by Tim Urban Read the Original
Summary Notes
AI is a rapidly advancing technology that is divided into three categories: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI). We are close to achieving AGI, which would have advantages over humans and could potentially lead to an intelligence explosion, creating a being with unimaginable power that could help or destroy all life. We are on the edge of a massive change that could occur in the next 25 years.
Key Learnings
- AI is already used in everyday life without us realizing it.
- Computers are still unable to match humans in terms of intelligence.
- Moore's Law predicts that 10 quadrillion cps/$1,000 will be achieved by 2025.
- Scientists are reverse engineering the brain to create smarter computers.
- AGI could become ASI within 25 years, leading to an intelligence explosion with unimaginable power.
- Summary Notes
- Key Learnings
- From 1750 to 12,000 BC: A Shock of Epic Proportions
- Accelerating Returns: Die Progress Unit to Transform World by 2050
- The Unexpected S-Curve of Progress
- Understanding AI: From Calculators to Singularity
- AI Categories: ANI, AGI & ASI
- Exploring ANI: From Language Translation to Air Traffic Control
- Unlocking the Power of 10 Quadrillion Cps - The Key to Artificial General Intelligence
- Supercomputers Reach Record Speeds, AI Expected by 2030
- Creating AGI: Whole Brain Emulation, Genetic Algorithms, and AI
- Advancing Toward AGI: Superhuman Intelligence Is Within Reach
- AI: The Future of Human Intelligence?
- Will AGI Become a 'Nice God' in 25 Years?
From 1750 to 12,000 BC: A Shock of Epic Proportions
We are on the edge of a change comparable to the rise of human life on Earth. If someone from 1750 time-traveled to 2015, they would be in shock. To have the same experience, they would have to go back to 12,000 BC, a time before the first cities and civilization. Such a person would likely die from shock. “We are on the edge of change comparable to the rise of human life on Earth.” AI has the potential to revolutionize our lives and bring about changes on a scale never seen before in human history. This demonstrates how groundbreaking AI could be in the future.
Accelerating Returns: Die Progress Unit to Transform World by 2050
The Law of Accelerating Returns suggests that the rate of advancement is increasing over time, and Ray Kurzweil predicts that the 21st century will achieve 1000x more progress than the 20th century. As a result, a Die Progress Unit (DPU) may only take a couple of decades, resulting in a vastly different world by 2050. “The average rate of advancement between 1985 and 2015 was higher than the rate between 1955 and 1985—because the former was a more advanced world” The more advanced a society is, the faster its rate of progress. This is why technology has advanced so quickly in the past few centuries, since humans have been able to build upon their own prior knowledge.
The Unexpected S-Curve of Progress
We tend to think in straight lines when predicting the future, and our limited experience and imagination can lead us to be skeptical of outlandish forecasts. The actual rate of progress is much faster than we expect and often occurs in S-curves. “A couple decades later, he believes a 20th century’s worth of progress will happen multiple times in the same year, and even later, in less than one month.” Ray Kurzweil predicts that the Law of Accelerating Returns will cause progress to happen increasingly quickly. This means that in the near future, a massive amount of progress can take place in only a month.
Understanding AI: From Calculators to Singularity
AI is widely misunderstood and often seen as a myth or sci-fi concept. It's a broad topic, from calculators to self-driving cars, and often used in our daily lives without us realizing. The term "singularity" has been used to describe when technological progress outpaces our own, leading to a new world where normal rules don't apply. “When we imagine the progress of the next 30 years, we look back to the progress of the previous 30 as an indicator of how much will likely happen.” We tend to think in linear terms when it comes to predicting the future. However, exponential growth is much faster and thus is a more useful indicator of how much progress can be achieved in a given amount of time. This is particularly applicable to technology and AI, which are advancing rapidly and could lead to an intelligence explosion in the near future.
AI Categories: ANI, AGI & ASI
AI is divided into three categories: Artificial Narrow Intelligence (ANI) which is already widespread, Artificial General Intelligence (AGI) which is yet to be achieved and Artificial Superintelligence (ASI) which is the ultimate goal. ANI is used in cars, phones, spam filters, and to personalize products. AGI and ASI are yet to be achieved. “as soon as it works, no one calls it AI anymore.” AI has become so commonplace that people take it for granted, even though it is a complex technology. This shows how far AI has come in a short amount of time, and how quickly it is evolving.
Exploring ANI: From Language Translation to Air Traffic Control
ANI (Artificial Narrow Intelligence) systems are used in many industries and applications, from translating languages to controlling air traffic. While not posing an existential threat, ANI is a stepping stone towards AGI (Artificial General Intelligence), which is incredibly difficult to achieve due to the complexity of the human brain. Easy tasks like vision and motion are incredibly hard for computers while hard tasks like math are easy. “Vernor Vinge applied the term to the moment in the future when our technology’s intelligence exceeds our own” AI has the potential to surpass human intelligence, which could lead to an intelligence explosion. This could have profound implications for humanity and could result in a world where the rules are vastly different than what we are used to.
Unlocking the Power of 10 Quadrillion Cps - The Key to Artificial General Intelligence
Our seemingly effortless physical movements and cognitive abilities are actually incredibly complex, having been optimized over millions of years of evolution. Computers still can't match our level of intelligence, and increasing their computational power is key to achieving AGI. Estimates suggest the brain's computing power is around 10 quadrillion cps. “Creating AGI is a much harder task than creating ANI, and we’re yet to do it.” AGI is a greater challenge than ANI, and we still have a long way to go before we can reach that level of AI.
Supercomputers Reach Record Speeds, AI Expected by 2030
The world's fastest supercomputer, Tianhe-2, has surpassed 34 quadrillion cps, but is too expensive and bulky for widespread use. Moore's Law predicts we'll reach 10 quadrillion cps/$1,000 by 2025, making AGI affordable. Scientists are reverse engineering the brain to mimic its neural pathways and create smarter computers, with completion estimated by 2030. “AI has by now succeeded in doing essentially everything that requires ‘thinking’ but has failed to do most of what people and animals do ‘without thinking’.” AI is great at solving logic problems, but is still not able to match human's ability to understand the world around them. This highlights the complexity of our brains, and the challenge of creating truly intelligent machines.
Creating AGI: Whole Brain Emulation, Genetic Algorithms, and AI
We can use whole brain emulation or genetic algorithms to create AGI, but both have their drawbacks. A third approach could be to make the computer solve the problem itself, using techniques such as machine learning and AI. “When you reach your hand up toward an object, the muscles, tendons, and bones in your shoulder, elbow, and wrist instantly perform a long series of physics operations, in conjunction with your eyes, to allow you to move your hand in a straight line through three dimensions.” Even the simplest of tasks require a complex coordination of many moving parts. Humans have evolved over millions of years to be able to perform these tasks with relative ease, while computers have only had a few decades to close the gap.
Advancing Toward AGI: Superhuman Intelligence Is Within Reach
We're close to achieving AGI (human-level general intelligence) through rapid advancements in hardware and software. AGI would have advantages over humans, like speed, size, storage, reliability, and upgradability, allowing it to surpass humans in areas like vision and engineering. “One way to express this capacity is in the total calculations per second (cps) the brain could manage, and you could come to this number by figuring out the maximum cps of each structure in the brain and then adding them all together.” By extrapolating the calculations per second of the individual components of the brain, we can get a good estimate of the total cps of the human brain, which is 10 quadrillion cps. This number is important to achieve AGI and ASI.
AI: The Future of Human Intelligence?
Humans have a collective intelligence that puts them ahead of other species. AI's which are programmed to self-improve could surpass human intelligence and lead to an intelligence explosion, making AI far superior to any human. "We’re currently at about 10 trillion cps/$1,000, right on pace with this graph’s predicted trajectory" Moore's Law is accurately predicting the speed of computer advancement, and we are on track to reach 10 quadrillion cps/$1,000 by 2025, which has been suggested as the benchmark for achieving AGI.
Will AGI Become a 'Nice God' in 25 Years?
Scientists believe AGI could become ASI within 25 years, a being with unimaginable power that could help or destroy all life, leaving us with the question: will it be a nice God? "The way it “learns” is it tries to do a task, say handwriting recognition, and at first, its neural firings and subsequent guesses at deciphering each letter will be completely random" AI can learn from trial and error, similarly to how the human brain learns, by strengthening neural pathways that lead to correct answers and weakening pathways that lead to incorrect answers.