It is likely that by the time children born today go to kindergarten, artificial intelligence (AI) will already have surpassed humans in all cognitive tasks, from science to creativity. In 1999, I predicted that we would have that kind of artificial general intelligence (AGI) in 2029; and, at the time, most experts thought he had begun writing fiction. However, after the spectacular advances of recent years, many experts now think that we will have IAG even before then, so technically I have gone from being an optimist to being a pessimist without having changed my prediction at all.
Having worked in the field for 61 years (longer than anyone alive), I am pleased to see AI at the center of the global conversation. Now, almost all the comments overlook how big language models like ChatGPT and Gemini fit into an even bigger story: AI is about to leapfrog from revolutionizing the digital world to transforming it too. the physical world. This will bring countless benefits, but in three areas the repercussions will be especially profound: energy, manufacturing and medicine.
Ray Kurzweil
Energy sources are some of the most basic resources of civilization. For two centuries, the world has needed dirty, non-renewable fossil fuels. However, the use of just 0.01% of the sunlight that the planet receives would be enough to cover all human energy consumption. Since 1975, solar cells have become 99.7% cheaper per watt of capacity, allowing global capacity to multiply by 2 million. So why isn’t solar energy already imposed?
The problem is twofold. First, photovoltaic materials remain too expensive and inefficient to completely replace coal and gas. Second, given the large variation in solar generation, both on a diurnal (day/night) and annual (summer/winter) scale, enormous amounts of energy need to be stored for when it is needed, and current battery technology is not profitable enough. The laws of physics indicate that considerable improvements are possible, but the range of chemical possibilities to be explored is so enormous that scientists have made painfully slow progress.
Read also
AI, on the other hand, is capable of rapidly sifting billions of chemicals in a simulated manner and is already driving innovations in photovoltaic panels and batteries. And such evolution is about to accelerate dramatically. Throughout history until November 2023, humans had discovered about 20,000 stable inorganic compounds for use in different technologies. And then Google’s GNoME AI discovered many more; Their numbers increased overnight to 421,000. However, that barely scratches the surface of applications in materials science. When a much smarter IAG finds completely optimal materials, photovoltaic megaprojects will be viable, and solar energy may be so abundant that it will become almost free.
Energy abundance allows another revolution: in manufacturing. The costs of almost all goods (from food and clothing to electronics and automobiles) come largely from a few common factors such as energy, labor (including cognitive labor such as R&D and design) and raw materials. AI is on its way to greatly reducing all these costs.
Read also
After abundant and cheap solar energy, the next component is human labor, which often must perform backbreaking and dangerous work. AI is making great advances in robotics that can greatly reduce labor costs. Robotics will also reduce the costs of extracting raw materials, and AI is finding ways to replace expensive rare earth elements with common elements such as zirconium, silicon and carbon-based graphene. Taken together, that means that almost all kinds of goods will become very cheap and abundant.
Advanced manufacturing capabilities will allow the price-performance ratio of computing to maintain the exponential trajectory of the last century: a 75 trillion improvement since 1939. And this is due to a feedback loop: today’s most modern AI chips are used to Optimize next-generation chip designs. In terms of calculations per second per constant dollar, the best hardware available last November could do about 48 billion operations. Nvidia’s new B200 GPUs exceed 500 billion.
Read also
As we build the titanic computing power needed to simulate biology, we will unlock the third physics revolution in AI: medicine. Despite 200 years of spectacular advances, our understanding of the human body is still based on haphazard approximations that are generally correct for most patients, but probably not completely correct for a specific person. Tens of thousands of Americans die each year from reactions to medications that studies show should help them.
Now, AI is starting to turn medicine into an exact science. Instead of the laborious process of trial and error in an experimental laboratory, molecular biosimulation (precise computer models that help study the human body and analyze how drugs work) can quickly evaluate billions of options in order to find the most promising drugs. Last summer, the first drug designed from start to finish using AI entered Phase 2 clinical trials to treat idiopathic pulmonary fibrosis. Dozens of other AI-designed drugs are now entering the trial phase.
Read also
Both the discovery and testing processes will be supercharged as simulations incorporate the increasingly comprehensive data enabled by AI. Throughout history until 2022, science had determined the shape of about 190,000 proteins. That year, DeepMind’s AlphaFold 2 discovered more than 200 million, made freely available to researchers to contribute to the development of new treatments.
Much more laboratory research is needed to accurately populate larger simulations, but the path is clear. The AI will go on to simulate protein complexes, then organelles, cells, tissues, organs and (eventually) the entire body.
Read also
Ultimately, this will replace current clinical trials, which are expensive, risky, slow and statistically underpowered. Even in a phase 3 trial, there is probably not a single subject that matches another on all relevant factors related to genetics, lifestyle, comorbidities, drug interactions, and variations in disease.
Digital trials will allow us to adapt medications to each patient. The potential is impressive: not only to cure diseases such as cancer and Alzheimer’s, but to alleviate the harmful effects of aging itself.
Read also
Today, scientific progress gives the average American or Briton an additional life expectancy of between six and seven weeks each year. When IAG gives us a full grasp of cell biology, these advances will accelerate sharply. When the annual increase in life expectancy reaches 12 months, we will reach “longevity escape velocity.”
For people who are methodical about healthy habits and use new therapies, I think that will happen between 2029 and 2035, at which time aging will not increase their annual probability of dying. And, thanks to the exponential improvement in the price-performance ratio in computing, AI-powered therapies, initially expensive, will soon become widespread.
Read also
That is the most transformative promise of AI: longer, healthier lives without the limits of scarcity and fragility that have limited humanity since its inception.
Ray Kurzweil is a computer scientist, inventor, and author of books such as The Age of Intelligent Machines (1990), The Age of Spiritual Machines (1999), and The Singularity is Near (2005). His new book, The Singularity is Nearer: When We Merge with AI, will be published on June 25.
© 2024 The Economist Newspaper Limited. All rights reserved
Translation: Juan Gabriel López Guix
#Ray #Kurzweil #transform #physical #world
