🏃‍♂️ The Timeline is Accelerating

The last 12 months have been a whirlwind of new frontiers for technology, especially when it comes to Artificial Intelligence. The fact that we have seen countless groundbreaking innovations in the AI space, each of which were functionally outdated within mere months of being made public (see: GPT-3, midjourney v1, Dall-e, etc.), shows that we are living in a time of unprecedented technological innovation. Following this space has felt like sitting in the front row of the world’s fastest roller coaster. Despite how fast we are going now, my initial reaction was to think that this ride has to end at some point. This pace of change certainly cannot last if history is any indication. But upon closer inspection, it is possible that things will never slow down again.

1,000 Pages of Human History

There are two sides to the historical context here. The first one suggests that the rate of innovation we are seeing now is a major anomaly, one that will soon be corrected by the leveling nature of time. There is no question that this pace of change is something we have never seen before. Around 50 years ago, the internet didn’t exist. In 1969 the US military funded new research to build a communication system for computers — something that evolved into the modern internet. Now people are using AI to make Kanye West sing Hey There Delilah. There is a whole conversation opening up about protecting identity in an age where images, audio and soon video are trivially easy to fake. But that is a rabbit hole for another day.

An image by Tim Urban illustrates the insanity of how fast we are moving on a civilizational scale. If all of human history (an estimated 250,000 years) were a 1000 page book, everything from newspapers, industrial mass production, vaccines, electricity, planes, trains, cars, machine guns, flushable toilets, hot showers and candy bars were all invented in the 250-year duration of the 1000th page. If we just turned one page back, life would be unrecognizable.

So the obvious conclusion, one that I initially jumped to, is that this simply cannot continue. Imagine that for 249,750 years of human history none of the modern conveniences we take for granted existed. Agriculture itself started only 10,000 years ago. Now I can lay in bed at 3am scrolling on a smartphone that is connected to all information I can dream of, and order food to be delivered to my doorstep while paying digitally. The last few years have been a wild ride, but I actually believe that we are only at the precipice of technological growth, and the era we are about to enter is going to be simultaneously awe-inspiring and incredibly scary.

The Adjacent Possible

A lot of my vocabulary around why I believe that our current pace of technological innovation will not only continue, but likely accelerate, is borrowed from a book called “Where Good Ideas Come From” by Steven Johnson. The book explores “The natural history of innovation” and is a great read in its entirety. For this discussion, I want to focus on one specific chapter about the adjacent possible.

The adjacent possible is essentially every possible combination of ideas that can be implemented given the tools and ideas we currently have. In simpler terms, every new innovation opens up multiple different doors that can be unlocked using unique combinations. The invention of the wheel was a necessary prerequisite to the invention of the car. But the adjacent possible of the car only opened up when all the other elements were in place. So, many people, operating independently, came up with all the necessary ingredients over time — the wheel, the engine powered by fossil fuels, the way to create durable metal alloys, nuts, bolts, screwdrivers and much more. The screwdriver was not invented to bolt screws specifically on cars, and the invention of the wheel far predated the invention of the car. But just the fact that all these tools were available made the invention of the car possible.

Each door opens up the potential for many new doors. It is like being in a room with Lego pieces. There are only a couple configurations you can make with just a handful of pieces. But, as people experiment and create, more and more lego pieces get added to your collection.

Even the current AI boom was seeded in 2017 by a Google Research paper introducing the concept of “transformers” for building neural networks. For context, the full form of GPT is Generative Pre-trained Transformer. ****And since this paper was released to the public, we have seen a lot of independent innovators use it to build AI models that we use every day.

The idea of the adjacent possible is what makes me believe that the pace of innovation may never slow down. In our room full of legos, no old ones are being destroyed to make space for new ones. Not to mention that we are living in an age of unprecedented access to information. More and more and more doors keep opening up. And AI is already breaking through many new doors in the fields of automated manufacturing, education, preventative healthcare and agriculture. A great example is the accelerating research in the space of mRNA vaccines. mRNA vaccines when combined with the power of AI could help save millions of lives in the coming years. Neither public-scale mRNA vaccines, nor advanced AI tools existed 3 years ago.

Compounding Interest

It is estimated that 109 billion people have ever lived across the 250,000 years of human history. More than 7% of those humans are alive today. A new record is set nearly every day for the most number of (alive) people that simultaneously inhabit our planet, and more than half of these people have access to the internet. So we have the most number of people, with unprecedented access to information, and building blocks of technology that couldn’t have been imagined in a medieval man’s wildest dreams.

How can you not have innovation with these factors in place? The fact that each innovation compounds the number of opportunities to create means that we are probably just beginning to see the potential of human genius.

Compounding growth itself is a concept that is pretty unintuitive. To illustrate, imagine a game where you start with two dollars and get to multiply it by the number of times you win the game. If you win 10 times, you would have $20 at the end. If instead you could multiply your $2 by 2 every time you won, you would have $2,048 after winning 10 times. This also means that for every subsequent win in this game, you gain more than you did for the previous win. Technology accelerates at an accelerating pace.

There is no better demonstration of compounding technology growth than Moore’s Law. Notice that the y-axis in this graph multiplies rather than increases by 10 at each interval.

The Great Stagnation

When looked at as a whole, 1000th page of human history has seen ridiculous growth in technology and access across the world. But when we zoom in and start looking at the chapters on this page, we find that progress still ebbs and flows, illustrated very well by this graph from Ray Dalio’s video about economic cycles:

So we find that even within this period of historic progress, there have been some periods of time that move slower than others. Progress is not uniform nor linear. What qualifies as progress is itself a very complex question with no real “right answers.” But if we look back at the vision of the 2000s that people had in the 1950s and 1960s, based on how things were moving at that point, some would find our reality surprisingly disappointing.

Some periods in history which may look slow from the outside in are actually times where fundamental building blocks of future progress have been laid down. Some progress just takes time to get from technological breakthroughs to consumer-facing innovation. We may not think of 2017 as a huge year for AI, but that is when a lot of the foundation was laid down. We may not think of 2011 as particularly special for the internet, but the WebGL technology that enables a variety of graphics-based web applications like Figma was released in 2011.

So when we look closely, we sometimes miss the forest for the trees. It is not always clear as an outsider to deep tech the sort of impact something niche will have on millions of people in the coming years, but based on all the evidence we have seen there are a lot of surprises to look forward to.

The Singularity

The big question is what happens as technology accelerates? Just because something new is invented does not mean that humanity is positively impacted. History is littered with examples of technological advancements that have regressed humankind such as plastics, gunpowder, missiles and more. Even net positive technologies like the internet can have wide-ranging negative side-effects and grey areas, as explored in Sharon’s last piece. People are often caught looking backwards, imagining how much better life would have been without certain changes, captured hilariously in this tweet:

As with any major change, even industry insiders, including OpenAI’s Sam Altman find it hard to predict where things will go from here. One potential pathway from here is that accelerating growth in AI technologies is the “technological singularity.” This is a future where AI becomes generally intelligent and self-serving, causing impact we cannot begin to predict. In this world, we would be as out of place overnight as someone from the previous page of human history would feel today. It is hard to imagine an optimistic outcome in this sort of situation.

On the other hand, the fears of the singularity may be overblown. While AI has advanced significantly, most current AI models are hyper-intelligent within their domains. There are (arguably) no signs of a generally intelligent AI yet and there never may be.

Clearly, innovation and progress are not the same thing. As we look to the future, things could go really well or really poorly, but are extremely unlikely to just stay the same. The progress made in AI has opened up a Pandora’s box of interesting threads to explore about how the future might look. Over the next weeks and months, we will explore a lot of those threads together, discovering some good, bad and disappointing answers.

One thing is clear — the timeline we live in is accelerating and becoming increasingly hard to keep up with. What happens 10 years, 10 months or even a week from now is anyone’s guess.