By Sanjay Bahl
CEO & MD, Centum learning
Workplaces were traditionally made for humans. Typewriters and computers joined them later. But with the advent of Artificially Intelligent Machines, humans must outsmart their replacements. Today’s fear that machines could replace humans in the workplace is far from fiction.
An Inevitable AI Future: Are there genuine reasons to believe there are some real concerns about artificial intelligence (AI), whereby robots could attain enough autonomy to render humans unemployable? The dispute is hot, but the optimists vociferously argue against human replacement. Most AI industry experts stress that robotic process automation can never replace humans, and as the genie in Aladdin, is merely a larger-than-life slave.
Most people I know think of “when” as opposed to “how” it will happen. Yet the rhetoric may be preceding the reality of AI, if not hyping the limits of its ubiquity. When computers were first introduced in Indian banks in the 1980s, they arrived after a massive pushback from employee unions. Suspicion and fear of job losses reigned for many years. In the 1990s, Natural Linguistic Processing (NLP)-based inventions led to the modern virtual assistant. Machine learning makes these virtual assistants learn continuously using artificial intelligence. With the introduction of robots and Artificial Intelligence (AI), another uncertainty-ridden acquiescence in the services and manufacturing workplace seems to be already afoot. But can mathematics solve a non-mathematical problem?
How AI Works: Machines learn through “cognitive technology”. Organizations across industries already use cognitive technology and apply it to product, process and insight. However, in the scheme of the scope of the technology, these are infant days. Because of the qualitative, non-mathematical nature of AI’s predictions, recommendations and decision-making abilities, much needs to be done in deep learning, which uses large-scale neural networks that can contain millions of simulated neurons. But precisely because they are based on training and not programming, deep learning requires enormous amounts of labelled data to perform complex tasks effectively, and that is the crux of its challenge. Large data sets are often not available and when they are, the human challenge in putting them together for the machine is massive.
The Real Risks: While quantifiable and traceable processes are easier to replace, non-mathematical processes may indeed be difficult to replace; businesses can invest more in the human brain for the more qualitative, scalable and not necessarily mappable neural applications like creative solutions, innovations and transformational rather than transactional processes. Siri and Alexa have grown to be popular machine-learning applications, but much remains to make them replace humans. Yet, from Elon Musk to Bill Gates, top technologists have warned about the potentially dangerous outcomes of super-smart AI applications. Musk has even invested in an organization that will work on AI for the benefit of humanity. So are there real risks of AI in a “human workplace”, and if so, what are they?
- Intelligence Explosion: Scientific literature suggests the possibility of a sudden explosion of intelligence—an existential risk involving recursive self-improvement as a result of artificial super-intelligence. This would mean that while humans are both endowed and limited by a time-bound and experiential expansion of the brain’s horizons, a machine’s clock may be programmed to evolve at a much faster pace.
- Cockroach Intelligence: In 2009, scientists agreed that some computer viruses had acquired the capability to resist destruction or elimination. Whether that uncontrollable level of autonomy would actually be accepted at legitimate application levels is very doubtful; however; there is some agreement to the apprehension that machines may “over-learn” precisely to achieve the task they have been assigned in a superior fashion.
- It is Hard to Debug Everything: A BBC report in 2016 revealed a startling case where a system trained to classify pneumonia as closer to death than patients with a history of asthma. “People with pneumonia and a history of asthma go straight to intensive care and therefore get the kind of treatment that significantly reduces their risk of dying. The machine learning took this to mean that asthma + pneumonia = lower risk of death.” The real risk, therefore, is our inability to see how a computer-derived the solution that it did.
Learning to Learn: As a recent industry report states, data size, explainability [of, say, decisions], the generalizability of a solution or decision in a specific case to all similar cases, and other human bugs such as bias will remain sticking points in automation based on deep learning. This means that the utility of deep learning will
not be universal. What is the one thing that machines have been programmed to do, which professionals can do well to replicate in order to avoid being irrelevant? It is learning to learn—constantly evolving from the past to create the future. Our education systems currently do not offer learning ability as a pedagogic process, and that creates a real hurdle for those who must learn to learn.
To help us with the constancy of learning and to avoid being obsolete, repeated up-skilling is an available solution to professionals. There is an increasing number of training courses in NLP, deep learning and machine learning. Up skilling will therefore no longer be optional. At the same time, businesses will look to up scaling their operations, and this gives the human workforce the opportunity to stay relevant. The challenge before business leaders of tomorrow will be how to take advantage of AI while keeping intact the human workforce, not disrupting the economic balance in the very society for whom they are creating wealth.
AI is a reality that business leaders must face. While confronting the reality of AI, good business leaders distinguish between replacement that brings genuine advantages and replacement for technology’s sake. Transformative technology replacement leads to precision, scaling up, and growth.
And growth creates jobs. As we scramble to “learn to learn”—train ourselves to constantly up-skill and future-proof ourselves—we must also realize that the real threat from machines to our work cannot occur in the current paradigm, unless we deem it so.