Time.

It is a concept that was discussed throughout the ages. Fields such as philosophy, physics, and literature all took stabs at explaining this not-so-simple mechanism.

Many feel time is an illusion. Certainly, there is a personal touch to it. When we are heavily involved in something, time flies. The flip side, when doing something boring, such as sitting in math class, it drags on.

Humans, in spite of efforts to figure it out, have always viewed time the same. These days, with more available to us, we find time being a premium. There simply is not enough to do.

We have a growing list of emails and messages to reply to. Zoom calls are now a part of our days. There are a multitude of applications to watch movies, with a list growing that meet our desires. We bookmark (or open tabs) on dozen of articles we want to read. There are more research papers (articles) in our field.

None of this includes family or other obligations.

To the human, time is compressing. We simply have to do more with the same time. Things are different in the AI world.

It is this difference that makes AI revolutionary and why society will change. We can also see how it feeds into the Age of Abundance.

Image generated by Ideogram

Time Expands For AI And This Is What Is Revolutionary

What if we reversed this paradigm? How would things progress if time was expanding as opposed to compressing?

Naturally, it is not literally expanding. Our basis is being able to do more in an allotted period.

This is what AI does.

We start by breaking it down. At the fundamental level, AI is compute. This is the driver. Here we see an acceleration with regards to the cost/compute. Each year, processors realize gains in terms of speed. Then we add in algorithm enhancement, providing even greater efficiency. This can also be layered with improvements in telecommunication systems.

From this perspective, it is easy it grasp how these systems are improving. We see an exponential boom in "intelligence". I think most will agree with this.

Basically, more compute power provides greater output of AI systems, which are units of cognition.

There is another area where AI differs from humans. On this side of the equation, we see humans still dominating machines.

It is intent over time.

Humans have the capability to implement strategies over months. It is not uncommon for a team to work on something for 3-6 months months. Certainly, some things are missed or forgotten but each individual have the ability to maintain an intention (goal) for months on end.

Machines cannot do this. Experts are hoping that models can reach a level of 1 week. This means that while the compute (resulting in cognitive output) is exploding, intent over time capabilities is crawling along.

Even still, since the compute advancements are so large, even small gains in intent over time have enormous results.

Physical AI

Many have viewed the demonstrations of humanoid robots. While early demonstrations are impressive, there is no way anyone would believe these are at human level.

There is now a "robotic Turing Test". The idea is to have a situation where a robot took care of physical activities to the point where it is impossible to tell if a human did it or not.

For example, could you see a robot doing the dishes, moving the furniture, pickup up thrown clothes, and organizing the cupboard in a way that nobody could tell? With today's technology, it is impossible.

To even train a robot on this scale would be enormous.

That said, Nvidia is using simulations to compress time. It claims to have taken 10 years worth of physical training and simulated it in 2 hours.

Regardless of the accuracy of these numbers, even if it is off by an order of magnitude, that is still very impressive.

It also illustrates AI relationships with time. There is no way a human could compress 10 years worth of training into 2 hours. We could not even reduce it down to a year.

When we flip the compression over, how much can be learned over a 10 year period of time? For humans, it is fairly static. With AI, this expands greatly.

Another way to look at it is to take the same scenario by Nvidia, but run it on processors that are available 5 years from now. We will see a massive increase over today.

That is time expansion. It is a radically new concept as compared to most of human history. We did see some of this impact with the Industrial Revolution as it pertained to machinery. However, that pales in comparison to what we are seeing now.

For this reason, the impact of 100 years of the Industrial Revolution will be seen in under a decade.

Posted Using INLEO