Stay ConnectedWhat is Artificial Intelligence? It’s a big part of our daily lives and you want to know. You need to know. But the explanations are so long and boring. Let me give you something short and sweet.
Join me, Dr. Peper, for 5 minute, pleasing, and easy to understand flash talks about everything artificial intelligence. Short and Sweet AI.
We all have thoughts of the future. Some of us will only think of it in passing, but others will spend months or even years contemplating the endless possibilities.
Kazuo Ishiguro’s vision for the future, beautifully presented in his latest book, ‘Klara and the Sun,’ shows an excellent level of thought and research. The British novelist presents an emotionally nuanced concept of what it means to be human or non-human.
In this episode of Short and Sweet AI, I discuss Ishiguro’s latest book and its depiction of robots and artificial intelligence. I also delve into what immortality could look like for humans – will it be robots in our future or something different?
In this episode, find out:
Important Links & Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI, I’m Dr. Peper.
We all have thoughts about the future, some of us in passing and some spend months and years thinking about it. Kazuo Ishiguro’s vision, beautifully presented in his latest book, Klara and the Sun, shows much thought and research. This British novelist presents emotionally nuanced concepts about what it means to be human and not human. I’m not an artificial intelligence expert nor a Nobel prizing winning author like Ishiguro. But I am someone who’s fascinated by artificial intelligence and want people to understand what AI means for our future. From that perspective, I’ve identified three things Ishiguro got right, and two things I think he got wrong, in his new book Klara and the Sun.
First, his depiction of Klara, an artificial friend, or robot, meshes with my understanding of what robots will be like in the future. They will have the ability to understand and integrate information and read and understand human emotions. This ability will surpass the ability of the humans around them at times. With exposure to more human situations and more human observations, robots will increase and refine their emotional abilities. They’ll have true feelings, not simulate them.
The second thing Ishiguro gets right is the future of work. There will be substitutions of humans with machines as machines do more and more of the work. Humans will be displaced and just as in the novel, people will struggle to redefine their role in society and find new meaning.
And the third thing that Ishiguro accurately writes about is the inequality created by those who choose and can afford to have gene-edited children, described as the lifted kids compared to the non-lifted kids, and those whose parents can’t afford or choose not to have their children’s genes edited before birth. I think this will be a real possibility in the near future. There will also be major inequalities in wealth, employment, and opportunity as depicted in the novel.
But one thing that doesn’t make sense is that Klara is able to learn and understand her surroundings so exceedingly well and yet make a very major wrong conclusion. In the book, Klara reasons that people, like robots, need the sun to sustain, nourish and heal them after she misinterprets one example. In the future, robots will have onboard databases that would have quickly given Klara correct information about how humans die from illnesses different than robots. This part of the plot did not seem to fit.
But even more frustrating and inaccurate, I think, is how the author depicts immortality in the future. The longevity culture thrives today so we know it will be prominent in the decades ahead. In the novel, Ishiguro has immortality being carried out through robots trained to learn and replicate everything about a person they’re going to replace. Klara has the ability to exactly replicate the way a human speaks, walks, and sees the world. As a robot, she can capture a human’s personality, but Ishiguro makes the point that a robot is still unable to capture ‘the human heart”.
Established futurists, even today, have said that in the future we will achieve immortality in a different way. We won’t rely on a robot to use their deep learning algorithms to learn and mimic everything about a person so a human can live on as a robot after death like in this book.
No, our immortality will come by uploading our consciousness to the cloud. It’s already begun. Brain–computer interfaces exist today as I’ve discussed in my episode on Neuralink. Highly complicated fields such as synthetic biology, microscopic robots, nanomaterials, and quantum computing, along with others, are merging. It’s inevitable that we’ll have the capability to connect our neocortex and the totality of who we are to a dedicated computer. We can thus choose if we want immortality in a synthetic cloud and choose when to be downloaded at some future time.
As a final point, I think what Ishiguro addresses so eloquently in the novel by what happens to Klara in the end is one of the most important parts of the book. He makes us feel connected to Klara so that her end kindles in us ethical questions. Ethical questions about how to treat, care for and dispose of robots. When a robot truly feels, rather than simulates emotion, then doesn’t a robot have rights?
Thanks for listening. Be curious and if you like this episode, leave a comment, or a thumbs up to let me know you like the content. From short & Sweet AI, I’m Dr. Peper.
What is Liquid AI, and could it prove more effective than other types of AI?
New research into neural nets and algorithms has revealed what some call “Liquid AI,” a more fluid and adaptable version of artificial intelligence.
In my previous episode, I discussed the basics of AI and the limitations that hold it back. It looks like Liquid AI could provide the very solutions that the AI community has been searching for.
In this episode of Short and Sweet AI, I explore the new research behind Liquid AI, how it works, and what it does better than other types of AI.
In this episode find out:
Important Links & Mentions
Resources:
Episode Transcript:
Hello to you who are curious about AI, I’m Dr. Peper. Machine learning algorithms are getting an overhaul from a very unlikely source. It’s a fascinating story.
Neural Nets have Traditional Limitations
Neural nets are the powerhouse of machine learning. They have the ability to translate whole books within seconds with Google Translate, change written text into images with DALLE, and discover the 3D structure of a protein in hours with AlphaFold. But researchers have struggled with neural networks because of their limitations.
Neural nets cannot do anything other than what they’re trained for. They’re programed with parameters set to give the most accurate results. But that makes them brittle which means they can break when given new information they weren’t trained on. Today the deep learning neural nets used in autonomous driving have millions of parameters. And the newest neural nets are so complex, with hundreds of layers and billions of parameters, they require very powerful supercomputers to run the algorithms.
A Neuroplastic Neural Net based on a Nematode
Now researchers from MIT and Austria’s Science Institute have created a new, adaptive neural network they’re describing as “liquid” AI. The algorithm’s based on the nervous system of a simple worm, C. elegans. And elegant it truly is. This worm has only three hundred and two neurons but it’s very responsive with a variety of behaviors. The teams were able to mathematically model the worm’s neurons and build them into a neural network. I’ve explained neural networks in my previous episode called A Simple Explanation of AI.
Computer Software with Neuroplasticity
The worm-brain algorithm is much simpler than the huge neural nets and yet accomplishes similar tasks. In current AI architecture, the neural net’s parameters are locked into the system after training. With liquid AI based on the mathematical models of the worm’s neurons, the parameters are able to change with time and with experience. This is a fluid neural net. As it encounters new information, it adapts. It’s an artificial brain created out of computer software but shows a kind of built-in neuroplasticity like a human brain.
When the algorithm was tested on the task of keeping an autonomous vehicle in its lane, it was just as accurate and efficient as more advanced and complex machine learning algorithms. The worm-brain model also adapts new pathways. In one example the researchers found the algorithm could change its underlying mathematical equations when it had new information like, there was rain on the autonomous vehicle’s windshield. This “neuroplasticity” means the neural net is less likely to break when it’s given data it hasn’t been trained on.
Liquid AI Uses Less Parameters, Less GPUs
Also, with this new approach, the researchers reduced the neural net’s size. It has only 75,000 trainable parameters instead of the million or billion parameters in some machine learning algorithms. This decreases the GPUs or computing power needed to run the algorithm. You can appreciate the excitement this has generated. Liquid AI is an adaptable machine learning algorithm that consumes less power, uses a smaller neural net, while being as accurate as the larger machine learning systems.
New Liquid AI is More Transparent
But I saved the best for last. For many years AI ethicists and researchers have been deeply troubled by machine learning systems being “black boxes” meaning how they work and arrive at their results is largely impenetrable. No one can determine exactly what’s going on within the neural nets that lead to the successful results. This can be a big problem when unsupervised machine learning models are trained on the unfiltered internet because there’s no way of knowing or controlling what they learn.
But this AI system was designed differently. It’s a new type of AI architecture. This liquid neural net is more open to observation and study. Researchers are able to analyze the neural net’s decision making and diagnose how it arrived at the answers. It’s more transparent. It’s adaptable, efficient, smaller, accurate, and transparent. Liquid AI, now that’s a good thing.
Thanks for listening. I hope you found this helpful. Be curious and if you like this episode, click the thumbs up button and leave a comment. From Short and Sweet AI, I’m Dr. Peper.
What is AI really, and how does it work?
If you are interested in AI, you’ll undoubtedly know that many of the concepts are a bit overwhelming. There are plenty of terminologies to understand, such as machine learning, deep learning, neural networks, algorithms, and much more.
With the world of AI continually evolving, it’s good to go over some of the basic concepts to better understand how it’s changing.
In this episode of Short and Sweet AI, I address some of the questions that I get asked a lot: what is AI? How does AI work? I also delve into some of the limitations of AI and their possible solutions.
In this episode find out:
Important Links & Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper.
If you’re listening to this, you probably think AI’s interesting and important like me. But sometimes I find the concepts are a little overwhelming. I want to go over something I get asked a lot. People ask me, what is AI really, how does it work? Actually, there’re new things going on with how AI works. So, it’s good to go over some of the basic concepts in order to understand the way AI is changing.
How does AI work?
Artificial Intelligence happens with computers. They’re programed using algorithms. Algorithms are step by step instructions telling the computer what to do to solve a problem. Just like a recipe has specific steps you follow in sequence, to bake a cake, or cook something. Computer scientist write algorithms using a programming language the computer understands. These computer languages have strange names like Python or C plus, plus.
The computers also perform math calculations or computations to analyze the information and give an answer. This is known as computational analysis. Basically, the programing language and math calculations are computer software. Using this software, the algorithms come up with an answer from data sets fed into the computer.
Machine Learning is a type of AI
The major AI being used today is called machine learning. Machine learning is carried out by artificial neural networks, or nets for short. Neural nets underpin the most advanced artificial intelligence being used today. They’re called neural networks because they’re based in part on the way neurons in the brain function. In the brain the neuron receives inputs or information, processes the information, and then gives a result or output.
Artificial intelligence uses digital models of brain neurons. These are artificial neurons, based on the computer binary code of ones and zeros. The digital neurons process information and then pass it along to other higher layers of processing. Higher, meaning the results become more specific, just like in the brain.
Deep Learning is a type of Machine Learning
Before computers can give us the answers, they have to be trained on large amounts of data. As the computer processes more and more information, it learns from the data. This is called training the machine. Then when you give the computer completely new data, the machine knows what to do with it and can give you a correct answer to your specific question.
If you have many, many layers of neural networks, each processing and passing the information to another layer, it’s called a deep neural network. When machines learn from deep neural networks, it’s called deep learning.
Present day AI has Limitations
All of the software and computer calculations used in machine learning, especially deep learning, require absurd amounts of data and computer power. The neural nets can be hundreds of layers deep with billions of parameters like AlphaFold or GPT-3 which I talked about in previous episodes. These are gargantuan machine learning algorithms and require very powerful supercomputers to run them. This limits who can use machine learning to only large tech companies and corporations.
Yet as mighty as these neural nets appear, at a core level they are very narrow. And that’s another limitation. They do exactly the one thing they’re trained to do such as recognize an image, steer a car to the left or right, or translate something from one language to another. When you ask a neural net to do something that deviates from its training, it acts brittle and breaks.
New AI Neuroplasticity
But there’s something new in artificial intelligence. AI has reached a point where it’s less artificial and more biological. Like the human brain, AI has developed “neuroplasticity.” Now that you understand basic artificial intelligence, next time let’s discuss something called “liquid” AI which is so cool. It solves a lot of these limitations with a type of artificial neuroplasticity.
I hope you found this helpful. Be curious and if you like this episode, please follow my channel. Or you can leave a comment and click the thumbs up button, which lets me know you like the content. From Short and Sweet AI, I’m Dr. Peper.
Microscopic robots might sound like the plot of a futuristic novel, but they are very real.
In fact, nanotechnology has been a point of great interest for scientists for decades. In the past few years, research and experimentation have seen nanotechnology's science develop in new and fascinating ways.
In this episode of Short and Sweet AI, I delve into the topic of microscopic robots. The possibilities and capabilities of nanobots are something to keep a watchful eye on as research into nanotechnology starts to pick up speed.
In this episode, find out:
Important Links & Mentions
Resources:
Episode Transcript:
Today I’m talking about microscopic robots.
In the book Super Sad True Love Story by Gary Shteyngart, set in the future, wealthy people pay for life extension treatments. These are called “dechronification” methods and include infusions of “smart blood” which contain swarms of microscopic robots. These tiny robots are about 100 nanometers long and rejuvenate cells and remodel major organs throughout the body via the bloodstream. In this way the wealthy live for over a century.
That book was my first introduction to the idea of microscopic robots, also known as nanobots, more than a decade ago. Nanotechnology is more than a subplot in a futuristic novel. It’s an emerging field of designing and building robots which are only nanometers long. A nanometer is 1000 times smaller than a micrometer. Atoms and molecules are measured in nanometers. For example, a red blood cell is about 7000 nanometers while a DNA molecule is two and a half nanometers.
The father of nanotechnology is considered to be Richard Feynman who won the Nobel prize in physics. He gave a talk in 1959 called “There’s Plenty of Room at the Bottom.” The bottom he’s referring to is size, specifically the size of atoms. He discussed a theoretical process for manipulating atoms and molecules which has become the core field of nanoscience.
The microscopic robots are about the size of a cell and are based on the same basic technology as computer chips. But creating an exoskeleton for robotic arms and getting these tiny robots to move in a controllable manner has been a big hurdle. Then in last few years Marc Miskin, a professor of electrical and systems engineering, and his colleagues, used a fresh, new design concept.
They paired 50 years of microelectronics and circuit boards to create limbs for the robots and used a power source in the form of tiny solar panels on its back. By shining lasers on the solar panels, they can control the robot’s movements. In fact, you can see a battalion of microscopic robots in a coordinated “march” on a video linked in the show notes.
The genius of Miskin’s work is that the robot’s brain is based on computer chip technology. The same technology has powered our computers and phones for half a century. This means the tiny robots can be integrated with other circuits to respond to more complex commands.
The nanobot can be equipped with sensors to report on conditions in whatever environment it’s in. These are truly miniaturized machines capable of being injected through a syringe and still maintain their structure and function. And since they use the same well understood manufacturing process as computer chips, they are easy and cheap to produce. Millions of tiny robots can be made at the same time. The end result is electronically integrated, mass manufactured, microscopic robots.
Like in the novel Super Sad True Love Story, we could have smart blood with nanobots injected into our bloodstream. The nanobots could be used to deliver cancer drugs in humans right where they’re needed and avoid harmful side effects to other tissues. They could be used to reduce plaque which has built up in arteries, or for treating hard to reach areas of the human body with microsurgery.
And by the way, the author of that book, Gary Shteyngart, has credited his ideas to Ray Kurzweil, whom you’ve heard me speak of many times. Kurzweil is convinced that nanotechnology is the way we can someday merge humans and technology. As he explained just several years ago, “These robots will go into the brain and provide virtual and augmented reality from within the nervous system rather than from devices attached to the outside of our bodies. The most important application … is that we will connect the top layers of our neocortex to the synthetic cortex in the cloud.”
Leave a comment and let me know if you could access all its power and knowledge, would you connect your brain to the cloud? What if it meant you could store your consciousness in the cloud after you die? I’ve come across a short story called Remembrance which chillingly depicts this in the future. These may seem like crazy ideas, but microrobots are real and soon may be flowing through our bloodstream.
Is a world without work a reality we need to prepare for?
In my last episode, I discussed whether the fear of machines taking over jobs was truly misplaced anxiety, as experts say. Experts believe that there’s no cause for alarm, but not everyone agrees.
Some believe that a future where human workers become obsolete is a real possibility we need to prepare for.
In this episode of Short and Sweet AI, I delve into the theory that our future will be a world without work. I discuss Daniel Susskind’s fascinating book, ‘A World Without Work,’ which explores the topic of technological unemployment in great detail.
In this episode, find out:
Important Links & Mentions
Resources
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper and today I’m talking about a world without work.
In my last episode, I talked about the future of work. Economists, futurists, and AI thinkers generally agree that technological unemployment is not a real threat. Our anxiety about machines taking our jobs is misplaced. There have been three centuries of technological advances and each time, technology has created more jobs than it destroyed. So, no need for alarm.
But Daniel Susskind, an Oxford economist and advisor to the British government, thinks this time, with artificial intelligence, the threat really is very real. He wants us to start discussing the future of work because as he sees it, the future of work is A World Without Work, which is the title of his recent book. He explains why what’s been called a slow-motion crisis of losing jobs to machines and automation, needs to be discussed now because it really isn’t slow-motion anymore.
Despite increased productivity and GDP from artificial intelligence, Susskind presents evidence technological unemployment is coming. As he says, we don’t need to solve the mysteries of how the brain and mind operate to build machines that can outperform human beings.
Machines have been taking over jobs requiring manual abilities for decades. It’s happening now. Although the American manufacturing economy has grown over the past few decades, it hasn’t created more work. Manufacturing produces 70 percent more output than it did in 1986 but requires 30 percent fewer workers to produce it.
More importantly, machines are increasingly being used in the cognitive skills areas, too. AI deep learning is used to read x-rays, compose music, review legal documents, detect eye diseases, and personalize online learning systems. And in the controversial area of synthetic media, AI systems can generate believable videos of events that never happened.
Machines also have human skills such as empathy and the ability to determine how someone feels. Algorithms are making headway into effectively and accurately reading human emotion through facial recognition and language. I talked about this in my episode on Affective AI.
The most significant point Susskind makes, in my opinion, is that we think machines can’t perform some human tasks because they can’t perform them the same way humans do. Many doctors use gut instinct and vast hands-on experience when treating patients. Machines won’t be able to diagnose this same way, the way doctors do. Machines will be able to accurately perform the task but in a different way.
So, the three capabilities that humans use to earn a living, manual skills, cognitive skills, and emotional intelligence, are all being replaced by machines. Susskind doesn’t know exactly when this will happen, but he thinks it will be sooner than most people realize, within just decades, and certainly during the 21st century because during the next 80 years, machines will become a trillion times more powerful.
In a future world without work, Susskind asks how we will earn enough to live on and how will we all find meaning in our life. His assessment is that government, or the big state as he calls it, will be needed to redistribute income and wealth. And even more importantly, governments will need to introduce programs to nudge us into behaviors that will give us fulfillment rather than yielding to Netflix, boredom and despair. Instead of labor market policies, governments will need to form leisure policies that shape the way people use their spare time because the future of work will be the future of leisure.
At different points in time throughout history large groups from the Greeks to the English have lead life with meaning but without work. For instance, in Victorian England, the upper classes were far from depressed by their idleness. Indeed, they created some of the greatest poetry, literature, and science the world has known.
According to many AI experts, as well as Susskind, AI advances will continue at an exponential rate. It’s inevitable. Machines will ultimately do most of what humans do. In the science fiction series, The Expanse, in the future there won’t be work for the majority of people on Earth. People exist on a type of universal basic income. But some families feel strongly that they want their children to live a life with meaning that comes from work. Their only option is to move to Mars which exists to defend Earth, and where everyone has a job, working for the military.
I highly recommend reading A World Without Work. David Susskind goes into much more detail with entertaining examples and comprehensive discussions of universal basic income, the age of labor, the limits of education, and much more.
But what do you think? Could you cultivate a meaningful life without work?
Thanks for listening. I hope you found this helpful. If you liked this episode, please leave a review and subscribe. From Short and Sweet AI, I’m Dr. Peper.
Are you anxious that a machine will one day replace your job? It’s a common enough fear, especially with the rate technology is advancing.
If you have watched any of my previous episodes, you will know that technology is accelerating exponentially! We have seen the equivalent of 20,000 years of technology in just one century.
Naturally, people worry about what this means for the future of work. Will human workers become obsolete one day?
In this episode of Short and Sweet AI, I explore “technological unemployment” in more detail and whether it’s something we should be concerned about.
In this episode find out:
Important Links & Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper and today I’m talking about the future of work.
For centuries there’ve been predictions that machines would put people out of work for good and give rise to technological unemployment. If you’ve been listening to my episodes you know that technology today is accelerating exponentially. We are living at a time when many different types of technology are all merging and accelerating together. This is creating enormous advances which some have said will lead to the equivalent of 20,000 years of technology in this one century. And experts are asking what does that mean for the future of work?
Historians, economists, and futurists describe the anxiety about new machines replacing workers as a history of misplaced anxiety. Three hundred years of radical technological change have passed and there is still enough work for people to do. The experts say, yes, technology leads to the loss of jobs, but ultimately more new jobs are created in the process. Automation and the use of machines increases productivity which leads to creation of new jobs and increased GDP.
A well-known example would be the rise in the use of ATM machines in the 1990s which led to many bank tellers losing their jobs. But at the same time, the ATMs enabled banks to increase their productivity and profits and led to more branches being opened and more bank tellers being hired. The bank tellers now spent their time carrying out more value-added, non-routine tasks.
In the early industrial revolution, when mechanical looms were introduced, many highly skilled weavers lost their jobs, but even more jobs were created for less-skilled workers who operated the machines.
People who study economics and AI are optimistic. They think machines can readily perform routine tasks in a job but would struggle with non-routine tasks. Humans will still be needed for their cognitive, creative, and emotional skills that machines don’t have. In this way, workers will complement machines and will always be needed.
The World Economic Forum, headed by Klaus Schwab who wrote the 4th Industrial Revolution, released a recent report on the Future of Work. They estimated by 2025, 85 million jobs will be lost through artificial intelligence, but 97 million new jobs created. This goes along with the mainstream thinking that technological unemployment is not something to worry about in the foreseeable future. But when you read the report in more detail, some red flags emerge.
Surveys show 43% of businesses are set to reduce their workforce due to technology, 50% of all employees will need reskilling in the next 5 years, and job creation is slowing while job destruction accelerates.
Many articles on the world economic forum website also dampen the prevailing optimism for the future of work. One example is the profession of psychologists.
Previous projections assumed the work of a psychologist requires extensive empathic and intuitive skills. Initially, it was thought very unlikely to be replicated by a machine during our lifetime. But experts have found artificial intelligence has become woven into the fabric of our daily lives at an accelerating pace. With the pandemic, the use of meditation and mindfulness apps such as Headspace and Calm has soared, as well as other technology-mediated forms of therapy. The most recent report concluded it’s almost certain the work of psychologists will be replaced in large part by artificial intelligence.
So what’s going on here? Is anxiety about technological unemployment misplaced or will machines be able to perform most human tasks, and how soon?
Well, I’ve uncovered another penetrating viewpoint on the future of work in a book by Daniel Susskind. The book’s been described as “required reading for any presidential candidate.” His premise is the future of work is A World Without Work, which is the title of his book. And I had a glimpse of what a world without work looks like in the science fiction series, The Expanse. I’ll be talking about both in my next episode.
What do you think? Do you feel anxious that your job will be replaced by a machine during your lifetime?
Please leave me your thoughts in the comments, and a review if you like this episode. From Short and Sweet AI, I’m Dr. Peper.
What is the protein folding problem that has left researchers stuck for nearly 50 years?
Knowing the 3D shape of proteins is so important for our understanding of various diseases and vaccine development. However, these shapes are fantastically complex and difficult to predict. Researchers have spent years trying to determine the 3D structure of proteins.
Thanks to AI systems like AlphaFold, it’s now much easier and faster to predict protein shapes. AlphaFold is currently leading the way in protein folding research and has been described as a “revolution in biology.”
In this episode of Short and Sweet AI, I explore the protein folding problem in more detail and how AlphaFold is accelerating our understanding of protein structures.
In this episode, find out:
Important Links and Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper and today I’m talking about AlphaFold.
One of Biology’s most difficult challenges, one that researchers have been stuck on for nearly 50 years is how to determine a protein’s 3D shape from its amino-acid sequence. It's known as “the protein folding problem”.
When I first came across the subject, I thought, ok, that’s a biology problem and maybe AI will solve it but there’s no big story here. I was wrong.
Some biologists spend months, years, or even decades performing experiments to determine the precise shape of a protein. Sometimes they never succeed. But they persist because having the ability to know how a protein folds up can accelerate our ability to understand diseases, develop new medicines and vaccines, and crack one of the greatest challenges in biology.
Why is protein folding so important? Proteins structures contain as much, if not more information, than stored in DNA. Their 3D shapes are fantastically complex. Proteins are made up of strings of amino acids, called the building blocks of life. In order to function, the strings twist and fold into a precise, delicate shapes that turn or wrap around each other. These strings can even merge into bigger, megaplex structures.
Only then can these proteins function in the way necessary to build and sustain life. A protein’s shape defines what the protein can do and what it cannot do.
But there’s an astronomical number of ways a protein can fold into its final 3D structure. It’s called Levinthal’s paradox. Cyrus Levinthal, a molecular biologist, published a paper in 1969 called “How to Fold Graciously.” He found there are so many degrees of freedom in an unfolded chain of amino acids, the molecule has an astronomical number of possible configurations.
There’re an estimated 200 million known proteins with 30 million new ones discovered every year. Each one has a unique 3D shape which determines how it works and what it does. For the last 50 years, biologists discovered the exact 3D structure of only a tiny fraction of known proteins.
The protein folding problem led to a global competition called CASP, which stands for Critical Assessment of Structure Protein. Scientists measure and compare their research efforts using computer-based predictions. The competition started in 1994 to improve computational methods for accurately predicting a protein’s 3D shape.
DeepMind, an AI research lab owned by Google, has made headlines for creating deep learning neural networks AlphaGo and AlphaZero, which beat the world’s leading chess and Go champions. I’ve talked about them in previous episodes. Protein folding has been called the challenge of a lifetime, and the researchers at DeepMind wanted to use AI, not for only game playing but to make a real-world impact. So, DeepMind went to work creating AlphaFold, a deep learning computer system, to solve the protein folding problem.
In 2018 AlphaFold entered the CASP competition for the first time. It achieved the highest score for accurately predicting various protein structures, scoring 60 out of a possible 100 points. But the AlphaFold researchers thought it could improve the accuracy and developed the deep learning neural network even further.
In addition to using a data set with 170,000 protein structures, DeepMind supercharged the algorithm. They added data about physics, geometry, and evolutionary history into their training model. The algorithm analyzed any buried relationships or patterns and was able to determine highly accurate structures in a matter of days, even hours. It could predict a protein’s shape down to the width of an atom.
The turning point came in the CASP competition in November 2020. AlphaFold, as well as teams from Microsoft and the Chinese tech company Tencent, competed to predict protein structures considered to be moderately difficult. The best performance of the other teams was 75 points on a 100-point scale. AlphaFold performed so unbelievably well, it was called a revolution in biology. AlphaFold scored 90 out of 100.
One researcher had been looking for the structure of a protein for 10 years, an absolute decade. AlphaFold’s predictions gave him the protein’s 3D structure in half an hour. You can’t make this stuff up. His exuberance is understandable when he says: “This will change medicine. It will change research. It will change bioengineering. It will change everything.”
Here are a few more comments made by experts that convey why Alphafold is not just a big story but rivals the discovery of DNA.
One researcher said, “I nearly fell off my chair when I saw these results.” Another proclaimed, “It’s a breakthrough of the first order, certainly one of the most significant results of my lifetime.” Another commented, “…a stunning advance…It’s occurred decades before many people in the field would have predicted.”
And John Moult, a professor who helped to create the CASP competition, describes it as a dream come true. He said “I always hoped I would live to see this day. But it wasn’t always obvious I was going to make it.”
Thanks for listening. I hope you found this helpful. If you like this episode, please leave a review and subscribe, because then you’ll receive my episodes weekly. From Short and Sweet AI, I’m Dr. Peper.
One of the founding principles of OpenAI, the company behind technology such as GPT-3 and DALL•E, is that AI should be available to all, not just the few.
Co-founded by Elon Musk and five others, OpenAI was partly created to counter the argument that AI could damage society.
OpenAI was originally founded as a non-profit AI research lab. In just six short years, the company has paved the way for some of the biggest breakthroughs in AI. Recent controversy arose when OpenAI announced that a separate section of its company would become for-profit.
In this episode of Short and Sweet AI, I discuss OpenAI’s mission to develop human-level AI that benefits all, not just a few. I also discuss the controversy around OpenAI’s decision to become for-profit.
In this episode, find out:
Important Links and Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper and today I’m talking about a truly innovative company called OpenAI.
So what do we know about OpenAI, the company unleashing all these mind-blowing AI tools such as GPT-3 and DALL·E?
Open AI was founded as a non-profit AI research lab just 6 short years ago by Elon Musk and 5 others who pledged a billion dollars. Musk has been openly critical that AI poses the greatest existential threat to humanity. He was motivated in part to create OpenAI by concerns that human-level AI could damage society if built or used incorrectly.
Human-level AI is known as AGI or Artificial General Intelligence. The AI we have today is called Narrow AI, it’s good at doing one thing. General AI is great at any task. It’s created to learn how to do anything. Narrow AI is great at doing what it was designed for as compared to Artificial General Intelligence which is great at learning how to do what it needs to do.
To be a bit more specific, General AI would be able to learn, plan, reason, communicate in natural language, and integrate all of these skills to apply to any task, just as humans do. It would be human-level AI. It’s the holy grail of the leading AI research groups around the world such as Google’s DeepMind or Elon’s OpenAI: to create artificial general intelligence.
Because AI is accelerated at exponential speed, it’s hard to predict when human-level AI might come within reach. Musk wants computer scientists to build AI in a way that is safe and beneficial to humanity. He acknowledges that in trying to advance friendly AI, we may create the very thing we are concerned about. Yet he thinks the best defense is to empower as many people as possible to have AI. He doesn’t want any one person or a small group of people to have AI superpower.
OpenAI has a 400-word mission statement, which prioritizes AI for all, over its own self-interest. And it’s an environment where its employees treat AI research not as a job but as an identity. The most succinct summary of its mission has been phrased “… an ideal that we want AGI to go well” Two specific parts to its mission are to avoid building human-level AI that harms humanity or unduly concentrates power in the hands of a few.
But there’s a big controversy. OpenAI recently reorganized to form a separate section that’s for profit. It never released the software for GPT-3 as open code for programmers to use and build on. Instead, it licensed GPT-3 exclusively to Microsoft for a billion dollars. OpenAI realized staying a non-profit was financially untenable. It defends its decision explaining it needs billions of dollars to build AGI and fulfill their mission. Personally, I see the necessity for this. I’ve said elsewhere, “If you’re dedicated to your mission, you first have to find consistent funding. We don’t always need more ideas about how to make the world better. We need more ways to consistently fund the ideas we have.” It’s a huge challenge when you realize DeepMind, Open AI’s main competitor, spent 442 million dollars on research the same year OpenAI spent only 11 million.
But there’s been an outcry from critics who say switching to a for profit model is inconsistent with OpenAI’s mission to democratize AI for all. I’d be interested to know what you think about OpenAI’s decision. Do you think its non-profit mission justifies it becoming for profit? Let me know your thoughts and leave a comment.
Thanks for listening, I hope you found this helpful. Be curious and if you like this episode, please leave a review and subscribe because then you’ll receive my episodes weekly. From Short and Sweet AI, I’m Dr. Peper.
Is DALL·E the latest breakthrough in artificial intelligence?
It seems there’s no end to the fascinating innovations coming out in the world of AI. DALL·E, the most recent tool developed by OpenAI, was announced just months after unveiling its groundbreaking GPT-3 technology.
DALL·E is another exciting breakthrough that demonstrates the ability to turn words into images. As a natural extension of GPT-3, DALL·E takes pieces of text and generates images rather than words in response.
In this episode of Short and Sweet AI, I discuss DALL·E in more detail, how it differs from GPT-3, and how it was developed.
In this episode, find out:
Important Links and Mentions:
Resources:
Episode Transcript:
Hello to you who are curious about AI. I’m Dr. Peper and today I’m talking about DALL·E.
In a previous episode, I highlighted a new type of AI tool called GPT-3. GPT-3 is a machine learning language model trained on a trillion words that generates poetry, stories, even computer code. Within months of announcing GPT-3, OpenAI released DALL·E. DALL·E is not just another breathtaking breakthrough in AI technology. It represents the ability, by a machine, to manipulate visual concepts through language.
DALL·E is a combination of the surrealist artist Salvador Dali and the animated robot Wall-E. What it does is simple but also revolutionary. It’s a natural extension of GPT-3. The AI system was trained with a combination of the 13 billion features of GPT-3 added to a dataset of 12 billion images.
DALL·E takes text prompts and responds not with words but images. If you give the system the text prompt, “an armchair in the shape of an avocado” it generates an image to match it. It’s a text-to-image technology that’s very powerful. It gives you the ability to create an image of what you want to see with language because DALL·E isn’t recognizing images, it draws them. And by the way, I would buy one of those avocado chairs if they existed.
You can visit OpenAI’s website and play with images generated by this astounding technology: a radish in a tutu walking a dog, a robot giraffe, a spaghetti knight. The images are from the real world or are things that don’t exist, like a cube of clouds.
How does It Work?
Text-to-image algorithms aren’t new but have been limited to things such as birds and flowers or other unsophisticated images. DALL·E is significantly different from others that have come before because it uses the GPT-3 neural network to train on text plus images.
DALL·E uses the language and understanding provided by GPT-3 and its own underlying structure to create an image prompted by a text. Each time it generates a large set of images. Then another machine learning algorithm called CLIP ranks the images and determines which pictures best match the text. As a result, the illustrations are much more coherent and reflect a blend of more complex concepts. This is what makes DALLE the most realistic text-to-image system ever produced.
Unintended But Useful Behaviors
DALL·E also demonstrates another example of “zero-shot visual reasoning.” Zero-shot learning or ZSL, is the ability of models to perform tasks that they weren’t specifically trained to do. They’re unintended but useful behaviors.
In the case of GPT-3, it can write computer code even though it wasn’t trained to do coding. DALL·E “learned” to generate images from captions or if given the right text prompt it can transform images into sketches. Another task it wasn’t specifically trained to do was to design custom text on street signs. Essentially DALL·E can behave as a Photoshop filter.
It also shows an understanding of visual concepts. It can, in a sense, answer questions visually. When given hidden patterns and prompted to solve an uncompleted grid with images to match, DALL·E was able to fill in the grid with matching patterns without being given any prompts.
Creativity is a Measure of Intelligence
Experts agree language grounded in visual understanding like DALL·E makes AI smarter. This machine learning system has the ability to take two unrelated concepts such as an armchair and an avocado and put them together in a coherent, new way. This is stunning because the ability to coherently blend concepts and use them in a new way is key to creativity. In essence, the machine stores information about our world to use and generalize in a very human-like way. And in the AI world, creativity is one measure of intelligence. So, is this how machine intelligence becomes human-like intelligence?
Thanks for listening, I hope you found this helpful. Be curious and if you like this episode, please leave a review and subscribe because then you’ll receive these episodes weekly. From Short and Sweet AI, I’m Dr. Peper.
Some have called it the most important and useful advance in AI in years. Others call it crazy accurate AI.
GPT-3 is a new tool from the AI research lab OpenAI. This tool was designed to generate natural language by analyzing thousands of books, Wikipedia entries, social media posts, blogs, and anything in between on the internet. It’s the largest artificial neural network ever created.
In this episode of Short and Sweet AI, I talk in more detail about how GPT-3 works and what it’s used for.
In this episode, find out:
Important Links and Mentions:
Resources:
Episode Transcript:
Today I’m talking about a breathtaking breakthrough in AI which you need to know about.
Some have called it the most important and useful advance in AI in years. Others call it crazy, accurate AI. It’s called GPT-3. GPT-3 stands for Generative Pre-trained Transformers 3, meaning it’s the third version to be released. One developer said, “Playing with GPT-3 feels like seeing the future”.
Another Mind-Blowing Tool from OpenAI
GPT-3 is a new AI tool from an artificial intelligence research lab called OpenAI. This neural network has learned to generate natural language by analyzing thousands of digital books, Wikipedia in its entirety, and a trillion words found on social media, blogs, news articles, anything and everything on the internet. A trillion words. Essentially, it’s the largest artificial neural network ever created. And with language models, size really does matter.
It’s a Language Predictor
GPT-3 can answer questions, write essays, summarize long texts, translate languages, take memos, basically, it can create anything that has a language structure. How does it do this? Well it’s a language predictor. If you give it one piece of language, the algorithms are designed to transform and predict what the most useful piece of language should be to follow it.
Machine learning neural networks study words and their meanings and how they differ depending on other words used in the text. The machine analyzes words to understand language. Then it generates sentences by taking words and sentences apart and rebuilding them itself.
Supervised vs Unsupervised machine learning
GPT-3 is a form of machine learning called unsupervised learning. It’s unsupervised because the training data is not labelled as a right or wrong response. It’s free from the limits imposed by using labelled data. This means unsupervised learning can detect all kinds of unknown patterns. The machine works on its own to discover information.
In supervised machine learning, the machine doesn’t learn on its own. The machine is supervised during its training by using data labelled with the correct answer. This method isn’t flexible. It can’t capture more complex relationships or unknown patterns.
Open AI first described GPT 3 in a research paper in May 2020. Then it allowed selected people and developers to use it and report their experiences online of what GPT 3 can do. There’s even an informative article about GPT 3 written entirely by GPT-3.
Judge for Yourself
One researcher used GPT-3 to generate a Harry Potter parody in the style of Ernest Hemingway. Take a listen: "It was a cold day on Privet Drive. A child cried. Harry felt nothing. He was dryer than dust. He had been silent too long. He had not felt love. He had scarcely felt hate. Yet the Dementor’s Kiss killed nothing.”
I think that sounds pretty good!
And there’s a twitter feed called gptwisdom which generates quotes using GPT-3. Here are a few examples:
“Dull as a twice-told tale.” Or: "The point at which a theory ceases to be a theory is called its limit.” Or this thoughtful gpt3 generated quote: “The truthfulness of your simplicity can only grow, as you improve your character.”
Things to Know About This Technology
In essence, GPT-3 is a universal language model. The model learned to identify more than 175 billion different distinguishing features of language. These features are mathematical representations of patterns. The patterns are a map of human language. Using this map, GPT-3 learned to perform all sorts of tasks it was not even built to do.
Unintended Abilities
One unexpected ability is GPT-3 can write computer code. Makes sense, because computer code is a type of language. But this behavior was entirely new. It even surprised the designers of GPT-3. They didn’t build GPT-3 to generate computer code, they trained it to do just one thing. Predict the next word in a sequence of words.
All in all, people discovered it can do many tasks that it wasn’t originally trained to do. They found it could build an app by giving it a description of what they wanted the app to do. It can generate charts and graphs from plain English. It can identify paintings from written descriptions. It can generate quizzes for practice on any topic and explain the answers in detail.
The Best But Flawed
GPT-3’s ability to generate text is the best that has ever been seen in AI. Yet it’s far from flawless. It can spew offensive and biased language and struggles with questions that involve reasoning by analogy. It isn’t guided by any coherent understanding of reality because it doesn’t have an internal model of the world. Sometimes it produces nonsense because it’s essentially word-stringing. Other AI researchers say it’s like a black box and it’s hard to figure out what this thing is doing.
A Machine Like Us
And yet, the consensus is GPT-3 is shockingly good. But because it can generate convincing tweets, blog posts and computer code, people think of it as being like them. They are reading humanity into the GPT-3 system and, as such, run the risk of ignoring its limits. Sam Altman, one of the founders of OpenAI which developed GPT-3, has thanked everyone for their compliments. But he urges caution about the hype. He says, “AI is going to change the world but GPT-3 is just a very early glimpse. We still have a lot to figure out.”
Thanks for listening, I hope you found this helpful. Be curious and if you like this episode, please leave a review and subscribe because then you’ll receive these episodes weekly. From Short and Sweet AI, I’m Dr. Peper.
Our TOPPODCAST Picks 
