In a rapidly evolving world, technology is at the forefront of innovation, and artificial intelligence (AI) is at the centre of attention of global tech pioneers. AI refers to a computer’s ability to exhibit signs of intelligence. This intelligence manifests itself in part in a machine’s capacity to make decisions as if it were human, using what data it has collected before, to provide the most optimal solution to a command—a process known as machine learning.
Given that the practical applications of AI technologies are not widely understood yet, some media outlets project an exaggeratedly negative representation of the uncertainty that comes with a rapidly-developing technological future. Shows like Black Mirror (2011) and films like Ex-Machina (2014) make it is even easier to imagine a world where humanity’s well-being is threatened by the existence of sentient machines. In reality, however, AI has many practical applications that are a lot less scary.
In today’s digital age, leading tech companies such as Apple, Google, and Microsoft are all investing in AI in an effort to enhance machine learning technologies. Currently, these technologies focus on data collection. In 2014, Google acquired the company DeepMind—a world leader in AI research and its applications. DeepMind’s team of researchers and engineers focus on the development of neural networks—a computational model that partially imitates the structure and functions of biological neural networks. DeepMind uses these artificial neural networks (ANNs) to expand a machine learning method based on learning data representations known as deep learning.
An ANN is built around a collection of nodes or “artificial neurons,” which transmit signals among one another, in the same way that neurons in a human brain do. When an ANN receives an input of information, the network also takes into consideration the many other inputs it has received in the past. Using this catalogued information, the software formulates a solution and forms a pattern so that if a similar situation arises again, the program can work out an answer faster. If the AI program were to play a game of Space Invaders for example, it would play round after round learning through trial and error until it found the optimal strategy to go about blasting all of that space scum.
This kind of learning is not just useful for winning at video games; DeepMind’s main goal is to implement these technological advancements in the real world to help solve tangible issues. One of DeepMind’s primary focus points is health care. An AI machine could help doctors diagnose illnesses faster. A program would be able to formulate a diagnosis, drawing conclusions based off of evidence such as medical images. AI’s appeal stems from its potential for improving tasks and decisions: Making a doctor’s work more accurate and efficient is a win for all parties involved.
Antony Phalen, a strategic partner development manager at DeepMind, gave his insight about the company’s projects on health care at an open talk entitled “Into the Future: Progress Towards AI-enabled Healthcare,” organized as part of SUS Academia Week at McGill on Feb. 9.
“We want to help doctors and hospitals get patients from test to treatment as efficiently as possible,” Phalen said during his talk.
However, this technology is not yet ready for implementation in the real world because of the complexity of the information the program would have to deal with. AI needs access to a collection of data that is being constantly updated, while its interface must be tailored to its user—for instance, doctors acquiring and relaying information in a medical setting. It also requires a strong feedback loop so that the AI is continuously learning how to improve its approach based on what it has experienced thus far. This means that the AI itself has to organize a lot of information in a very specific context as fast as it can, while simultaneously relaying the data to the user so that they can understand what’s going on.
“It’s a 20 to 30 year journey to have a fully digitized environment […but] interoperating streams in which patients and doctors can interact [more efficiently are] imminent,” Phalen said.
As AI already becomes more present in society, many individuals are expressing skepticism about applications of the technology in our day-to-day life. Phalen reassured his audience that with the appropriate regulations in place, AI can be a useful mechanism.
“AI is not good or evil,” Phalen said. “It’s a tool. The way that you make sure it is used for good is [educated] government regulation […that is] heavily focused on [countering] algorithmic bias.”
Algorithmic bias refers to the AI making its own decisions based on the data sets it has accumulated regardless of what is morally or factually correct. We can let AI do all the heavy lifting when it comes to mundane tasks but when it comes to complex situations such as health care, it is imperative that we monitor the decisions it makes.
Element AI, a Montreal-based AI company, aims to shed light on the benefits AI has to offer society. As Canada’s largest privately-owned AI research and development lab, the company helps other businesses to better understand the implications of modern technology and conducts extensive research on future AI models.
Jacob Shriar, a marketing manager for Element AI, believes that an AI-centric world is one to look forward to, not to cower away from.
“The everyday person [...overestimates] how intelligent AI actually is,” Shriar said in an interview with The McGill Tribune. “[They] think it is going to take over jobs [and] go rogue, but that is not the case. [AI] will actually help people with their jobs, making them more efficient.”
Current innovations in AI allow for optimizing processes like banking transactions on smart phones, and self-governing entities like autonomous vehicles. These applications represent the obvious perks of integrating AI into our everyday lives. Nonetheless, the fear of unregulated, or “rogue” AI dominates debates surrounding the progress of AI technologies. Most of the public’s qualms stem from the technology's power to automatically and independently execute tasks, as well as uncertainties surrounding the implications of that power. Authority figures like Elon Musk, CEO of SpaceX and Tesla, encourage such thinking in their public declarations of anxiety at a future where robotics turn the world upside down.
John Giannandrea, senior vice-president of engineering at Google, wants to dispel sweeping doomsaying statements like Musk’s, who has declared on the record that “robots will be able to do everything better than us.” He believes that the fear-mongering rhetoric perpetuated by the media and some tech figureheads falsely influences the public’s perception of the technology, when there is still too much uncertainty about the future of AI to be able to make such predictions.
One thing we do know is that the efficiency of AI is contextual: AI programs run best in the situation they were specifically designed for. One AI machine couldn’t possibly replace the entirety of the workforce in an industry.
“There is general AI and narrow AI,” Shriar said. “General AI is […] where they have a general sense of human intelligence, whereas narrow AI is what we realistically have right now. It can do a lot of things in one narrow path extremely well.”
Shriar suggested that traditional labour-intensive jobs will be the first ones to get the AI makeover. He used the banking world as an example of an industry where menial tasks such as filing paperwork could easily be performed by machines.
Working alongside AI is still a ways away; there are substantial obstacles that stand in the way of AI developing at as fast a pace as it could. One of the stumbling blocks in the growth of AI is the quality of the data currently available.
AI cannot learn without a quality set of data. A computer can’t account for every single parameter in a given situation. The more high-quality data sets the program has access to, the more situations the machine can understand, making it that much more likely to take the correct course of action.
“Data quality is […] not as organized as it could be, not as clean as it could be,” Shriar said. “[Some AI programs] are narrowly focused, [so] if there’s even one problem that the program hasn’t come across, it’s just going to break.”
As AI develops, executives are realizing that it can be used for more than just automation purposes. Decision makers within IT companies are starting to use AI to complement their current technological tools to provide virtual customer assistance on their websites or even make personalized shopping recommendations on online stores. Shriar emphasized that students and future job seekers should invest more time into understanding how AI works in order to gain a competitive edge in an employment market increasingly preoccupied with the growth of AI.
“The whole education system needs to change [because] the whole game is changing,” Shriar said. “Employees will be doing much less data entry, so there’s going to be [a need] for more knowledge workers.”
Knowledge workers manage and use information, identifying the important parts to disseminate to the public and contributing on various levels to the transformation and trade of information. The field of knowledge management can include both programmers and marketers of data.
The education system needs an overhaul that acknowledges the use of AI as it becomes more mainstream. Because, whether we like it or not, AI technologies will become an important part of every sector of society. According to Shriar, educating future members of the workforce about AI’s potential as well as its current limitations can help make the transition into a world more reliant on technology a little smoother.
“We should have discussions on campus and in [different] classes,” Shriar said. “Every program should be putting an emphasis on how AI [might] affect that industry and [construct] their curriculum with that in mind.”
Mathieu Rundström, U2 Science, is majoring in Mathematics and is fascinated by AI and machine learning. He agreed that a better understanding of AI benefits us all.
“There should be more exposure to AI,” Rundström said. “Not everyone knows a lot about technology, and it’s becoming more a part of our everyday culture. Our generation is undergoing a revolution because of how technology with AI is rapidly changing, especially in Montreal.”
Indeed, Montreal is at the forefront of AI learning with four major universities (Concordia, McGill, Université de Montréal, and Université du Québec à Montréal) hosting more than 150 deep-learning researchers. Companies including Google, Microsoft, and Samsung are investing in Montreal’s artificial intelligence sector by setting up research labs in the city. On Sept. 15, 2017, Facebook announced at McGill that it plans to invest $7 million into Montreal’s AI scene. As these big tech companies invest more capital and manpower into developing AI, the technology is only gaining momentum.
AI has the potential to help propel the human race to new heights. Machines that overcome the physical limitations of the human body can facilitate the exploration of dangerous areas such as space and the ocean floor. But AI can also facilitate more mundane, everyday applications of the technology in mobile assistants like Apple’s Siri and Amazon’s Alexa, or even in a typical GPS, which becomes more efficient at selecting route options. AI can assist users with filing tax returns, adjusting the temperature of their home, or simply organizing their day-to-day schedule. There are already a myriad of different technologies in existence that incorporate AI and that we all take for granted.
AI is an augmentation tool that humans have the ability to control. A careful understanding of AI is indispensable when dealing with the future automation of data collection and use. It’s important to pursue future endeavours with clear guidelines and goals in mind. Nonetheless, when envisioning the future of AI we shouldn’t forget that it is here to help—not hinder—our efforts.
“The idea that AI has to take over and massively displace everyone is a myth,” Shriar said. “It has unbelievable benefits from a societal point of view [and will help make] the world a better place.”