Why the IT Devices Of the Future Have to Ditch 1 and 0 Logic For Good

It’s time to throw the idea of arithmetic logic unit (ALU) into the trash, because it is the reason our tech cannot match up with Ridley Scott movies.

What is ALU anyway?

If this is the first time you see the term ALU, check Wikipedia or Study.com section on it, and come back to reading.

Every article about arithmetic-logic units (just like this one) boils down to the fact that ALU does a pretty basic job. That doesn’t mean the device itself is primitive or worthless — it’s just that the operations it performs are quite primitive.

So, what’s the problem with it?

Because today’s computers run on the logic that is not optimal, and that stumbles our progress. No one actually needs the set of commands used in ALU. You don’t use binary system for decimal arithmetic, let alone calculus — it would be an extremely inconvenient way to solve “7 × 8”, you must agree. And if it’s not “7 × 8”, but adding or, God forbid, multiplying billions, you will never be able to replicate the computer’s calculation logic, even if you wanted to — a lifetime won’t be enough.

For such operations people have come up with a very convenient column method for addition, subtraction, multiplication and division — the method that is taught in elementary school. Some prodigies can add, multiply, and divide any number in their head instantly, but our guess is they never do it the way a computer would.

But computers have no other way of doing it. And that’s a problem.

I still don’t get it…

ALU is about using a computer to execute a required operation in a way that is convenient for the computer itself.

And for some reason, this has been considered the only possible way for too long now. That’s because in the beginning of the 20th century someone really clever invented the flip-flop circuit based on the two-hundred-year-old Boolean algebra logic, which is based on the binary system invented by Leibniz more that three centuries ago, and Leibniz has borrowed the concept from the I Ching (Chinese Book of Changes), which is at least 2700 years old.

Few decades after the invention of the flip-flop, the first computers emerged. No other system except for the binary was suggested, which was fine: billions added and multiplied correctly, and at the time it was the only thing electronic computing machines had to compute. For that reason, most devices around us today still read and store data as 0 and 1.

Binary logic seemed convenient in the middle of the 20th century. But today’s computing challenges are not quite the same, are they?

The amount of data processed by computers now is absolutely insane — think, for example, of the Full Self-Driving system, a massive challenge that the best minds of big tech industry are yet to solve.

There’s no way ones and zeros are the only, or the best way to compute on such scale. It seems we got used to these elementary data units so much that we continue to use them despite the obvious inconvenience. There must be another option.

To say that the whole IT industry is stuck in the past, and is trying to build the tech of the future on the logic of the past is a bold statement. But it is true.

Zeros and ones are like the alpha and omega of IT — they set the trend, define the computer architecture, offer the one and only solution to any task, and, eventually, define how any task is thought of.

Zeros and ones influence the end result in a way that’s often called “major impact” — those two tiny entities determine the direction for information technology development, as well as its limitations. It seems like this never bothered anyone (and still doesn’t) — partly because the scope of the practical problems that can be solved using this logic is impressively big.

The steam ALU engine

You may ask yourself: are steam engines worthless? You’re right, the question is silly. The invention of the steam engine was a major breakthrough in the human history, and influenced the human civilization much like Promethean fire, wheat domestication, metal tools and weapons, soccer, penicillin, or emojis (which made it possible to express emotions without actually expressing them).

And yet no one in their right mind would claim that steam power is the most efficient, ergonomic and otherwise optimal way of solving modern practical problems, despite the fact that steam engine once propelled automobiles, locomotives and machinery, being one of the key symbols of the industrial revolution.

It’s quite obvious that a steam engine would not be enough in the modern days. However, it is possible to imagine every single device around us today running on steam. It might be even possible to reach space using steam power, but what’s the point? Even the internal combustion engine is no longer enough for one of the most conservative industries in the world — the automotive industry.

Nevertheless, industry giants nowadays are still trying to make fundamental innovations using a technology that almost a century old — flip-flops with their ever-lasting ones and zeros. Companies, R&D teams and scientists all over the world have to solve complex contemporary challenges while being limited by the same inputs, trying to address fundamentally new challenges in fundamentally old ways. It is, indeed, like launching a steam-powered moon rocket — possible, but at what cost?

As the tasks become more complex and challenging, tons of financial and intellectual resources are flushed away solving them just because of how inconvenient 0 and 1 logic is. No one thinks about it, and no one counts the costs, nor calculates project efficiency.

A huge bubble of extensive IT development is constantly being inflated. Many tasks are being solved just for the sake of solving, not to find the solution that can scale into the future (although it’s often impossible due to retro logic, and all the money in the world won’t fix that).

Let’s face it: what once was a high-speed highway, has now turned into a country road that leads into a muddy swamp. Modern technology has exhausted itself. Sure, flip-flop circuit and ALU were great inventions for the 20th century, just like steam engine was for the 19th. Sure, you can still invent devices, develop applications and perform tasks with AI & ML. But building today’s tech on ALU is like trying to conquer space by putting hundreds of the most powerful steam engines together.

Transistors have certainly evolved — from huge tube-shaped apparatuses to nanoscale semiconductor devices, — but their internal logic remains the same.

Companies are racing for processor performance by increasing the number of transistors per unit area, but that’s not what progress looks like. It seems that mankind has come to a breaking point that cannot be overcome by conventional methods.

Now we have zeros and ones invented by Leibniz (who picked it up from the ancient Chinese), the Boolean algebra of logic, and that’s pretty much it — sure, the number of internal operations increased in ALUs compared to the original flip-flop. Modern ALUs are integrated into CPUs, successfully solving their amazingly primitive tasks…and nothing more.

Today’s computers are bound to translate the simplistic algebraic logic to complex mathematics, because that’s the only way today’s IT specialists can formulate tasks and control the process. They call it ‘coding’ for a reason.

So we need to build a better ALU?

Yes, but not just a better ALU. We need an ALU that is built and operates more like a human brain, forget the 0 and 1.

Just like our brain never uses advanced math or binary logic to make decisions (‘computations’ to keep the comparison going), we need a device that would compute without using the algebraic logic or calculus.

We need something else — not ALU, but a unit of some kind, a device that could acquire information. Let’s call this device a perceiver for now, referring to the verb “perceive” in the same way as “processor” refers to the ‘process’ verb.

Why do we need a new word for it? Because using the term information processing ignores quite a few meaningful levels of working with data. Information has to be collected as is, before being processed, calculated, or used in decision-making and other things our brain is so good at.

Our hypothetical device must be able to receive and process data from the initial level, images as images, sounds as sounds, and so on — not in the form of ones and zeros. Just like our brain does. Such a device (aptly named “perceiver”) will have to use other elementary units that would form much more complex and internally connected system, than the binary logic of ALU.

What can it be, if not 0 and 1?

It seems we may have the answer. The Theory of the Active Perception (beware of clicking the link, cause there’s even more nasty lengthy boffinhead article on the subject that may or may not introduce the new era of computing) explains how the information is perceived by human brain. TAPe is based on group theory and scientific discoveries made by Comexp team. Basically, TAPe is a new way of processing information, which is what we propose to use for building the new generation of ALU — the perceivers.

Now wait a minute. When we say that perceivers can process information much like human brain, there’s a lot to be clarified.

First of all, there are quite a few theories explaining how our brain thinks, and not a single one of them is accepted by the majority of scientists — this means that no one knows exactly how our brain handles information.

Our only claim is that TAPe goes much closer to the human brain when it comes to perceiving and processing information, than any other technology or algorithm.

In other words, TAPe is not a theory of brain structure and function (yet), but a logic explaining thought operation much better than the algorithms based on higher mathematics — those, in fact, have nothing to do with the brain.

And the perceiver could be based on TAPe, using TAPe units instead of 0 and 1 (more information on TAPe is available here, but don’t click yet).

Proof of concept

A bunch of technologies and software products based on TAPe principles are already up and running — addressing the challenges of processing and understanding video data. Not to boast, but addressing quite efficiently.

Comexp is not a unicorn (although it would be nice). But so far, we’ve been growing by the good old rule: “It costs $1 to come up with a brilliant idea, $10 to develop products based on it, and $100 to sell them” (in real life this funnel is even worse). We’ve successfully completed the $1 and the $10 stages, and have entered the $100 game.

We’ve created the world’s first (and only) reverse video search (based on videos, not descriptions or frames).

The tech allows our clients to look for matches in quite big (very big) collection of videos in real time — for example, TV channels used the service for comparing their broadcasts to other channels in order to track pirate broadcasts. Years of video were processed.

And we did it without the $100M that Google spent on developing a similar technology called Content ID for monitoring copyright compliance on YouTube. The best minds in the field of AI, including the legendary Jeffrey Hinton, were called in to work on the algorithm to solve this basic problem. What other illustration of binary disfunction does this industry need?

Using TAPe logic, our video searching and recognition technology can solve the same problems without gradient descent, convolution operation or any other fundamental AI/ML concepts. They are simply excessive — TAPe does just fine without them.

Another example: TAPe allows toinstantly identify the key features for image recognition — moreover, TAPe always suggests an optimal number of such features, and identifies them automatically. If you didn’t understand a single word above, go ask your fellow IT specialist (everyone has a friend in IT nowadays). If you work in the field of neural networks and image recognition, please don’t think we’ve gone nuts. Feel free to contact us for a demo.

So, you’re saying TAPe can improve things for ALU as it is?

The problem is that the modern “steam-powered” computers with their ALU logic don’t allow to use TAPe to its full potential. TAPe is based on fundamentally different units of information, which are meaningful themselves, unlike 0 and 1. Among other things, it means there is more meaningful information per unit of data in TAPe, than in a good old-fashioned bit.

Right now we have to deal with the data represented as ones and zeros, convert the input into the TAPe format, then convert the output… And still, the efficiency of TAPe-based video technologies is superior to the traditional ones. For now we can only dream of the results we could achieve on a computer with a TAPe-based percevier instead of a processor, operating not on 0 and 1, but on TAPe’s elementary units.

What would it look like? Should this computer be quantum? Or, perhaps, neuromorphic? Go ahead and read any article about quantum or neuromorphic processors. You will see that the so-called innovations are based on the same steampunk ALU, with their traditional algebraic logic, the same ones and zeros.

It’s time we stop building rockets with hundred steam engines. It’s time to build fundamentally new devices based on fundamentally new mathematical methods. We believe that TAPe could save the world from ones and zeros, or at least be used in a brand new device (perceiver) for the sake of the future tech.

RIP ALU!