Advances in artificial intelligence will not impact all industries equally.
Depending on your perspective, artificial intelligence is either a boon for business or an unstoppable force that consumes jobs. Despite the gravity of a White House report predicting that vehicle automation threatens to replace as many as 1.7 million jobs in the trucking industry, this sea change will not impact all industries equally. The type of labor done in that industry—driving over long stretches of uniform highway, with fairly predictable traffic patterns and well-defined lanes in which to navigate—is exactly the kind of job the current generation of artificial intelligence is well-suited to perform.
While these kinds of jobs with predictable labor are highly subject to automation, jobs that require management or expertise are not. It’s the jobs with high expertise or human interaction where automation can augment, not replace, the potential of skilled workers.
A Current Snapshot
Given the rapid advancement in the field, it’s important to establish some definitions. Artificial intelligence (AI) is a broad term that can be understood as using computations to mimic human thought for decision making, problem solving and learning. Machine learning is a particular subset of this, wherein a computer is made to behave similar to the process of learning as we observe it in humans and other animals.
One way to achieve machine learning is through deep learning, where computational models composed of multiple layers of processing learn representations of data. Recent advancements in deep learning stem from the application of artificial neural networks, which model the brain by connecting a very large number of simple processing units, much like neurons. While an individual node in a neural network may only perform a simple algebraic operation, the aggregate result of these many operations can mimic what we consider learning.
While the aforementioned report on vehicle automation might seem an impossible future scenario, similar automation is impacting the workplace today. Starship Technologies has begun to deliver meals via wheeled robots in Washington, DC. Amazon made its first Prime Air delivery via drone in late 2016. Factory work is already highly automated, and new orders of industrial robots are at an all-time high. McDonald’s is replacing cashiers with touch-screen ordering kiosks. A variety of platforms exist to take raw data related to everything from the stock market to tennis and generate news stories.
It’s easy to forget that automation has already provided self-checkout lanes, traffic lights that ticket violations, high-frequency stock trading, and numerous other applications. It may sound as if artificial intelligence, robots and automation will only serve to replace workers in the workplace of the future. Thankfully, that is not the case.
Avenues for Growth and Insight
The jobs that will be replaced are those that are repetitive or physically demanding—areas where robots excel and human satisfaction is often low. In today’s information economy, there are a number of jobs that require extensive knowledge and expertise and are not easily supplanted by artificial intelligence, but could be augmented by it.
Japanese farmers have automated the time-consuming process of sorting cucumbers by quality, but the art of cucumber farming still requires vast expertise. Code-sharing site Github uses a bot to automate tasks such as managing conference calls, distributing builds of software or handling server load, freeing up software developers to fix critical bugs and work with customers. IBM has partnered with hospitals to deploy its Watson artificial intelligence platform, where it will help analyze information in a field with a rapidly expanding corpus of medical knowledge. Automation services like IFTTT and Zapier can free up time consumed by repetitive daily tasks. This is just the beginning, as we are at the cusp of a fusion of artificial intelligence and workers.
The future brings a number of growth avenues in this area. One is computer vision, where machine learning could assist security in the monitoring of video feeds, identify trends in customer movement through a retail store to optimize layout, or evaluate customer sentiment during video conferences, where subtle cues are hard to notice. In the field of textual analysis, chatbots that respond to conversational input could expand to provide information a worker will shortly need based on current conversations, or provide action items based on an email chain. (Google already does this to recommend auto-replies.)
Of course, the cloud and big data offer great potential for artificial intelligence, where it can provide new insights based on collected data, improve sales and marketing outcomes by analyzing past experience and making recommendations, or analyze software to fix defects. The potential for teamwork between AI and humans has yet to be fully realized; expect rapid growth in this area in the near-future. Despite the possibilities for collaboration, there are some concerns with artificial intelligence.
An Existential Threat?
A number of prominent individuals have expressed grave concerns about the future of artificial intelligence. Elon Musk has stated that artificial intelligence is “our biggest existential threat” and has compared this field of research to summoning a demon. This concern is related to Ray Kurzweil’s concept of a singularity—when a computer is capable of building an even more intelligent version of itself, ad infinitum—and the resulting irrelevance of humans in a world in which we are unable to comprehend the dominant intelligence. Far from being a far-flung concept, Kurzweil predicts this will happen in our lifetimes.
Some worst-case predictions include the scenario in which a computer is given the task of maximizing the output of paperclips, and thus turns the entire planet into paperclips. Other noted tech luminaries, such as Mark Zuckerberg, have downplayed the threat posed by artificial intelligence, describing Musk’s alarmism as irresponsible. Today’s neural networks are a far cry from the complexity of the human brain and run on much less powerful hardware; however, as the technology improves, these risks will only increase.
More pressing is the simple issue of understanding neural networks, which have proven to be extremely capable of modeling and understanding information simply by spreading calculations over many layers. However, they are often so complex that it’s hard to know with any certainty how or why they work, so they are commonly thought of as “black box” solutions.
Traditional attempts to get a computer to recognize a face required identifying predictable key features of the human face and then searching for those features in an image. In contrast, a deep learning approach involves providing a very large dataset of example images and allowing the network to find an efficient solution that recognizes when faces are present. We are able to identify the input data and the effectiveness of the resulting network, but it’s very difficult to say with certainty how any given node in a graph of thousands contributes to the final network. In order for neural networks to be safe and verifiable, we must improve on this front.
The Future of Automation
Automation and artificial intelligence in the workplace is a vast topic of discussion. Deep learning and neural networks are advancing at an astonishing pace, and new research in this field is available almost daily. While Musk and Kurzweil offer pessimistic views on the future of high-powered artificial intelligence, the state of the art works with machines of limited power to solve limited problems.
We’ve seen that robots are already completing fairly complex tasks, but the future of automation will not be limited to companies like Google and Github; artificial intelligence has the potential to benefit workers in security, retail, communications and more. The ideas offered here only scratch the surface. Explore the potential of automation in your own workplace by using an automation service, or discover the potential of software you already use to handle repetitive tasks and save time every day. iBi