Artificial Intelligence and Data Science: levers to transform the way we do things

No one can deny the media hype surrounding artificial intelligence (AI) and data science. With a very active academic leadership and an ecosystem dedicated to the subject, Montreal is already recognized as a dominant pole in North America.

But, between the alarmist visions of imminent human obsolescence and the denunciations of fake news, what is the effective contribution of these specialties for companies?

Between myth and reality

Despite its promise, AI is still a long way from a replica that equals the human being. It is essentially an algorithmic application of methods known for half a century in statistics, linear algebra and differential calculus.

Deep learning, the most comparable form of the human brain, is only applicable to the achievement of very specific objectives. We are still unable to replicate the power and versatility of the biological brain fed in real time by the multiple senses of the human body.

Machine learning in the age of Big Data

AI has historically been limited by the capacity to process and access data. Each company jealously guarded its data, but the advent of the Internet changed everything. In a few years, data has become massively open and public. With the emergence of powerful vector processors and widespread wireless networks, innovation in mobility, cloud computing and interconnected objects has increased the possibilities tenfold.

This simplified sharing has enabled the massive creation of public data that facilitates supervised training of machine learning algorithms. Combined with the evolution of data science, a multitude of bots and innovative solutions are now shared at the click of a button on our mobiles, tablets, watches, vehicles and other electronic devices.

Data: the key to AI’s potential

Since AI is an application of algorithms that focus on finding affinities in data, its potential cannot be realized without it. Even the most “intelligent” algorithms are useless without relevant, reliable and sufficient data.

Machine learning is powerful for the interpretation of complex or large data, when the classification criteria are stable. It is particularly cost-effective when used where humans lose their advantage:

• Predict and fill in missing data in large existing databases based on similarities detected with other public sources;
• Process electronically received requests to determine eligibility, identify the subject matter or service, provide a pre-determined response or redirect the non-standard request to an expert;
• Analyze a multitude of data in real time to coordinate rapid actions over long, continuous periods (e.g., production lines, warehousing, supervision);
• Identify patterns or predict trends from large volumes of structured or unstructured data (e.g., stock market buying/selling, fraud detection, consumer opinion analysis, traffic coordination);
• Assist human expertise for better decision making in complex, risky areas or when expertise is scarce (e.g.: anti-terrorism, cybersecurity, medical diagnosis, piloting, climate or weather forecasting);
• Simplify interaction by voice translation or interpretation of posture, gestures or brain waves;
• Improve and refine repetitive programmed actions (RPA, WPA, bot) or virtual interaction scripts (chatbot) by continuous reinforcement.

Towards employment 4.0

In the age of the knowledge economy, few organizations take advantage of the intellectual capacity of their employees, many of whom are condemned to mind-numbing daily tasks. With the proven ability of intelligent algorithms and automata, any employer would do well to equip their employees with such tools to increase their competence and efficiency.

 

This article was originally written for E3 Consulting Services’ Spring/Summer 2018 Newsletter.