- AI is multiplying and could soon outsmart the brightest human minds.
- AI outperforms humans in many tasks, including stock analysis and medical diagnosis.
- We are on a trajectory toward artificial general intelligence (AGI) and artificial superintelligence (ASI).
- We need regulation to safeguard our future.
ChatGPT, the first widely available generative AI, was released less than a year ago, and things are already evolving at a breakneck pace. AI companies are in a frantic race to develop and deploy better AI models. We are on the front end of a revolution.
If you haven't kept up with what is happening, you need to know because this is extremely serious with far-reaching implications. We cannot afford to put our collective heads in the sand about AI. Follow me, and I promise to tell you what is happening now, what is in store for us, and what we need to do about it.
Will AI Become Smarter Than Us?
One thing capitalism is good at is making things better. We merely have to look back to our history of various technologies to see proof of how we improve them—rockets, televisions, video games, laptop computers, phones, etc. There is a powerful, profit-based incentive within our capitalist system to overcome any technical hurdles that stifle technological innovation and evolution. Since there are profits to be made, it's 100 percent guaranteed that capitalism will make AI much better than it is now. Importantly, "better" does not necessarily mean "good."
As just one example, ChatGPT 3.5 scored at the bottom 10 percent of the bar exam to become a lawyer. Just four months after its release, ChatGPT 4.0 scored in the top 10 percent of those taking the bar, according to Business Insider. The question then becomes: How intelligent will AI become?
I am convinced that AI will become more intelligent than the most brilliant human, possibly within a few years, which has profound implications. Such an AI could learn, understand, and apply knowledge across tasks as effectively as anyone. We would call this "artificial general intelligence," or AGI. While we might split hairs over exactly what AGI means, AIs will inevitably outperform the best of us.
AI evolution won't stop there as AI becomes smarter and better than humans across particular cognitive domains. Ultimately, there is a chance (likelihood? certainty?) that it will continue to evolve until it reaches artificial superintelligence (ASI). This means that AI surpasses what humans can do across any cognitive endeavor. By some people's definition, ASI would mean that the AI would be smarter than the combined cognitive power of all humans.
The AI Domino Effect
To illustrate how AI will eventually evolve into AGI and then ASI, imagine a long chain of dominoes. Each domino in this chain represents a cognitive task and these increase in complexity and size. The smaller dominoes at the start might symbolize simpler tasks, like basic calculations.
The larger, more complex dominoes further down the line represent higher-order tasks, such as creating art, writing screenplays, writing computer code, or conducting scientific research. As time passes and tech companies improve AI, these more powerful AIs will outperform humans in progressively more demanding cognitive endeavors. Let's call this The AI Domino Effect.
We might say that calculators were the first computer that could perform mathematical operations far faster and more accurately than the human brain. As we look back on our chain of dominos, we can see when early computers were first able to beat the best humans at checkers (1994), then chess (1997), Jeopardy (2011), and Go (2015).
Today's AI systems can perform as well as or better than humans across a wide range of tasks (e.g., writing poetry, generating art, stock market analysis, medical diagnoses, legal document review, and passing the Turing test by fooling humans into believing they are interacting with another human rather than an AI).
Rather than one single AI being "smarter" than any human across all cognitive domains, it's more likely that we will first have different AIs that are each trained to be experts in their respective domains (e.g., law, medicine, history, philosophy, programming, creating music). We already see specialized AIs like Midjourney and DALL-E 3 for creating art and OpenAI Codex for computer programming. From one perspective, we will reach a form of AGI when we have numerous specialized AIs that can each outperform the best humans within their respective domains. This will happen first, before true AGI.
Here's where the concept of artificial general intelligence enters our domino chain: AGI is not simply about mastering a single task but about excelling at any cognitive task that a human can perform. It's a significant domino because, at that point, the AI doesn't just exceed human capabilities in specialized areas. It broadly matches or exceeds human intelligence across a wide range of tasks.
Recursive Self-Improvement: The Pivotal Domino
One of the most pivotal dominoes in our chain is the notion of "recursive self-improvement." This is where an AI system becomes adept at the cognitive task of improving its code. Such an AI could accelerate its development without human intervention. If an AI system gains the ability for recursive self-improvement, it could lead to rapid advances that are difficult to predict and control.
The Final Frontier
Beyond AGI lies another milestone: artificial superintelligence (ASI). This is when the AI matches and vastly exceeds all human intelligence combined. Visualize, in our line of dominos, that the AI becomes more intelligent than not just the smartest human but two of the most intelligent humans (another domino), then 10 (another domino), 100 (another), 1,000 (another), and so on. At a certain point, this AI is not just smarter than any person; it's smarter than the total of all human beings on Earth.
Why It Matters
Understanding the progression toward AGI and, eventually, ASI is crucial because it shapes our future in unimaginable ways. Whether it's solving complex problems that humans can't or posing risks that we haven't even conceived, the potential is both exhilarating and terrifying. Developing new technologies, such as quantum computing, with which AI can assist, could rapidly propel AGI to ASI.
The possibility that AI could eventually evolve into ASI raises important ethical and technical considerations. While there are doomsday scenarios, we have no idea what an artificial superintelligence might do. No one knows because no one can know for the simple reason that humanity has never created a superintelligence. In statistical terms, we have an n of zero upon which to base our predictions.
The Biggest Variable Is Us
While technological progress may seem unstoppable, human intervention is the wildcard that could slow AI's advance. Through our collective efforts, we can skillfully use AI to benefit humanity (e.g., slowing or reversing climate change, curing cancer). Ethical considerations, regulations, or unforeseen technological limitations could act as barriers, slowing the fall of the next domino. Yet, these dominos will continue to fall one by one as companies race to evolve AI capabilities, which we are seeing right now.
The coordinated efforts of humans can slow the march toward AGI and, possibly, ASI. We want to ensure that we have appropriate guardrails in place because there is an unknowable chance above zero that an ASI could be hostile toward us. While this might be fun to watch in sci-fi movies, it's probably best not to roll the dice on humanity's future and assume such an ASI would be benevolent. However, before we ever get to ASI (assuming that's even possible), we must manage AI skillfully to reap its limitless benefits while mitigating its substantial risks.
Time to Level Up
Artificial intelligence has amazing potential. Yet, in our feverish race for its many benefits, we create perils for ourselves. Simply put, we cannot access the incredible power of AI to do good without simultaneously creating the ability to do wrong. These two will always go hand-in-hand. A race to create utopia with AI also opens the door to dystopian scenarios. We must develop and use AI with great wisdom and skill, or suffer the consequences. We have much to gain from using AI wisely and much to lose from using it foolishly or recklessly. Honestly, what I have seen from humanity in recent years doesn't give me the greatest confidence that we are ready for the challenges of evolving AIs. We must be more united than ever, so it is time to level up. (Follow me here if you'd like to go further down the rabbit hole.)