Subscribe to our newsletter

Join our subscriber list to get the latest news, updates and special offers delivered directly in your inbox.

Inside the Black Box

Everybody knows that AI means computer aided artificial intelligence.

Some may even be aware of the existence and power of quantum computers.

But apart from that, AI and Quantum Computing are just black boxes that spit out technocratic magic, through means impossible for the uninitiated to understand.

Let’s open that box.

Artificial Intelligence is a computer system that can mimic human intelligence by learning from data, recognizing patterns, and making decisions or predictions based on that learning.

It’s a “system,” and not a “program” because there are a lot of moving parts working together. There has to be a way to get data from various sources. There has to be infrastructure that supports continuous learning and adaptation – something a program with a fixed set of code can’t do. It has to be able to grow, to scale across cloud computing, distributed data processing, and things like that as the demands driven by its learning increase. Lastly, it has to be versatile enough to be able to use what it has learned on a wide range of tasks – again something beyond a coded program’s abilities.

AI excels in several areas. It can digest and learn from large amounts of data very quickly. It can understand and converse in human language through a method called Natural Language Processing (NLP). It can refocused images and speech, automate repetitive tasks, do predictive analysis, personalize recommendations and preferences on platforms like e-commerce or media sites, and give robots the ability to perform complex tasks.

IBM’s AI “Watson” surprised the world in 2011, when it competed on the TV quiz show “Jeopardy!,” and defeated two of the broadcasts greatest champions. It showcased AIs ability to understand both language and unstructured questions in real time, then deliver an accurate response.

AI is weak, however, on the softer side of intelligence. It struggles to understand context, idioms, and sometimes, even common sense. While it can generate art, it is neither creative nor original – all AI products come from existing data processed by the system. AI has neither emotional intelligence nor the ability to make an ethical or moral decision. It has a difficult time adapting to new situations it has not been trained on, and may have some issues doing tasks that require fine motor skills. Lastly, it’s not transparent – it’s difficult for humans to understand how AI arrives at its decisions – hence it being called a “black box.”

Almost all AI currently runs on classical digital computers. These are those hot things with motherboards, sucking electricity and pumping coolant.

While this setup works right now, the amount of car power consumed by classical computers to train AI models doubles every 3 to 4 months. There will come a time when the power we will need for AI will be more than the power we are able to generate.

The hard limit to the future of AI is how much energy we can feed it.

This is where quantum computing comes in.

Warning: crazy big numbers.

A classical computer understands data in terms of bits. These bits represent a single state of data: either a 1, or a 0.

To increase this capacity, more bits need to be added. For every bit that is added, the capacity goes up linearly.

That means that one more bit allows the computer to represent 2 states. Add another, and the computer can represent 3.

Counting ram, storage, and video ram, a typical gaming computer might have 1 trillion bits at its disposal.

A quantum computer, on the other hand, processes data through cubits. Unlike a bit, a cubit can represent 2 states at once: a 1, a 0, or a 1 and a 0. The “how” is complex, so for the sake of this unboxing, let’s take this at face value.

Just like classical computing, adding more cubits makes a quantum computer more powerful. Unlike classical computing, however, each cubit added provides an exponential increase instead of a linear one.

That means that if 1 cubit is 2 states, 2 cubits is 4 states, 3 cubits is 8 states, and so on.

The IBM System One quantum computer in their Shin-Kawasaki complex has 127 cubits. Thanks to the exponential increase in computing power, this unit has 127 undecillion units of computing ability at its disposal.

That’s a “127,” followed by 38 zeroes.

This single machine has the power of 170 septillion gaming computers.

It also only consumes around 15 kilowatts per hour of power. This includes all the things needed to keep it running, like the cryogenic system.

On the classical side, a gaming computer may consume around 0.4 kWh. Not a lot.

170 septillion of them, however, will suck down 50.8 septillion kWh.

It would take 72.8 trillion whole years for the world’s most powerful nuclear plant, the Kashiwazaki-Kariwa Plant, to produce that 50.8 septillion kWh.

72.8 trillion.

The universe is only 13.8 billion years old.

And those gaming computers would drain that supernova in 60 minutes.

Quantum computers are both ridiculously more computationally powerful, and more power efficient than classical computers.

AI is ridiculously more efficient at understanding large data sets and generating meaningful responses than human intelligence.

The marriage of the two will allow the human-machine ecosystem to grow both exponentially, sustainably.

Unicorns and rainbows, yay.

Time to bring out our pots of gold.

How much?

Can I have one, with fries on the side?

The real black box – the one we can’t open – is cost.

Joseph Broz, IBMs VP for Quantum Growth and Market Development, shut down the costing question with a response to the effect of “we do not divulge our finances.”

Glen Thomas, IBM Japan’s VP of Marketing and Communications, was more open, but no less opaque. “There is no one-size-fits-all pricing,” he said. “We tailor each solution to the needs of the business requesting the service.”

In 2017, a D-Wave 2000Q quantum computer was estimated to cost around 15 million USD.

That can buy you 6,000 gaming laptops.

And the D-Wave offerings then are primitive compared to IBM’s now.

It’s cheaper, of course, to rent than buy. IBM details their plans here:

There is a free tier which gives 10 minutes a month, and a pay-as-you-go tier for US$1.60 per second.

A quantum computer happens to be 3 septillion 39 sextillion times more powerful than a regular gaming computer. One second is a lot of time.

These are followed by those “please contact us” tiers, with forms that angels fear to fill.

As far as quantum computing is concerned, the sky’s the limit for both possibilities and pricing.

That’s fine with IBM.

Their core business doesn’t directly touch regular folk.

IBM works with sectors like insurance, government, banking, and the like. These giant organizations need to understand vast amounts of data, and return solutions in a flash. Economies of scale work out in favor of these entities as well: given the huge number of computations they need to perform across their portfolios, IBM says the cost per item addressed becomes very affordable.

The benefits of quantum computing and AI are then passed on to us – better service, financial access, etc.

We might never know to what extent IBM gives us an improved quality of life.

We might never realize how much time AI actually saves us, nor how little energy quantum computers consume in exchange for the kind of computational power they provide.

But those black boxes are open now, and there’s no putting that genie back in.

And if we know anything, we know at least what that genie can do.