Two common threads tying together 2018 tech trends

Maybe 2018 is the year where we think of AI not as artificial intelligence, but as “assistive information” technology

artificial intelligence
geralt (CC0)

Many of the technology trends that drove us into 2017 will continue into 2018: connected devices, digital transformation, the internet of things (IoT), machine learning, artificial intelligence, and automation.

These hot-button issues will remain part of the technology vocabulary in 2018 and beyond. Where I see a substantive difference is in the union of the technologies. AI and IoT are transformative by themselves; now imagine digital transformation in a connected and automated world empowered by an artificial intelligence of things.

Going into 2018, I see two common technology characteristics: intelligence and automation.


If we want smart factories, cities, cars, and homes, the systems that drive them also need to be smart. That requires active, real-time learning systems that can generalize and optimize from a common set of rules.

The impressive advances in AI over the past decade are supported by supervised deep learning: training deep neural networks to perform narrow, single-domain tasks. The learning is supervised because you are telling the algorithm the correct answer (the label) as it is exposed to many examples (big data).

This type of deep learning is powerful and can create systems with superhuman capabilities. Scientists at Stanford University trained a neural network to diagnose skin cancer with accuracy of board-certified dermatologists. It required more than 129,000 medical images.  

People learn differently; we do not need such large volumes of data. On the other hand, machines can process data much more quickly than we can. As a result, it is often faster to train an algorithm than it is to train a human expert. A limiting factor is the need to provide correct labels for large volumes of high-quality training data.

We are now seeing unsupervised learning systems that learn more quickly, require less data, and achieve impressive performance. Beginning in 2018, we will see systems—based on reinforcement learning with little supervision—go beyond game play. (DeepMind’s AlphaGo Zero system was trained entirely by playing itself, starting only from the rules of the game.)

Supply-chain optimization, customer journey, predictive maintenance, datacenter operation, and building automation are examples of domains where there are rule-based systems. You can now quickly train systems that apply these rules better than human-generated logic. 

A tremendous effort has gone into optimizing how supervised algorithms train. Now the focus shifts to optimizing how unsupervised methods can relate the patterns of the world and take the best actions in complex environments.


If there is a way to automate a task or process, chances are you will. You automate not only because you want to focus your attention on higher-value tasks, but to gain operational efficiency and repeatability. Automation has become a necessity. We live in the era of big data, and data volumes will continue to increase, amplified by IoT.

Analytics is key to deriving insight and value from IoT. Automation of analytics has become a necessity to tackle the deluge of data-driven problems.

Data scientists report that 80 percent of their time is spent wrangling, cleaning, and joining data in preparation for analytics. How much would job satisfaction improve if you could automate data management so that these highly trained individuals can focus on deriving value from data?

Machine learning and AI are the technologies that help you achieve automation. And that makes many people uncomfortable:

  • People are not automating simple tasks that require only few skills and little training. The town crier is long gone. In this period, automation aims squarely at functions that require extensive training and preparation on people’s part—for example, the paralegal and medical professional.
  • People have not developed the level of trust in machine learning and AI that they have in other technologies that automate tasks—everyone trusts a copier or fax machine. The inability of machie learning and AI systems to explain how they work and how they make their decisions is a major trust inhibitor.
  • People sometimes confuse automation with autonomy. An automated task does not have to make decisions for you; you can remain in control. People are not ready to change their relationship with mathematics and logic. People like to own the logic, understand and codify it. Software is information processed by computer systems in a known way—people know it because they wrote the code. When things go wrong, they can debug the logic and correct it.

The predictions, classifications, and moves of these AI systems might surprise you not because they have intuition, but because you do not understand their logic. They are not intuitive; they are inscrutable. That is why fair, accountable, and transparent AI is a major research area right now.

Precisely because algorithms learn differently than people, they look at things differently. They can see relationships and patterns that escape us.

As Tom Gruber points out in his TED talk, the conversation we should have is how machines and algorithms can make us smarter, not how smart we can make the machines. Maybe we do not let the algorithm operate the supply chain autonomously but look to it to suggest the next move. It might surprise us.

Like all technology, AI is assistive technology. Maybe 2018 is the year where we think of AI not as artificial intelligence but as “assistive information” technology.

This article is published as part of the IDG Contributor Network. Want to Join?