The Top Four Most Interesting Questions in Computer Science Engineering and Technology



Introduction

Computer Science is a fascinating topic, full of boundless potential. One of the extra advantages of being a researcher in CS is that there is no shortage of possibly solvable open questions that are available. Compare that to the field of Mathematics which has become so convoluted and arcane that just understanding the proof of a currently solved problem is a project that might take months. For one single proof. Computer Science, on the other hand, is developing its scope and the range of the field on a minute-by-minute basis. It is not unlikely that a solution to an existing problem could be found or someone else’s work rediscovered by an independent party simply because the field is expanding incredibly fast, and a huge range of unsolved problems exist that can actually be tackled even by a gifted student. And we see examples of students who do exactly that every single year, consistently.

So, while Mathematics and Physics reign as the most fundamental research subjects, it is definitely worth your time to get into Computer Science research instead, since there is a much greater likelihood that you will discover something new. To have a bird’s eye view of the current scenario, here are four of the most interesting and high-potential problems (that could be solved by a novice) that have both massive industrial applicability and ramifications for the world at large. The four questions that we have chosen are:

1. What are the Capacities of a Quantum Computer?

2. What is the Most Optimal Human-Computer Interface Design?

3. Is Hypercomputation Possible? (It is!)

4. Can We Create Artificial Life with AI and Deep Learning?

This list is rather general and perhaps not to a computer scientist’s taste. True! Totally true! However, my desire is not to list a set of mathematical problems formally. There are plenty of those lists on the Internet. My aim is to make the interested reader understand the appeal and the vast potential of Computer Science for humanity as a whole, including all the most appealing possibilities. And in doing so, I hope to garner interest for Computer Science research in anyone who reads this article. Additionally, if I can explain some technical jargon in plain English for everybody and not just for computer scientists — all the better!

1. What are the Capacities of a Quantum Computer?

Quantum computing was once the subject of science fiction. Now, quantum computation is very much a reality, but only for research purposes — at least, so far. Theoretically, the potential of a quantum computer with quantum bits (qubits for short) is massive — with a mere 100 qubits, one can manipulate over 2 to the power of 100 elements — more than the total number of subatomic particles in the entire universe! However, the main issue with quantum computation is the phenomenon of wave function collapse and also, decoherence. Wave function collapse occurs because, by definition, when the output of a quantum computer is observed, it collapses into one single stable value. So, the massive parallelism is obscured, since the calculation becomes serial the minute we observe it. So, a quantum computer would be useless, right? Actually, the answer is — no. A quantum algorithm by Peter Shor has been discovered that uses entanglement to solve an NP problem (see previous topic) of factoring large numbers. In theory, a quantum computer of 100 qubits could render modern encryption schemes useless by solving the factoring problem in polynomial time. However, that landmark is still some years away — experts believe a minimum of five more years is required to overcome the current problems in implementation.

Furthermore, no one has hit upon a standard way to implement a qubit. There are multiple research implementations — ion traps, spin-2 particle systems, lattice systems, nuclear magnetic resonance (NMR) methods, but there has been no model that is practical for even 100 qubits. The problem is that a fundamental particle is in a single state only for around 1 X 10–8 seconds, meaning that the states have to be persisted using an additional mechanism to keep the particle in the same state for any length of time that a single computation requires. What is more, the rate at which errors and noise is generated while measuring quantum phenomena is so massive than quantum error correction schemes have been implemented to work around this problem. But whether a model will be discovered that will be both practical and not too expensive to maintain and whose coherence times will be measured in at least seconds rather than nanoseconds is a question that is an open mystery. If this problem is solved, then watch out, for problems that are completely unsolvable for conventional computers will be wide open and easily solvable with these quantum computer systems.

2. What is the Most Optimal Human-Computer Interface Design?

Human-Computer Interface in its standard form involves a keyboard and a mouse that you use your hands for. However, recent advances in technology have opened the door to much more invasive but equally effective devices. Researchers have been successful in implanting a chip in the brain of a paralytic that enabled him to move his hands and feet. More recently, a device was implanted in a completely paralyzed individual so that he could control an external keyboard — believe it or not — by his thoughts. Communicating with a computer via a chip in your head sounds extreme, but controlling a computer with our minds may soon be possible for the general public in the near future. Of course, that raises a whole lot of ethical questions. Would such a implant allow a hacker to control you remotely? It’s a valid and an infinitely troubling question! What is worse, if an external device is controlled by our thoughts, what happens when people become mentally ill and thoughts spiral out of their control? Perhaps the scariest application of this technology is a human completely controlled by a criminal — without any external signs of it! The person could commit any crime, and only a log of the commands in his brain can show that he was not the responsible party. For all we know, this may be reality in some counter-intelligence agencies. The ethical questions that surround this issue are considerable. Is such a device even worth implanting? Is not the risk too much to take?

Even with all our reservations, it remains clear that there are several applications to this technology that can be allowed. As mentioned, paralyzed or crippled patients can control their prosthetics with their thoughts. A person incapable of movement can type onto a computer by thinking the appropriate commands, allowing him to communicate. In such extreme cases, the technology seems valid. However, the risk of hackers and the very real risk that the system controlling the security of this implant could be compromised — every system, regardless how secure it may be, has a potential to be hacked — regardless of all precautions — makes this technology a weapon, and a terrifying and scary one at that. The risk might not be worth it. For the conspiracy theorists among us, there is always the possibility that this has already been done to someone without their knowledge. And that anyone — any human being — without their knowledge — could be operating via remote control instead of autonomously. Scary!

3. Is Hypercomputation Possible? It is (!)

What is hypercomputation? General computing falls under the category of Turing Machine Computable functions. To avoid explaining what a Turing Machine is and its formal description, let’s abstractly consider the essence of the concept. A Turing Machine-Computable function refers to any arbitrary computation that can be described by a clerk who can write the output values one at a time on a roll of text. We’ll assume the value is provided by an atomic operation in the underlying language. Alan Turing, the father of the Theory of Computation, postulated that a Universal Turing Machine could be built using an oracle satisfying certain conditions that could perform any arbitrary function. It is possible to achieve this situation in specialized environments.

Specifically put, hypercomputing refers to performing computations that cannot be performed on a Turing Machine. This model can associate infinite time or infinite space resources to the computation model. Even using theoretically finite time and space, there are several approaches which perform hypercomputing. So, hypercomputing is possible — under the right conditions. We shall cover three particularly esoteric approaches to computation.

The first is an Accelerated Turing Model that performs computing using superluminal particles. By this we mean, particles travelling faster that light. Before you laugh it off, do understand that even special relativity has very rare conditions in which particles can be measured to travel faster than light. But this phenomenon is achieved in the lab using quantum tunnelling, in which photons cross a finite distance between two prisms instantaneously, achieving almost perfect speed in the process. This phenomenon reflects the EPR paradox (search Google or Wikipedia) and was one of the first reasons that quantum mechanics was considered incomplete (even if remarkably accurate).

The second is a Relativistic Double-Black Hole Space Computation, which uses the unique conditions that Kerr black holes cause (two black holes slowly orbiting each other) that cause two event horizons to happen, and the area in between the event horizons will support computations that cannot be done on a standard Turing Machine. This particular configuration also gives an infinite speed up, since if the process is frozen, events in one black hole seem to happen arbitrarily fast when observed from the event horizon of the second slowly orbiting black hole. Exotic!

The last one is not theoretical, it is a working machine, known as the Infinity Machine, or the Quantum Adiabatic Computer from D-Wave called D-Wave One. This unique machine uses quantum entanglement and superposition to explore all possible configurations on a certain optimization process called quantum annealing. This system finds its lowest possible energy state automatically, thus providing computation almost instantly using through evolution of its time dependent Hamiltonian of the system (a fancy mathematical name for a mathematical object made up of partial differentials that describe a quantum mechanical system).

D-Wave went on to build a successor, the D-Wave Two, that the company claims can simulate optimization on a 50-qubit quantum computer. These machines claim to be the start of a new quantum computing revolution — the result of which only time will tell.

4. Can We Create Artificial Life with AI and Deep Learning?

Once upon a time, ‘artificial life’ was a computer game that consisted of reproducing cellular automatons that the mathematician John Conway invented in his free time. Artificial life and artificial intelligence have come a long way since then, mostly thanks to the advent of deep learning. Deep learning refers to the construction of neural networks with multiple hidden layers that may also include feedback (convolutional) networks (with often as many as 9 or 10 hidden layers). These additional layers allow the network to search along multiple combinations and configurations of the input variables. This avoids the problem of local minima by using hundreds of variables and a very complex state space which has so many dimensions that local minima are statistically incredibly rare. Recently, companies like Google, Microsoft, and IBM have made remarkable achievements in this field.

Google bought DeepMind, a company focused on ‘creating a sentient intelligent being’. In many applications, you can almost imagine that they’ve succeeded. Until now AI has been under the control of human beings. Respected figures such as Stephen Hawking and Elon Musk have stated that a self-evolving autonomous AI could cause the extinction of the human race (Terminator 3’s Skynet, anyone?) and deep neural networks have achieved remarkably human behaviour in various simulated environments. These mathematical models can recognize images, dream, play board games, play video games, and almost without exception, outperform their human counterparts on every single scale.

IBM built AlphaGo, a program that defeated the world’s best Go player (Go is not just a programming language, it’s also the name of an ancient Japanese board game) in a tournament. Quite remarkably, the network trained itself, by playing several billion games against itself, and learning and improving itself from the results of each game (a technique called reinforcement learning). Microsoft has also invested heavily in AI and provides AI-as-a-Service through Microsoft Azure on the cloud. IBM also has a notable cognitive computer, called Watson, that performs nearly every human function that is possible via computer, and does it better than humans.

So, seeing all this, can we argue that artificial life will actually develop itself? The AI Singularity that Musk and Hawking warn against with such dire consequences? This researcher actually believes that it is likely, even inevitable if we continue developing AI as actively as we do now. Simply put, machines are evolving. While I agree that we cannot create a ‘soul’, as it is considered in Western philosophy, a sufficiently complex machine can definitely keep up the appearance of autonomous intelligence. This reflects the reasoning behind Google’s effort to develop DeepMind.

The approach Google has taken is to simulate human vision, in as human a way as possible. They have been successful in developing a human-like intelligence, and this is the core of my argument. A sufficiently capable entity capable of autonomous behaviour (self-driving cars are just an inevitable start) could develop very human-like behaviour, especially in interaction with real-world issues. Yes — issues. Deep neural networks are capable of actual reasoning, depending upon the inputs. Some properly (or improperly) trained deep neural networks are capable of demonstrating even morality, both good and bad, depending upon the training set of inputs. The danger to the human race is real, since life on Earth has been a self-repeating history of evolution with more highly evolved animals replacing those with inferior characteristics. Only this time, man himself is the one who is inferior, in comparison with an intelligent machine (the next evolutionary cycle in the course of time) that has enough autonomous behavior to live forever when provided with infinite supplies of fuel.

Deep learning has opened vistas into AI like never before. No other branch of AI has done so much to advance the field in such a fundamental way, in so many foundational areas. The future is here — and it is intelligent. A mad scientist focused on creating a superhuman intelligence has all the ingredients he needs. We just hope and pray is no one is mad enough to risk creating such an entity. Is it capable of evolving into a superhuman intelligence without our direction? That is, without thinking, would researchers out there end up creating a machine with intelligence superior to human beings? Elon Musk and Stephen Hawking believes that the answer is yes. And when personalities of their caliber agree on something, there is good reason to believe that that ‘something’ is true. So far, it has not yet happened. But the danger is real. At least, according to Elon Musk.

Final Words

We have covered a lot of material. There is a seemingly infinite range of material on the Internet that covers all of this and much more. I hope that this effort has put the spark of hope into your heart that there is much fundamental work yet to be done in Applied Computer Science, simply because the field is evolving at such a rapid rate. If you are looking for research — you’re in the right place.

Computer science is currently the most fulfilling area of applied research currently available. Both technology and the theory behind the tech progresses at such a rapid rate that it is difficult even for professionals to keep track of everything new to the field. And the field is still in its infancy — and especially so in the case of quantum computation and artificial intelligence.

Source: Deep Learning on Medium