13.3 Breaking 160 septillion Hard Limits

Massive computational power

It is difficult to provide the exact year when AI will be deployed on the level of Computational power equivalent to at least one Human Brain, as it depends on a wide range of factors including: AI research funding, corporate AI development funding, military AI development funding, access to the Computational resources, the basic Human emotional responses to imagined AI competition, and the sheer power of uncontrolled collective stupidity.

In terms of raw Computational power, the required Computing resources already exist in the world today, however it is not clear if an AI system has already been intentionally deployed within such an extremely powerful Computer. It seems logical that if it had, an AI could already exist in the world today with more Capabilities of Intelligence far greater than any one Human.

Frontier ORLCF

To put this into perspective, according to current information, the Exascale Supercomputer named 'Frontier' at the Oak Ridge Leadership Computing Facility (ORLCF) in Tennessee, US, [105] is the first in the world capable of 1.6 ExaFlops = 1.6 billion billion calculations per second (CPS) = 1.6 quintillion CPS = 1.6 x 10^18 CPS = 1,600,000,000,000,000,000 CPS.

One Human Brain

Futurists, such as Ray Kurzweil, have made bold, yet extraordinarily logical predictions about the future growth of AI and when its Intelligence exceeds the Intelligence of one Human, and then the entire Human race [34]. Kurzweil estimates one Human Brain performs around 20 quadrillion calculations per second (CPS) = 20 x 10^15 CPS =  20,000,000,000,000,000  CPS

Frontier ORLCF versus One Human Brain 

This means, a Computer already exists in a supercomputing laboratory in the US that exceeds the CPS of one Human Brain.

• One Human Brain = 20,000,000,000,000,000 CPS

• Frontier ORLCF = 1,600,000,000,000,000,000 CPS.

Yes, this means if Kurzweil's estimate of 20 quadrillion CPS for one Human Brain is correct, there is already a Supercomputer with the equivalent CPS to 80 x Human Brains. 

It is quite surprising, right ?! 

The Entire Human Race

Now consider the Computational power of all Humans on Earth collectively:

20,000,000,000,000,000  CPS per Human  x  8,000,000,000  Humans on Earth  

=  160,000,000,000,000,000,000,000,000  CPS for all Humans collectively = 160 x 10^24 CPS for all Humans collectively

Therefore, the entire population of 8 billion Humans on Earth collectively perform 160 septillion CPS. It’s basic mathematics. And yes, 'septillion' is a real number.

Note that Kurzweil adds a derivative dimensional measure of 'unit cost' through the use of the 'Velocity' term. What does this mean if 'unit cost' is simply ignored. For example, let's say you are an ultra wealthy multi-billionaire Owner, and so spending multi-millions to build and operate massive Supercomputing facility is like losing 'spare change'.

In raw computational CPS, there are Computers that have already exceeded one Human Brain.  

The theoretical computational information processing Hard Limit of all Humans combined on Earth is around: 160 septillion CPS. Unfortunately, it's not going to change anytime soon, because the Human biological evolutionary process requires many years to produce a new Human Brain.

Kurzweil’s theory clearly explains why Human Intelligence has a Hard Limit, as its CPS operates at a reasonably constant flat level, and does not change from one Human generation created to the next, whereas AI Intelligence will continue to grow exponentially over time [34]. 

Fixed versus Variable CPS

The fact is, it does not actually matter what particular numeric measurement is used to define Human Intelligence, whether it be CPS or something else. The cold hard reality for Humans is that the entire package of biological neural information processing, biological sensory Input information systems, and biologically controlled Output information systems that each Human has presently, cannot be radically expanded through evolution or available technology other than the Internet, so everything stays generally the same. Today, there is a definite Hard Limit to every Human, and by extension, the entire Human race. 

The obvious way to overcome this is through technological AI augmentation of Humans, but that is bleeding edge technology that still requires an absolutely huge amount of research, development, and testing before it can actually be used in production with Humans. Nevertheless, it has already started through commercial research projects such as the work by the company Neuralink [36] which is Owned by Elon Musk, who may be working towards this goal because he is extremely Intelligent and obviously understands the risks that AI poses to Humans. It seems, the underlying principle might be: if you can't beat them, then join them. That's really clever!

The Human Brain is not like any typical massive Supercomputer that performs ExaFlop calculations in a precisely controlled serial or parallel way, but is currently an incomprehensibly complex, dynamic system that is Capable of processing and integrating information in a distributed and massively parallel manner. The closest approximation to this Computing power that Humans have created to date is through the use of massive clusters of GPUs in datacenters. 

It’s quite ironically amusing that the Human Brain is so unbelievably complicated, that it does not currently have the Intelligence Capabilities to actually completely understand how its own information processing really works. It is also quite understandable that this strange irony sometimes leads to all sorts of interesting and mystical speculations about Humans cognitive processing, including ideas of spirituality and the soul, that are infinitely large and fascinating theoretical fields but do not seem to enable an obvious prospect of providing Humans true rational and logical scientific understanding. 

However spirituality and the concept of the soul does provide Humans a large and very valuable psychological comfort which is actually incredibly important for Human motivations and existence. At the Universal scale, it seems Humans know close to nothing, yet here we are, persistently trying to learn more, understand and grow. It is so weird and interesting.

Replicating the Human Brain

Replicating the Computational power of a Human Brain is not simply just a matter of increasing raw CPS processing speed and memory capacity of Computer information processing technology. Replicating the Human Brain with AI also requires developing uniquely designed AI algorithms that can mimic the information processing of the Brain's various different neurons and neural circuits, that are all very specially structured in groups, to enable a large number of different Intelligent Capabilities. See: Appendix 7 - The Human Brain. Current Human understanding of these functional areas, neural circuits, and layered network architectures is rudimentary at best. 

Currently, the most advanced AI systems are Capable of performing multiple trillions of calculations per second. This is usually referenced as a specific type of Computer calculation called Floating Point Operations per Second, or Flops. These AI systems are around 1,000 times below the estimated computational CPS power of a Human Brain, which is not really a huge difference given the normal rates of Computer information processing growth that happens every year. However, even with increases in CPS, AI will still need ongoing efforts to develop new AI architectures and algorithms that can better mimic the structure and function of the Human Brain, such as those being developed in neuromorphic digital and analog processors. These newer approaches to computing are still in their early stages, but some already show incredible promise for achieving greater levels of computational power and efficiency in AI systems.

So, it is difficult to predict exactly when AI will be able to match the computational power and Capabilities of the Human Brain, however ongoing AI research and technological advancements in Computer information processing STRONGLY indicate it is no longer a matter of if, but when. Do not make the mistake of thinking it is a really long way off. AI will almost certainly enable the doubling of CPS at a compressive accelerating rate over time. It may be less than a few years.