How AI Will Revolutionize Medicine and the Arts Over the Next Decade and Beyond
In the coming decade, AI will bring incredibly far reaching changes to movies, music, TV, medicine, and healthcare and almost nobody sees…
In the coming decade, AI will bring incredibly far reaching changes to movies, music, TV, medicine, and healthcare and almost nobody sees it coming.
There’s too much darkness in the world today.
Whether it’s a pandemic raging out of control, surging civil unrest, a crashing economy or the retreat of democracy, life is starting to feel like the mean streets of rain soaked cyberpunk novel.
Whenever I write or talk, I try to give a balanced view of where we are and where we’re going but today doesn’t feel very balanced. At times it’s easy to feel like there’s no good solutions to any of our problems and we’re nothing but a tiny, insignificant drop in a massive ocean.
But I’ve had enough of the darkness.
That’s why today, I’m going to focus only on the good AI will do in the world.
If you want to read about the bad things it will do then you can go scare yourself with breathless articles about AI taking over the world anywhere else on the Internet.
In the past I’ve talked about where AI is going in 5, 50 and 500 years. But right now I’m going to laser on where we’re all going in the next decade.
AI is the ultimate technology. It’s the one technology that changes everything, the one that accelerates every other technology on the planet.
AI can and will design new materials, change way we learn, how we fight, how we interact with everything around us, how we run the world and societies, how we create art and heal the sick.
In short, it will touch every single aspect of human existence.
If the last several thousand years were about mass scale of humans, mass industrialization, AI will accelerate that exponential growth like nothing else before it.
That’s because intelligence is at the heart of all our innovations. It’s what got us where we are today, living in great cities and traveling the world and building skyscrapers that stretch into the sky. Intelligence and ingenuity arethe reason you can stream this article on your phone right now. It’s the reason you can chat with your friends and family at any time, anywhere.
Intelligence is the one thing that makes us special and different from every other creature on the planet.
Humans are a young race. We’ve only been around a few million years and for most of that time we didn’t change all that much. We ran around in the forest, hunting and gathering for most of that time. Up to 90% of the people in the world for 95% of history worked jobs that revolved around getting basic food needs met.
But it was intelligence that changed all that.
Ancient peoples looked at the problem and figured it out in the agricultural revolution. That wasn’t a materials or machinery revolution, it was breakthrough in how we think. The agricultural revolution is only 12,000 years old. But that’s just the first agricultural revolution where we learned how to grow crops rather than chase them around. The real agricultural revolution came a few hundred years ago in Britain and it gave us one critical insight:
Crop rotation.
Before that, growing food was highly inefficient. We couldn’t grow food on every plot of land all at once because it exhausted the soil. In olden times we used to leave half of the land fallow, meaning empty.
In the Roman world half a plot of land grew nothing for a year and then after the harvest, farmers would rotate crops to their second plot of empty land. But farmers in England and Europe figured out how to grow different kinds of plants on every inch of land in a constant rotation that renewed the soil at every step.
That brilliant breakthroughs of crop rotation and nitrate production led to a population explosion, andless and less people needed to farm and could turn their interests to other things.
And turn to other things they did.
Today only 26.7% of the world makes it living through agriculture.
People moved to other professions. Science. Industry. Research. That lead to more and faster revolutions.
The scientific revolution is only 400 years old and the industrial revolution just 200 years. Now we’re living in the information revolution, a mere 50 years old. It’s getting faster and faster as each innovation builds on the last.
The next decade marks the beginning of the Intelligence revolution, the age of AI.
AI will become our doctor, a most intimate friend, our interface to the world. We’ll talk to it, ask it where to go, and when we’re sick we’ll ask it what to do. It will organize our lives and our societies. It will track the spread of disease and optimize how roads and cities are built. It will show us insights we can’t see ourselves, buried deep in the data.
And of course, AI will do both good and bad things: surveillance, war, weapons. But enough has been written about the bad things.
Today we’re going to talk only about the amazing things it will do and it’s already starting now.
The Machine Learning Platform of Tomorrow
Before I talk about AI for good, let’s talk about how we get to the good AI apps of tomorrow.
For all the breathtaking headlines and breakthroughs of the last decade, today’s AI software is still bleeding age. The infrastructure tools in AI/ML are rough and ever evolving.
AI remains the product of big tech companies and companies will billions of dollars in budget that can afford an R&D department or it’s the province of small, agile, super cutting edge teams, like at Stitch Fix, where data science powers every aspect of the business.
But it’s still in the early adopter phase for everyone else.
Big tech companies like Google have an army of coders who’ve build a general purpose cloud operating system for Google to run their web of data centers. When a new technology develops, they can easily weave it in, by updating and adapting their software or inventing new software to manage it. AI is just another cog in the machine to them.
But to break out of the hands of Big Tech, we’ll need powerful, open source platforms that put AI and ML into the hands of everyone else.
We need the Linux of AI.
Linux gave us a robust, baseline platform that powers all kinds of other innovations. When I started with Linux a few decades ago it was a messy, difficult to use, hobbyist system.
When I started out in software, open source was a rebellion.
Early open source was a cheap knock off of the proprietary way of building software behind closed doors. When I joined Red Hat a decade ago, we were still trying to sell it to enterprises who thought it might be a path to communism and the death of commercial software as we knew it. They thought it would die a quick death.
Instead, Linux dominates everything now, from the public cloud to embedded devices. Even Microsoft, once the dominate proprietary software provider, now builds clouds, video games and uses Linux to power their own machine learning efforts.
Of course, it’s too early to see a single platform to rule them all in AI/ML. The systems are too complex and evolving too quickly to get pinned down to a basic monolithic, integrated architecture yet.
Instead, we’ll get a suite of software, a stack, the MLCS, Machine Learning Canonical Stack, that weaves together to do everything we need it to do, the way Kubernetes and Docker came together to power the cloud or the LAMP and MEAN stacks came together to power modern web apps.
Today, the Canonical Stack (CS) in Machine Learning is still evolving.
A swarm of venture backed, open source companies are working to create general purpose software that will power AI innovation for the rest of us. They’re making in-roads at different layers of tomorrow’s infrastructure. They’ll battle it out and compete and consolidate. Some of them will emerge as titans and some will get gobbled up by the eventual winners in the space.
Once we have a clear winner in this area, that stack will form the bedrock of tomorrow’s AI innovation. When a CS forms it lets developers move “up the stack” to solve more interesting problems.
Over the last few decades we’ve seen traditional software development reach dizzying new heights as better and better stacks emerged. It once took a small army of developers to write a database with an ugly interface that could serve a few thousand corporate users in the 1980s and 1990s.
It took only 35 engineers to reach 450 million users with WhatsApp.
The same will happen in AI.
Data scientists will rocket up the stack to solve more interesting problems.
We won’t have engineers messing around at the lowest levels, labeling data by hand, transforming and loading data, cleaning it and massaging it and testing it over and over.
All of that will get rapidly automated over the next ten years.
ML will label 99% of data and humans will check it. Scientists will start with a bevy of ready made approaches to attack their problems. If they don’t have a data set they’ll pull from a library of encrypted, anonymized ones or they’ll synthesize one. They’ll drag and drop concepts together with beautiful interfaces or they’ll just tell the computer what they need to it do and it will understand. They won’t be using notebooks to create new systems. That will look primitive, as AI moves to the interface itself and bubbles up promising pathways and models and algorithms to pursue. It will help scientists evolve newer and better algorithms at each step and refine the ones that already work.
At the same time we’ll continue to see faster and faster chips, driven by video games and the rise of deep learning. Those chips will start as co-processors in our devices, or add in cards and quickly get woven into the general purpose architecture of every smart phone and game console and sensor.
All of this will tear AI/ML out of the hands of big tech and democratize it.
It will deliver a Cambrian explosion of AI apps big and small.
Fortune 500 companies and then smaller and smaller companies will unleash the power of AI in their own environments, building on the back of the Canonical Stack. The innovations will cascade down to smaller and smaller companies and then citizen level innovations in machine learning.
The solutions to various problems in AI, like how to spot objects in video or how to do sentiment analysis to see if your customers are angry when they call the support line, will become fully baked. They’ll be vast libraries of pre-trained models that every-day programmers can roll into their apps with ease. It’s already started to happen with libraries like Hugging Face that delivers pre-trained transformer models.
Companies won’t need an R&D department of data scientists just to get their machine learning practice off the ground.
Every business will start to use AI more and more. In the beginning of the Internet revolution, companies debated whether you needed a web page. Now nobody doesn’t have a web page and two or three social media accounts and people to manage them.
Tomorrow it won’t be a debate about whether you should use AI.
You’ll only ask where you need to use AI more.
AI will leave no industry untouched. There’s not a single business on Earth that won’t benefit from more intelligence.
Agriculture. Finance. Medicine. Defense. Security. Retail. Telecom.
AI will manage stores and supply chains. Why would you have people guessing whether they need to order more umbrellas, when AI can follow the weather and know a surge of winter storms are coming?
The machines will discover promising ways to attack viruses and diseases. They help us forge new cures and train our immune systems. They’re re-purpose old drugs to fight new diseases.
Your cars will pick you up and drop you off and there won’t be a driver.
You’ll walk into a store and walk out with what you need and the AI will know what you took and charge you.
AI driven drones will soar over fields and watch for pest damage and signal the systems to water more or dial it back.
The machines will watch our hives and call bee keepers when they spot a swarm of hornets coming to attack that precious yellow gold we call honey.
Tomorrow AI will finish your emails and suggest presents for your friends for their birthdays when you forget them. It will act as your personal assistant, scheduling meetings and figuring out a place to take your date to dinner.
We’ll also see an explosion of citizen AI projects, open source and ever evolving. Big tech and governments won’t just watch everyone, people will watch companies and the government right back. The camera a small bar owner installs won’t just take pictures of someone getting attacked outside a bar. It will ID the attackers, call an ambulance and call the police.
People’s cameras will do facial recognition on officials who abuse their authority and their power. Your glasses and the cameras in your car will record when you get pulled over and make sure nothing gets out of line.
And all of that will get built on the back of AI Infrastructure stack that makes it easier than ever to build the apps of tomorrow.
Some of them will be good and some of them will be bad, because AI is a dual use technology. But like I said earlier I don’t want to spend any more time thinking about what can go wrong.
Today I’m focused completely on how AI will change the world for the better and nowhere is that more obvious than in AI for the arts and in healthcare.
Arts
AI will radically change music, movies and TV in the coming decade. Already we’re seeing cutting edge research deliver breakthroughs and those breakthroughs will build on each other in the coming years.
There’s a fantastic, underrated sci-fi novel called Remake, by Connie Willis, about a young actress trying to make it in Hollywood, a classic coming of age story that every young actor or actress faces as they try to make it in the movie biz.
The just one little problem.
She’s trying to make it in a Hollywood where there aren’t anymore real actors.
The studios simply whip up any actor they want digitally and use AI put them into the latest blockbusters. They favor the iconic old stars of the silver screen like Humphrey Bogart and Marilyn Monroe and there’s no room for actual actors anymore.
Of course, the death of acting is premature, but AI will make it possible to recreate people much more easily in games and movies. Animators will have less to do, because deep learning algos will already have studied movements and expressions and recreate them on the fly.
Animators will tell it what emotions they want the characters to have and the AI will act it for us. The animator will tweak it to get it perfect or change a spot or two. Motion capture will gradually disappear and animators will have a repertoire of perfect movements to drag and drop into movies of tomorrow.
We’ve already seen advances in movies and video game graphics feed off each other and in the coming decade that convergence will accelerate. Video games have been pushing towards real-time photo realism my entire life, with each new generation getting closer and closer. Eventually they’ll succeed and there won’t be any more new game engines.
Why create a new engine when you can already recreate Newtonian physics and deliver graphics with perfect photo realism?
That’s when the technology will really spill over into the movie making world.
Eventually you’ll have movie studios in a box where animators and directors can quickly create whole movies by dragging and dropping stick figure like characters into scenes. Then they can layer any imaginary or real actor on top of those stick figures to create a brilliant new performance. AI will power those motions and expressions, the long sighs and bedroom eyes of Ava Gardner, that charismatic smile of Tom Cruise, and those convincing tears of Matthew McConaughey and Meryl Streep.
As this software gets better and better it will drive the price movies way down. Towards the end of the next decade a small studio will create a blockbuster movie with stunning special effects and AI driven digital actors and it won’t cost $400 million dollars, it will cost $20 million.
That may take more than a decade to really set in but we’ll see the roots of it start to take shape over the next ten years.
More and more of the actors you’ll see in the backgrounds will be digital creations, with AI doing most of the acting, the algorithms having learned about realistic movement from ingesting millions upon millions of hours of footage from the past.
Of course, I don’t see the AIs completely replacing all actors forever. Actors will evolve and we’ll probably always want to connect with real stars and see breakthrough performances that no machine could ever give us from studying the performances of the past. An actor can come up with something we’ve never experienced before when they get deep into their role, the way Heath Ledger did in The Dark Knight. We thought we’d seen all the possible interpretations of the Joker but he went somewhere nobody had ever imagined and created the way everyone pictures the Joker now.
But increasingly those actors will stand with totally digital creations. It will be a dance between AI and real life animators who tease the best performances of AI, the way directors tease performances out of actors now.
Actors will also see parts of their performance fixed or changed by AI. The algorithm could simply study a new performance and handle the reshoots instead of the actor needing to come back in for the shoot when the director wants to change something later.
In the music world, we’re already closer to fully realized digital singing stars and that will only accelerate in the coming years. The OpenAI Juke Box can already recreate Frank Sinatra, Elvis Presley and more. It’s not perfect but they’ve laid down the first dirt roads into the wild forests beyond today. As we get more neuro inspired chips, better algorithms and more time and money we’ll soon recreate the great singers of yesterday and today with ease.
It won’t be long before media companies have their own AI research teams.
Soon after that they won’t need research teams, they’ll just need traditional coders to layer the finished technology into mainstream apps anyone can use.
It won’t be experimental to recreate Sinatra, it will just be something you expect software to do right out of the box. You’ll have music editing apps already layered with the voices of the fantastic singing stars of old. You’ll write a song and try it out with Edith Piaf, Billie Holliday or Mariah Carrey.
My partner and I inherited her grandfather’s old record player and quickly rediscovered the incredibly warm sounds of old vinyl that’s too often missing in our infinite digital playlists. I had records when I was a kid and I’d forgotten how smooth and deep they can sound, almost like having the singer right in the room. It was a brand new experience for her because she’s in her 30s. I had to show her how to use it but she quickly fell in love with the old fashioned sound and we amassed a great collection quickly, as we hunted through the old record stores and flea markets in Berlin.
One of my favorites is the recently discovered Lost Berlin concert of Ella Fitzgerald in the 1960s. It’s an incredible 2 disk collection, filled with old favorites like Cry Me a River and Someone to Watch Over Me. The concert got recorded live and then the tapes went in a box as Ella and her team rushed off to more shows in more cities. They stayed in that box unopened for 58 years. It’s like having brand new music from the old master of jazz when you listen to it now.
But what if we didn’t have to find a lucky stash of old concerts anymore?
What if we could recreate Ella for the modern age?
What if we could create a song, tag the notes with “emotional” metatags, which tells the AI where to put emphasis and pathos?
What if we could tweak every single second of the song just the way we wanted it, drawing out the depth of Ella’s voice, making it sad and sweet, or soaring and majestic to fit the sweep of the song?
Or what if you could combine the incredible range of Ella’s voice with the melancholy of Amy Winehouse or Billie Holliday or Edith Piaf, three singers whose hard lives bled over into the sounds of the supreme voices? You can hear the joy in Elle’s voice as she skips over the beats and you can hear the sadness and heartache in Holliday’s voice as she sings of Strange Fruit or you can hear the defiance in Piaf’s voice as she sings Non, Je Ne Regrette Rien (No, I Regret Nothing). AI will give us the ability to blend the best voices, to mix and match them into voices that will soar.
We’ve already seen research into how to separate out instruments and voices from tracks. We can reverse engineer the tracks of old, isolate pieces of them and enhancing them and it wont be long before that’s a part of every music program on the planet.
Machines will become the next hit makers too. I did a Learning AI If You Suck at Math piece last year with a data scientist where we applied research from the Google Magenta team to make beautiful ambient music. It worked incredibly well and it will only get better in the next ten years, as our algorithms benefit from faster and faster parallel processors that can crunch smaller and smaller sections of songs and tease out their long term, deeply interconnected structure.
You won’t have to imagine it for long.
You’ll soon have an AI driven Photoshop for Music in your hands.
It will be the standard way people make music in the coming decades, adding background voices and crafting new stars out of thin air by combining voices.
If you’re working in AI now, don’t just spend time on your job, spend time in the arts.
Spend time on passions and fun.
There’s a lot of companies that will hire you right now to track their employees or optimize useless ads so they hit 3% more eyeballs but spend your free time exploring AI for the arts. At nights dig into the research and study it. Figure out what it’s missing and make it better. Unleash your creativity to create new kinds of algorithms and new ways to experience music.
Then one day you’ll wake up and find your passion project will make you a lot of money as media companies compete for your rare talent, a talent nobody thought they needed until now.
Of course, these new frontiers in music will also become the brave new battleground for intellectual property wars. Expect the big content owners, the distributors, the artists, and the tech companies to fight it out for who gets to control the voice of all the living and the dead.
We have no legal precedent for any of this and if the companies aren’t careful they’ll strangle a brilliant new technology before it gets off the ground.
But in the end it won’t matter. You can’t fight the future.
The record companies tried to stop piracy and force people to keep buying CDs but people didn’t want hard copies anymore. They wanted digital ones that put their entire collection right in their pocket. Napster gave it to them. The record companies had to adopt and now we all enjoy legitimate streaming services like Spotify that put the all the world’s music on our noise canceling headphones and Bluetooth speakers. We can listen to music anywhere and everywhere when ancient people might never have heard music or only have heard it a few times in their entire lives.
And the same will happen with AI voices and digital actors.
If Big Music and Big Movie/TV is selfish and stupid, open source developers will just release the AI driven music makers and actors for free and a revolution of new artists will create incredible songs and genres and beats and movies and TV they could never imagine now.
If Big Music and Big Movie/TV are wise, they’ll take a hands-off approach and let the new tech flourish so they can own the commercial software that makes it all possible to create the breakout digital stars of tomorrow.
Health Care
The second area I’m most excited about is AI for healthcare. AI will radically change healthcare in the coming decade and beyond.
It will speed up everything from drug discovery, to disease detection, to the way doctors interact with their patients, to how people get the care they need.
We’ll see machines crunching through drugs already on the market to find new uses against novel diseases. Algorithms will design novel compounds and new ways to attack viruses that researchers never dreamed of trying. It will detect cancer better than any radiologist and offer treatments that raise doctors to a whole new level of skill.
The fastest breakthroughs are coming in disease detection.
AI will swiftly bring the end of radiology as we know it in the next ten years.
Google researchers showed a 72% accuracy on detecting skin cancer with a pre-trained Inception V3 convolutional neural net (CNN) in 2017. By 2018 the best classifiers had rocket up to 85% on the ISBI open dermatological dataset. By 2020, the best in class systems had a 96% accuracy which is in line with the top radiologists on the planet.
The wisest radiologists in the world already see it coming. In the magazine Radiology Today, Robert Schier, MD, writes about a Google team’s algorithm detecting breast cancer better than today’s best radiologists. He knows exactly what it means for his profession:
“The appearance of this system marks the beginning of the end of the practice of diagnostic radiology.”
He was referring to CNN model that detects metastasized breast cancer from pathology images called the LYmph Node Assistant (LYNA), which achieved an incredible 99% success rate, compared to human doctors who sometimes scored as low as 38% on challenging slides.
In 2008, the FDA cleared a whopping 1 algorithm for use in medical imaging. By 2013, that dropped to zero. But in 2017 it was 4. By 2018, it was 18.
AI will be great for medicine. Patient care will get better, faster, less expensive care.
But as Schier writes, “AI will not end up being good for the specialty of radiology.”
The next big wave will be AI doctors and treatment advice. We won’t get the Star Trek robo-doc in the next decade but we’ll lay the foundation for machines that know how to detect and treat every day problems fast.
People will turn to their phone and their house/apartment AI and their smart glasses for advice before they ever see a doctor.
They’ll talk to the system and it will tell them to see a doctor or they have nothing to worry about. Listen to my cough. Check out this spot on my skin. Is my cholesterol too high?
Imagine a tiny little pin prick hook on your watch that you can’t even feel that will test your blood and give you constant feedback on what’s happening deep inside your body. You’ll see your blood pressure graphed out over time and your triglycerides and your cholesterol. The apps will get smarter and smarter and they’ll warn you days or weeks in advance that you need to see a heart doctor, long before you go face down in your oatmeal.
A team of Researchers at Boston University got together with local Boston hospitals and designed a system that predicts with 82% accuracy which patients with heart disease and diabetes will need hospital care in the next year.
AI will smell disease on our breath, the way honey bees can do now. Yes, honey bees can detect cancer in ten minutes or less. It won’t take AI longer than a few seconds. Imagine a little sensor on the phone that can detect things on your breath and what that would do for medicine?
It will listen to our coughs and tell us we have COVID or the flu or bronchitis and what to do about it. An MIT team already demonstrated a system that can do it now and that will only get stronger from here on out.
Young people tomorrow will turn to their phones before they turn to a doctor.
They’ll trust algorithms more than people and research shows they already do. Their phone will listen to their cough and peer into their eyes and tell them they need to see a doctor or that they just caught the flu and they need to stay home from work.
This will bring health care to the far reaches of the world faster. It will decentralize medicine, putting it in remote areas swiftly.
There aren’t enough doctors in the countryside. You might find a good GP, but if you’re far away from a city it’s hard to find a top notch specialist.
AI can bridge the gap and people in the deep countryside will ask their phone what to do and suddenly know they need treatment. They’ll drive to the next town over or into the city to get the help they need now.
And doctors and nurses will trust those algorithms too. They’ll want to see your report from your smart app before they prioritize your visit.
Right now when you call the doctor the nurse on the other end of the line has no idea if you’re a hypochondriac or whether you’ve got a real problem. So you go into the queue with everyone else. Even worse, if you live in Berlin an “appointment” means “come on down when we open and wait 2 or 3 hours and maybe you’ll see a doctor.”
But what if you had a little app on your phone or in your glasses and you pointed it at a worrisome spot on your skin and the app told you to call your doctor right away because you might have a real problem? As the health care systems develop around this new technology they’ll change how you get to see doctors. Now you can zap the results over to the nurse and she can put you at the top of the queue or get you in right away to see a specialist instead of waiting for a few hours or a few weeks or even months.
That means that people will get diagnosed better and faster and health insurance prices will drop as more people get the care they need when they need it. People make mistakes all the time. They put off going to the doctor. They don’t bother to see someone. They wait.
Waiting could be fatal.
It means they don’t get the treatment they need when they need it or they need it too late, which rockets up prices because we’re trying to help someone who’s already too far gone and its like piling sand bags against a flood instead of treating something early, which is like plugging a tiny leak in a dam.
But when AI lowers the barrier to entry and all you need to do is have your phone listen to your cough or take a picture with your phone or glasses then only a fool would wait. They’ll get the treatment they need faster and they’ll need less treatment. That will mean they have a much better chance of surviving.
But the most radical change in healthcare will come from a most unexpected place.
The intense pressure of the COVID pandemic will deliver the rise of Big Biotech.
If the last ten years so the rise of big digital tech companies, the next ten years will see an unprecedented surge in biotechnology power.
The pandemic is an existential threat to the world. Even if this disease isn’t as deadly as we imagined something else will come along that’s worse. The world is too hyper connected now, too interwoven for diseases not to run rampant around the globe.
But pressure on systems forces them to change to grow stronger. Diamonds are forged under the most intense pressure and society is too.
The roots of this rise are happening now.
We’ve already seen machine learning deployed to track the spread of the disease. The Center of New Data, a team of volunteer data scientists, used GPS data to trace how a massive bike rally spread COVID across the mid-west of the US. Their model crunched numbers that took weeks before and they got the time down to a few hours, so they could get alerts out to governments and watchdog groups in time to make a difference.
We’ve also seen the fastest time to vaccine candidates in the history of man, by an order of magnitude. Pfizer’s and BioNTech’s early success of 90% efficacy is a stunning achievement. It’s a testament to open science, rapid information sharing and AI accelerating the design and development of drugs, just as I predicted in my article Coronavirus, a Reason for Hope, at the start of the pandemic in March. Investors in BioNTech bet on BioNTech’s bringing together “biology with bioinformatics, robotics and artificial intelligence” to develop cures, Ugur Sahin, BioNTech’s chief executive officer and co-founder said in a company statement.
But they’re not the only companies delivering potential vaccines in record time.
Emily Waltz, writing for IEEE Spectrum tells us that “as of early September, there were 34 vaccine candidates being tested in humans, according to the World Health Organization (WHO). Another 145 candidates were being tested in animals or in the lab, says WHO, which keeps a running worldwide list. Those are astonishing numbers, considering that less than a year ago no one had heard of the novel coronavirus…It typically takes many years, or even decades, to develop a vaccine; until now, the speed record was held by the mumps vaccine, which went from a collected sample to a marketed product in about four years.”
The Pfizer/BioNtech vaccine is a new kind of vaccine too.
Traditional vaccines introduce dead copies of the virus into the system to provoke an immune response. But as this article in the Financial Times points out: “In contrast, mRNA technology — originally developed as a cancer therapy — injects genetic instructions into the body that tell cells to make viral proteins that prime the immune system. Although mRNA vaccines had been under development for several years for viruses including influenza, cytomegalovirus, HIV, rabies and Zika, the arrival of Covid-19 turbocharged the process.”
AI is a major factors driving the biotech revolution as this article in Frontiers in Artificial Intelligence lays out so beautifully. Almost every algorithm in deep learning has been applied to the field of vaccine development. Graph Neural Networks have proven a favorite in drug discovery. “Transformers have demonstrated the capacity to predict drug–target interactions (Shin et al., 2019), model protein sequences (Choromanski et al., 2020), and predict retrosynthetic reaction.” Deep learning has accelerated predicting vaccine candidates making it faster and faster. The pipelines do “feature extraction, feature selection, data augmentation, and cross-validation implemented to predict vaccine candidates against various bacterial and viral pathogens known to cause infectious disease.” AI also helps with predicting toxicity and side effects and interactions with other drugs.
Of course, AI can’t speed up the slowest part of drug discovery, trials in humans, but they’re helping scientists analyze the virus and its structure and showing scientists where to attack with laser accuracy.
AI is checking genetic combinations and potential outcomes. It used to take scientists years to study the structure of the virus and figure out where to hit it hard. They had to comb through existing drugs to find out if they had a chance of also killing off a new pathogen.
This time Chinese scientists synthesized the genome of COVID-19 and distributed it all over the world in days so it could be scrutinized by the brightest minds everywhere.
Remember that it took the Human Genome Project over a decade and 5 billion dollars to synthesize the human genome as recently as 2003.
Now we can sequence your genes for $200 dollars in 48 hours.
This summer we saw the Summit and Rhea supercomputers designed for deep learning and simulations at the Oak Ridge National Labs go to work on the COVID crisis. Summit is the world’s second fastest supercomputer with a whopping 27,000 Nvidia Volta GPUS and Rhea is 512 strong Linux cluster with 9 1 TB GPUs from Nvidia and both of these beasts can rip through heavy machine learning workloads. The AI monster machines crunched through 40,000 genes from 17,000 samples, more than 2.5 billion combinations, in a single week and helped deliver a “eureka moment” for understanding COVID and how it destroys the body.
As tech writer Thomas Smith wrote in his fantastic article on the supercomputer study, “The computer revealed a new theory about how COVID-19 impacts the body: the bradykinin hypothesis. The hypothesis provides a model that explains many aspects of Covid-19, including some of its most bizarre symptoms. It also suggests 10-plus potential treatments, many of which are already FDA approved.”
More than anything the machine revealed that researchers should start thinking of COVID as a vascular disease rather than a lung disease. While the virus enters through the nose and often destroys the lungs it doesn’t rest there. It quickly attacks other cells in the body that have ACE2 receptors, like the heart, arteries, kidneys and intestines. But the disease doesn’t just attack cells that have lots of ACE2 receptors, it hijacks the body’s own systems and tricks them into upregulating ACE2 receptors where there aren’t that many, like the lungs, so that it has more cells to attack.
That kind of insight lets scientists know what kinds of drugs stand the best chance to work in the real world and it took only a week.
There is big money in this vaccine.
The EU struck a deal with Pfizer to secure 300 million doses. The US inked a 1.95 billion dollar deal with them to get 100 million doses. India has a deal with AstraZeneca for 100 million doses by December. They’re partnered with the WHO’s COVAX program to buy doses for poor nations so that nobody is left out of the loop.
According to the CNBC article, that’s just the deals the nation states made with Pfizer and AstraZeneca alone. They’re also hedging their bets with other potential vaccine makes too. “The EU has already signed supply deals with AstraZeneca, Sanofi and Johnson & Johnson for their experimental Covid-19 shots, and is talking with Moderna, CureVac and Novavax to secure their vaccines.”
The infrastructure to find and create cures and deliver those cures won’t just disappear after the corona crisis is quelled. Biotech conglomerates will have become some of the biggest behemoths on the planet. They can take that money and pour it into the coffers of their R&D departments. Long shot, long term strategic plays like immunology therapy, that the trains the body to fight diseases, or antibody cloning will get a surge of new funding and they’ll get the support of big AI clusters and data science teams to boost their efforts.
These new efforts aren’t just looking to create drugs to beat one kind of disease. They’re looking to find ways to kill off whole categories of disease.
If you can train the immune system to attack cancer the way naked mole rats fight cancer then you don’t need to design one drug for breast cancer and another for colon cancer. The naked mole rat doesn’t get cancer. Its immune system kills off any cells that divide too quickly or that clump together. That’s how the body evolves defenses for entire categories of diseases and tomorrow’s biotech companies will train the body to fight more and more of those diseases rather than just letting people evolve their own defenses through brutal Darwinism.
Once you start looking at disease as a system it changes the way we fight everything and now Big Biotech will have the resources to deliver like never before in the history of the world through a combination of supply chain, logistics, spending power and AI.
The Exponential Curve of Progress
Of course, some of what I write won’t all come to pass in the next decade.
As a futurist I’ve come to realize that it’s easier to see what will happen, rather than when it will happen. Getting the timing right is tricky. It’s not hard to be off by decades or more.
I watched Back to the Future II again recently and it hasn’t aged nearly as well as the first one. While many of its predictions may still come to pass, like anti-aging rejuvenation drugs and flying cars, it could take decades more or longer.
As film critic Richard Nelson Corliss said, “Nothing ages so quickly as yesterday’s vision of the future!”
So take ten years with a grain of salt, but the big sweeps of what I’ve predicted will come to pass as more and more time goes by.
Of course, they’ll be bumps along the way to tomorrow’s greatest breakthroughs too. Progress is never a clean, smooth line. It’s a back and forth battle, a thousand ideas contending and competing to take root and grow.
We already saw some early failures in the last decade, like IBM Watson for Healthcare flaming out. It was ahead of its time, trying to do too much, too early. It offered treatment advice to doctors that just couldn’t reach the levels that veteran doctors reached. The system designers didn’t have enough data to really train the machine right and so they came up with synthetic data that didn’t match the real world and that lead to treatment recommendations that swerved into dangerous. At Sloan Kettering it recommended a drug that accelerates bleeding to a cancer patient with severe bleeding.
But too often people see these early failures and assume technologies will never succeed.
People aren’t great at seeing the long sweeps of history, the changes coming, the big patterns flowing out into forever like a fractal. They’re short sighted, seeing today and imagining tomorrow and tomorrow and tomorrow will always be the same.
But the only constant in the modern world is rapid change.
People don’t see engineers going back to the drawing board, learning from their mistakes and coming up with fresh ideas. That’s the unstoppable march of progress, moving like a winding river, but always rushing forward. Over time more and more of doctor’s knowledge will seep into our machines making them smarter and smarter as we learn more about the mind and how it works and how to recreate aspects of it.
At one point the steam engine was only good for getting water out of mines. Centuries later it powered goods around the globe.
We needed monks in a cave making copies of great texts to preserve old knowledge. Then the printing press let us create as many copies as we wanted and the Internet made even more copies so that the death of one couldn’t kill off the knowledge, just like cells in the body dying don’t kill off the whole organism.
Progress has only sped up faster and faster since then.
My parents used to tell me to go play and be home before the street lights came on. They had no idea where I was or what I was doing for most of that time. Now we can’t imagine not talking or messaging someone at any hour of the day and knowing exactly where they are right now because we all have a universal communicator in our pocket.
I used to collect CDs and watch movies on physical media and now the whole world of musical and movies and TVs streams right to my TV or laptop or iPad anytime I want it.
I had to drive to work in this funny little field call the Internet that none of my family or friends could understand and now nobody knows life without the Internet and 24x7 connectivity. I work at home and video call my friends like it’s Dick Tracy come to life.
Things are changing faster and faster and faster. We’re riding more than one exponential curve now. There are dark exponential curves like corona and exponential curves of light too like the rapid spread of cell phones and knowledge around the world.
The history of the world is often dark and brutal at times but it’s also majestic and brilliant. Humans are an incredibly adaptable and innovate little species. We’ve crawled out of the muck and now we’re sending spaceships to Mars and building vaccines in 9 months instead of a 10 years and making cars drive themselves.
Just remember:
Agricultural revolution 12,000 years.
Scientific revolution. 400 years.
Industrial revolution 200 years.
Information age. 50 years.
Faster and faster and faster.
And now welcome to the dawn of the age of intelligence.
Who knows where tomorrow’s roads really lead?
But don’t worry, “where we’re going, we don’t need roads.”
###########################################
I’m an author, engineer, pro-blogger, podcaster, public speaker and Chief Technical Evangelist at Pachyderm, a cutting edge data lineage and AI pipeline platform. I also run the Practical AI Ethics Alliance and the AI Infrastructure Alliance, two open communities helping bring the canonical stack into reality and making sure AI works for all of us.
###########################################
If you love my work please visit my Patreon page because that’s where I share special insights with all my fans.
Top Patrons get EXCLUSIVE ACCESS to so many things:
Early links to every article, podcast and private talk. You read it and hear first before anyone else!
A monthly virtual meet up and Q&A with me. Ask me anything and I’ll answer.
############################################
Lastly, you can join my private Facebook group, the Nanopunk Posthuman Assassins, where we discuss all things tech, sci-fi, fantasy and more.
############################################
This article includes affiliate links to Amazon.
###########################################