraordinary time for technology, with AI and its application across so many areas being the latest — and by far — the greatest development. Today, I want to share what we're seeing at Google and put some of the AI opportunities into context — in particular to support the diagnosis, treatment, care and cure of cancer.
Throughout history, there have been a limited number of technologies that change everything. Economists call them “general purpose technologies” — they can affect an entire economy at a national or global level. They can alter societies through their impact on economic and social structures. General purpose technologies are very rare: They include the steam engine, electricity, the internet and now AI. Their power isn’t in the initial invention: it’s in the applications it makes possible when, to use another word from the economists, there’s “diffusion” — it spreads across all industries, throughout the economy.
Much like the steam engine — which was initially developed to pump water out of mines — it demonstrated its full transformative power only when it was put to use powering ships, trains and factories.
AI is one of those technologies. As a general purpose technology, it has tremendous potential across four broad areas: First is driving economic growth. Economists estimate that if AI is applied across industries, we could collectively have an uplift to global GDP of about $20 trillion by 2030.
The remaining three areas are highly related to health, and that's what I'm going to spend time covering today: accelerating scientific breakthroughs, supporting better delivery and outcomes and strengthening cybersecurity.
Making cancer “manageable”
I’m coming to you today first as someone who’s privileged to work at Google with some of the planet’s most brilliant engineers, scientists and researchers. I’m a finance person — they humor me (mostly because they must). I recently spoke with one of my colleagues, a gentleman by the name of Vint Cerf. If you use any of the many search engines out there to learn more about him, you'll see that Vint is recognized as the father of the internet.
I asked the father of the internet for his perspective on the impact of AI relative to the internet. And he said he believes AI has greater potential than the internet, in his words, “because it can augment human capabilities.” It is a partner for all of you, for each of us.
I’m also coming to you as someone who has had cancer — twice. When I was first diagnosed, my children were very young: 5, 7 and 9 years old. For me, like many of your patients, everything was fine — until it wasn’t. I went in for my typical annual mammogram and found out I had breast cancer. And then the journey began, resulting in a double mastectomy and chemo.
The first thing I felt was fear — in particular, fear that I wouldn’t see them grow up, see who they would become, be there for graduations and so much more — as well as fear about the process of treatment.
Two years later, I was diagnosed with cancer again. That time was even scarier, because I thought I had done all I could possibly have done. And then there was more chemo, more surgery and this time, also radiation.
What I learned through the process: As difficult as my cancer was, for me it was manageable. And I learned that that word “manageable” was my friend.
I know I’m one of the privileged ones thanks to the extraordinary care I received at Memorial Sloan Kettering from people like my exceptional oncologist then, and my good friend now, Dr. Cliff Hudis and now ASCO’s CEO. I’m so grateful for everything — every step on the journey he took with me. And of course, luck. By coincidence Memorial Sloan Kettering was founded as New York Cancer Hospital on this date, May 31, in 1884.
Cancer is not yet manageable for all patients, and certainly not yet for all cancers. Even as AI is helping to make cancer “manageable” a reality for more people — which is, in and of itself, great progress — the ultimate goal, of course, is to look beyond manageable to preventable, and curable.
Today I want to take you through the areas of opportunity where I hope AI can help achieve just that, and move all of us closer to the goal best captured in ASCO’s great mission statement: “Conquering cancer through research, education, and promotion of the highest quality patient care.”
I'll be the first to say that AI is not a panacea, but I hope and believe that technology can support the work that you are leading. Looking around the room, I’m humbled to be with so many dedicated professionals who bring that mission to life each day.
Accelerating scientific breakthroughs
Returning to the areas where AI has potential as a general purpose technology, I’ll start first with accelerating scientific breakthroughs. At Alphabet, we're focused on two key areas in this category: drug discovery and early detection of disease..
Hopefully, you are already familiar with AlphaFold. I'll tell you, it is a joy to work with Nobel Prize winners – Demis Hassabis and John Jumper – my colleagues who were recognized this past fall for AlphaFold, which solved a decades-long problem of protein folding. As many of you know, until recently, mapping the structure of just a single protein used to take years of painstaking work — there are more than 200 million known proteins. At that pace, it would've taken humanity hundreds of thousands of years to map them all.
With AlphaFold, this was achieved in months, not years — and it has been called the most significant advance in biology in decades. The hope is that AlphaFold truly helps accelerate cancer research by helping scientists understand how mutations alter protein function, which means faster insights into how cancers form, and how to design drugs that disrupt the harmful protein interactions driving tumor growth. We also chose to open source it, which means releasing this to the scientific community in a freely accessible database. We're really proud that today more than 2.5 million scientists in more than 190 countries have used AlphaFold.
Helping with diagnosis and early prevention of disease
The second area that's benefiting from AI is work to help with diagnosis and early detection of disease. Clearly, as you all know so well, early detection is key to improve move to outcomes. I know the best thing each of us can do and can inspire everyone, all our loved ones, to do is to get that early test.
But the quality of those tests is also imperative. Google has pioneered using AI to identify tiny clusters of cancer cells that have spread to the lymph nodes. We developed a deep learning model trained to spot cancer on gigapixel pathology slides, including cases where it might be missed, in particular when reviewed under time constraints. This cut the time it took for pathologists to review slides in half and helps them detect small metastases with far greater accuracy.
The best results come when humans and AI work together; that outperforms both the pathologist who worked without AI and the algorithm that works without the pathologist. Since then, we've seen how this kind of pattern recognition can scale. AI is now helping specialists like radiologists, acting as a second set of eyes and helping them to move through thousands of scans faster, which can ultimately save lives if early detection is paired with more widespread delivery of treatment.
As I frequently do when trying to understand the significance of some of our AI work, I called my oncologist Dr. Hudis and asked him what he thought. He said: “AI is a critical part of democratizing healthcare, so that everyone everywhere can have access to the best insights.” That's what inspires us — this concept of democratizing healthcare.
We're already seeing at Google how early detection, benefiting from AI, can scale rapidly to make a huge difference. We're seeing this in work that we're doing with AI applied to early detection of diabetic retinopathy blindness from diabetes. Google has already carried out more than 700,000 scans working with our partners in places like Southeast Asia and India. Our goal is to scale that many multiples, to 6 million+ over the next decade, which results in the early intervention that saves eyesight. This points the way to the benefit of AI for interventions for so many diseases.
The next category is supporting delivery and outcomes. There already are a large number of delivery solutions being piloted and implemented. First, the world is seeing the emergence of what is called agentic AI. Agentic AI systems combine the intelligence of advanced AI with access to tools so they can understand, they can reason, and they can act across complex workflows. They can take actions on your behalf and under your control. Agents are being created to do everything from tracking the latest cancer trials, surfacing what's relevant to a specific patient, auto-drafting prior authorization paperwork so doctors can spend less time at their desks and more time with their patients.
You've already heard about one powerful example: Today, Google is so proud of our collaboration with ASCO on the new ASCO guidelines assistant. As Google Cloud CEO, Thomas Kurian put it, think of it as providing clinicians with an AI-powered assistant that “never tires, always learns and can sift through mountains of data in seconds.” What we're hearing from you is that it dramatically reduces the cognitive load of reading long pages of dense content, those 80- to 90-page documents by surfacing clear structured answers almost instantly.
Knowledge alone does not conquer cancer, it must be applied. And working with ASCO, this is knowledge applied.
Second — I don't need to tell you this — but medical professionals carry a heavy administrative workload. Doctors can spend a third of their time on paperwork. Clinicians report losing 28 hours each week due to administrative tasks. Nurses are carrying a substantial administrative time burden with substantial time lost during shift changes and more.
This is a problem in and of itself, and in particular because in the U.S. and around the globe, there is a meaningful shortage of healthcare professionals. This shortage will get worse as the global population continues to age. Generative AI tools have shown the potential to reduce documentation time and the administrative burden on medical professionals. It has the ability to handle tasks like summarizing medical exams, test results, scheduling appointments, everything you see on this slide here, including billing support, real-time patient outreach. Looking ahead, the hope is these tools will support clinical trials and more.
Now obviously, your work is not just text-based. You don't just look at a CT scan or a chart. You interpret not only patient records, but tone of voice, symptoms and many other inputs. In the tech world, this is called “multimodal.” Now, AI can process these multimodal inputs ranging from audio to high-resolution radiology and pathology images, and even to the molecular level with genomics, which can help you do your work with greater precision and efficiency. And as Dr. Hudis beautifully said to me when I checked in with him on this one too, that gives back time for “greater humanity in the doctor-patient relationship.” What we're seeing is that it can give medical professionals valuable time back. Doctors are seeing a 30% improved efficiency for documenting patient explanations. Nurses are freeing up 40% of their time for discharge reports -- both allowing more time spent with the patient. That is the gift of time.
Strengthening cybersecurity
The last category is strengthening cybersecurity. I've gone through how AI can help with early detection of data and pattern detection and disease. The same is true in understanding data to fortify you and your organization against cyberattacks, monitoring new threats, often with very limited data, and staying ahead of problems.
The reality is, if you are running a hospital today, you are also managing one of the most targeted digital environments on the planet. Last year, healthcare data breaches reached an all time high. Hundreds of millions of records were compromised, affecting more than 80% of the U.S. population. The frightening reality here is that these attacks are not slowing down. They are accelerating. Because healthcare data is so sensitive and valuable, negotiating with attackers can translate into lost time, and lost lives. Your organizations, understandably, prioritize speed of resolution unlike many other businesses. It's therefore imperative to ensure privacy and security begin by design. And that implies leveraging technology to build a strong cloud environment to protect against unwanted data intrusion with AI applied to infrastructure controls. Each one of us -- individuals, organizations and the public sector -- have to remain vigilant to this increasing threat. AI enables this early detection and fortification.
Getting started — and bringing hope — with new technology
Today I've shared with you only a small fraction of tools that can provide an uplift: accelerating scientific breakthroughs, supporting better delivery and outcomes and strengthening cybersecurity. I didn't even go into all of the consumer applications that enable each of us to have technology at our fingertips. Based on conversations I've had (and a lot of survey data!), I'm assuming that a number of you are wondering how to even get started engaging with the technology. It is important for each one of us to be at that forefront of change because the benefits are compounding, the pace is accelerating. I'm going to share with you a story I keep coming back to that really speaks to the speed of change. It's a story about Google Translate.
What people don't realize is we started on this journey to translate and offer Google translation 20 years ago. We now translate 250 languages, but in the last nine months we added 116 of those 250 languages. The benefit of AI is like a vertical lift, and you want to be at the front end of that lift — not chasing it from behind.
One of the beauties of AI solutions is they really should be quite intuitive. As a starting point, try the ASCO guidelines assistant, because a key to AI is learning how to pose a query to trigger AI to work for you. You don't need to master the technology all at once. The key is to build familiarity through real use, whether through the ASCO tool, or you can even just try AI Mode in Search. I call it “getting the puck on the ice” — just starting that journey puts you on a path to catch the uplift.
In closing, I stand here today grateful. Everything I feared I would not see, I have. Every birthday, every graduation. And now I want to share why I chose to call this session, “Why Not?” I'm going to return to my colleague Demis Hassabis. When Demis and his team embarked on that journey to predict all 200 million protein structures known to humanity, one scientist said to him, “how is this even possible?”
He responded: “why not?” And then he went on to win the Nobel Prize with his colleague John Jumper. So those two words — why not? — inspire me.
Why not imagine a world where earlier detection is available everywhere?
Why not imagine a world where the best care is not the exception, but it is the norm?
Why not bring the word “manageable” and the word “cured” to millions of people?
When I was going through cancer the second time, another cancer survivor shared a story with me that was deeply moving and motivating, and I share it every time someone shares with me that they have been diagnosed. It is a story about a bird. What my friend told me is that as a cancer patient, as many of you know, you have good days and you have bad days.
He said, “On those good days, this evil bird invariably will show up on your shoulder and say, ‘Why are you happy? You have cancer.’” But he promised me that eventually that bird would come less frequently. And he promised me that, finally, that bird would never come back. I found the story empowering, because even when that evil bird showed up, I learned to tell him to get lost. Even though I told myself the bird would one day go away, every time the bird showed up, it was scary.
In conclusion, why not imagine the day that bird never shows up? No first time. Never.
That is my dream.
Thank you for all the patients you have helped, and the many more you are helping and are going to continue to help.
Thank you.
