We are embarking upon a prolonged period of sustainable change. This is moving at such a rate which is making our present metric obsolete.
Many are trying to make sense of what is taking place. However, the reality is, quite frankly, too difficult for most to absorb.
If we look at just generative AI, we see a path where progress is being made at speed never witnessed before. When this is extrapolated over 10-15 years, the impact upon society will be profound.
In this article I will discuss the extreme pace of the world of bits and then compare that to atoms. For much of history, we see very little change in the latter.
This is about to be completely upended.
Image generated by Ideogram
Digital Warp Speed.
Do you think generative AI is moving at an accelerated pace? Are you one who thinks it will slow down?
Here are a few stats that might make one rethink that:
- algorithms improve enough to cut compute times in half every 16 months as measured on Imagenet
- cost of compute is doubling every 2.5 years. That means twice the compute for the same money
- total compute (CPU) is doubling every 6 months
- spending on compute models is doubling every 2 years
Groq state that a report by IDC, the global digital data volume is expected to grow from 64.2 zettabytes (ZB) in 2020 to 175 ZB by 2025, representing a compound annual growth rate (CAGR) of 44.3%. Gartner puts this figure between 30%-40%.
It is easy to see from these numbers why the digital world is expanding so rapidly. We did not even include the amount going into things such as inference, which could outpace the processing used in training neural nets by at least an order of magnitude.
Keep in mind that AI can be basically summed up as data wrapped in algorithms and powered by compute.
As we can see, each of this is advancing rapidly.
The world of bits is one of exponential. When focused upon a specific technology, such as generative AI, we see how a slowdown is not likely happening anytime soon.
Of course, as they say, you can't eat a semiconductor. Therefore, let us take a look at the world of atoms.
Physical World Acceleration
The physical world changed a great deal. In fact, it transformed as much as the digital, only over a longer period.
Go back a few hundred years. How long did it take one to cross Italy. Remember the days of horse and carriage? Compare that to now. We not only have automobiles, high speed trains, but also planes. Actually, the International space Station can circle the planet in about 90 minutes.
Consider digging a hole.
How many people with shovels would it take to dig an Olympic size swimming pool? I would be the answer is many and it would take a number of days.
Let us consider how two people with backhoes could do it. My guess is they could get it done in a day.
We have cranes, equipment in factories, and other physical items that accelerate tasks. Have you used a manual drill? Of course not. We have power tools.
All of this sped up transportation, manufacturing, and construction. However, it still lags behind, by a large degree, the digital world.
Closing the Gap
For the purpose of this article, we will use the digital world as a baseline 100. When comparing the physical world to that, maybe we are at a 5. This is the spread between the acceleration of the two realms.
This is going to close. The digital is about to be released into the real world, in the form of embedding. To put it simply, the robots are coming.
Naturally, we are never going to see the world of atoms match the digital. A 100 will never be attained.
That said, what will robots push this number to? When comparing economic productivity, i.e. output, it usually takes a new being (a baby) about 2 decades to make a contribution. We also have the manufacturing cycle of one (for the most part) every 9 months.
With robots, the "reproduction" time is much smaller. One scaled, it is likely a new robot will come off the line every minute or so. When coupled with AI, the learning process is simply an updating to the latest software.
This means the learning curve is next to nothing.
Hence, we are likely looking at a baseline of 25 or maybe 30. Naturally, these are just estimates of raw numbers that are immaterial. The point is we can easily see a 5x (or more) in the acceleration of our physical world.
What does this mean for output? At the same time, what does it mean for society?
If we take this concept and apply it towards 2040, things get very interesting. People like to debate whether we will have AGI, ASI, or some other form of intelligence. That is great for discussion but really matters little for the impact.
Whatever it is called, we are about to embark upon a massive acceleration of both the digital and physical worlds, The latter will always lag the former yet that it about to get a major boost.
By 2030, this will likely be in full swing.