We sat down with Tanzim Saqib, Naborly’s Principal Engineer, to unpack what artificial intelligence means today and how it will be used in the future. Banking on his long experience at Microsoft and British Telecom in the past, he has been building and shipping modern AI systems since 2016.

The first in a 2-part series, we will start by dissecting AI and then look into how it is penetrating the residential real estate industry.

What is AI?

Cloud, Internet of Things (IoT), Big Data, Artificial Intelligence (AI), Blockchain - these are a tall order of buzzwords have recently received much media attention. While these terms are sometimes misrepresented in scope and application, the media has done a remarkable job of educating people at scale and creating demand all over the world.

AI is an umbrella term for analyzing information, representing the extracted data digitally in such a way that a computer program can process, inferring new facts from it, and solving a predefined set of problems and goals using data. AI banks on many enablers and related technologies, so the heart of AI is inference, reasoning and problem solving.

We often judge AI by its applications. For example, when we see a robotic limb, it is not entirely AI: the electronics, sensing, actuating, control systems, digital signal processing involved are not part of AI. AI drives behavior of those electronic devices, exactly the way a computer program drives your computer. Typically, when we see a computer program being very smart about its surrounding and our needs, we call them AI. However, there are enabler technologies, such as interfaces that can receive and process data points from millions of internet connected devices (IoT) in real time, databases that can contain such a very large dataset, highly parallelized computer programs that can work with such a database, also known as Big data, rapidly commissionable and decommission-able elastic computer platforms, also known as cloud, that can run such programs, data visualizations technologies that enable rapid experimentations. Although the first work on AI was revealed in 1951 (Copeland, B. Jack., 2000) and AI’s application without having a title “AI” have been in software for several decades now, AI was never as bullish.

How come AI is all the rage now?

Two simple reasons: we and the systems we use now generate data about ourselves more than ever and enabler technologies have evolved in cloud/big data and hardware space so much so that now we can use those data, process and finish computation within a reasonable amount of time and augment some of the intelligent human behaviors as promised by AI (or promoters of AI). The amount of data points that we generate are pretty easy to see. Fitbits on your wrists produce crucial health and habitual data and upload to their servers. With the advent of ubiquitous GPS-enabled mobile technology location-based data is available to the phone companies at a scale it was never possible. Netflix films you browse, YouTube videos you watch, Facebook newsfeed item you spend time on just staring, Amazon products that you view and order, all of that contribute to building a persona for you. With the help of above-mentioned enabler technologies, now it is easier to ask bigger and more complex questions that require the ability to process huge amount of data, newer tools to finish processing those data within a reasonable time, cleverer AI algorithms that can offer incredibly high accuracy. These questions are often seemingly unrelated but may result in profits! What is the likelihood of people who watch action films, read about business, go to gym once a week, dine at the restaurant four times a week, click on money related news feed items, may have kids and buy diapers once a month? A good AI system typically processes sufficient data to be able to accurately predict, classify and categorize and unveil new data points.

What excites you about AI in North America?

I am fond of the concept “creative destruction” which essentially captures the essence of innovation. Instead of waiting for the corporations to solve existing problems in the market and society for us, start-ups take the initiative to start addressing them by freeing themselves from corporate obligations and board pressure for stable growth in share price. Disruptive innovation disrupts the business cycle of corporate capitalism. Tech start-ups and like-minded corporations have attempted and succeeded in revolutionizing our lives.

The “Applied AI Era”, as I like to call it, is a practical example of creative destruction. Like any disruptive technology (Christensen, 2006), AI challenges status quos, offers efficiency by automation and cost savings, and uncovers new opportunity for business. America is the top producer of tech start-ups in the world. There are 85 start-ups worth more than $1B in evaluation in United States according to a market insights agency (The Complete List of Unicorn Companies, 2019). Start-ups in America also enjoy access to a big and mature ecosystem including the largest venture capital funds in the world. North America is going to see AI contribute $3.7T to GDP by 2030 according to PwC (PwC, 2019). It is quite safe to assert that in both start-ups and enterprises AI is going to be one of major forces of future economic impact.

AI Research Citations (1997-2017)

Source: (Scimagojr.com, 2019)

Both United States and Canada have been preparing their workforces to take advantage of the Applied AI era with their top-notch education systems and production of impactful research. Canada is also home to top scientific thought leaders of modern AI. NIPS (now known as NeurIPS) is probably the top most AI conference in the world and almost all of the top contributors are North American industries and academia.

NIPS Conference Report

Source: (Microsoft Research, 2019)

It is all happening in North America. In my opinion this race is not an existence-threatening competition, because it is not a winners-take-all scenario and essentially, any new advancement can be achieved by other researchers as well. Therefore, it is critical to success that the ecosystem allows the innovators freely do their best work. For that, North America is one of the best places to be.

Is AI being used to its full potential?

We expect that AI should have done a wide variety of tasks already. However, there are generally two major roadblocks: hardware and software.

Hardware

Not many decades ago, many used to wonder, what would they do with a more powerful computer? The software that they used daily ran perfectly fine. There is a widely popular saying, "640K ought to be enough for anybody" is always wrongly quoted to Bill Gates. 640K memory, also known as RAM, was perfectly fine for those who used to use those ancient applications. Then, many creative applications of computing came into existence, and that memory was simply no longer sufficient. The smartphone that you use now probably has more than 2GB RAM. You can play high definition videos and games, do business in it, communicate, etc. We have come a long way from 640K RAM era and a personal computer system with 32GB RAM is not uncommon now. I’d argue that it was software that demanded further hardware innovation.

Therefore, we always hope to see a dramatic shift in hardware quite frequently. However, the progress of semi-conductor technology which powers today’s computers has slowed since 2015. Semi-conductors are used to create processors, also known as Central Processing Units (CPU). Processors are the modules in your smartphone and computer that do all the math, obey all your commands, runs the apps and simply renders the computing experience to you. Because it has become a huge challenge to accommodate even more semi-conductors into the same space as the earlier generation, hardware leaps every year do not seem to be possible anymore. For example, my first computer’s processor speed was 200MHz in 1998, why could we not go beyond 3.6GHz processor base clock speed yet, first introduced in 2013?

One way to address such a problem is looking out for alternate compute resources, similar to how we look out for alternative sources of energy. Quantum computers is one such option, but in my projection, it will take at least 10 years for this technology to be realistically usable and it will never replace your personal computer. Rather it will work similar to on-board modules, such as your graphics cards, math co-processors, etc.

Cloud technology has allowed us to scale our processing power and computation in multiple servers horizontally, but when you require heavy-weight computation like AI, your computer resources need to be vertically scalable, meaning one machine needs to be as powerful as a super computer, which is still very expensive. We cannot always distribute a task across multiple computers, because transferring data over the wires from one computer to another takes time. When you work with large datasets, these delays add up massively.

Graphics Processing Unit (GPU) has been always appealing to the gamers and audio-visual production enthusiasts, professionals and companies, but now AI firms are equally interested in it. GPUs have the ability to vertically scale computation in a single machine. Performance of the GPUs for AI is not too mind-boggling, but it is far better than traditional CPUs.

Now, is it possible to build a computer system entirely on GPU instead of CPU? No, but we can use one or more GPUs per system to accomplish in AI more than what we previously were able to. There was a very long dark time for AI especially due to hardware limitations and scarcity of data, popularly known as ‘AI winter’. The number of AI solutions that were and are possible, are being built now using existing hardware innovations. However, like the analogy that I made earlier, with the new, innovative and creative AI applications, hardware also needs to be evolved at a similar pace to meet software demand.

Software

The other major blocker is in the generality of AI. If you teach a computer to be the best chess/go/computer game player, it will probably be undefeatable by humans. Even in games, I have had a fascinating experience with AI. You may not be able to defeat a properly designed AI in games because its compute space and readily accessible memory are so large, but sometimes you can predict its next move, because you know for a fact it will not make a mistake!

But, if you task a games AI to classify cats and dogs, it will fail in all attempts. If you teach an AI to detect only all the breeds of cats, it will probably classify a cheetah, leopard and Bengal tiger a cat as well although they are different species. An AI can never be a full replacement of a person in our lifetime especially because of the amount of knowledge each individual human being possesses.

What if we created an amalgamated AI with all those individual and separate specialized modules? Would that be able to replace humans? Probably not. The orchestration of all our faculties is probably even more fascinating, complex and brilliant. The concept of building a generalized artificial brain orchestrating its specialized parts can be called Artificial General Intelligence (AGI).

Could AI replace all banal jobs of humans? Absolutely! That is what we should expect AI to do until we achieve significant breakthrough in AGI.

References

  • Copeland, B. Jack. (2000). The turing test. Minds and Machines 10.4: 519-539
  • Christensen, C. (2006). The Innovator's dilemma. New York: Collins business essentials.
  • The Complete List of Unicorn Companies. (2019). The Complete List of Unicorn Companies. [online] Available at: https://goo.gl/RYpNVG [Accessed 2 Feb. 2019].
  • Scimagojr.com. (2019). SJR - International Science Ranking. [online] Available at: https://goo.gl/oH3QMY [Accessed 2 Feb. 2019].
  • Microsoft Research. (2019). NeurIPS Conference Analytics - Microsoft Research. [online] Available at: https://goo.gl/uCAeC9 [Accessed 2 Feb. 2019].
  • PwC. (2019). PwC’s Global Artificial Intelligence Study: Sizing the prize. [online] Available at: https://goo.gl/RkE5nB [Accessed 2 Feb. 2019].

Learn more about how you can get a Naborly Report on your prospective tenants for free with Naborly.