Google DeepMind/UnsplashCC BY-SA
According to the Australian Economic Development Board, our country’s productivity growth is at its lowest in 60 years. And this economic downturn is reflected in most developed countries around the world.
So it’s no surprise that some see the rise of artificial intelligence (AI) as a productivity savior. Media articles herald a new era of high productivity enabled by AI, especially generative AI tools such as his ChatGPT and DALL-E.
Similarly, the world’s top journals are filled with descriptions of how AI has enabled transformative leaps in research. Machine learning is being used to predict the shape of proteins from DNA information and to control the shape of superheated plasma in nuclear fusion reactions. One team at CSIRO designed an AI-based autonomous system that allowed him to manufacture and test 12,000 solar cell designs within 24 hours.
Does that mean I can switch it on, leave it on auto, and go to the beach? Not perfect.
Not a productivity panacea
The example above is both hopeful and a distraction from the many poorly functioning AI applications. These are cases where the use of AI was costly, time-consuming, and did not produce the desired results, and is often not covered in journals and media.
In 2021, 62 published studies using machine learning to diagnose COVID-19 from chest scans will be unreliable and unusable in clinical practice, mainly due to input data issues. As it turned out, the AI community was forced to pause. It made me realize that AI can make mistakes.
This is not to say that AI cannot be used to improve productivity. It’s just that it’s not an off-the-shelf panacea for solving productivity problems. AI cannot magically solve problems related to inefficient processes, poor governance, and bad culture.
Bringing advanced AI to stupid organizations doesn’t make them smarter. It just helps organizations do stupid things more efficiently (in other words, faster). This does little to improve productivity.
Where AI applications work
A recent study by the U.S. National Bureau of Economic Research found a 14% increase in productivity for customer service agents who used AI tools to guide conversations. In Australia, Westpac says AI has made software engineers 46% more productive, with no decline in job quality.
In many ways these examples are not surprising. It is clear that effective use of AI can improve productivity. Google Maps is clearly better at getting from A to B than the old road atlas.
So what do the situations where AI performs well have in common?
Successful applications of AI tend to be characterized by distinct needs and capabilities of the AI system. They are well integrated within a broader business or organizational process and do not interfere with the employee’s other tasks.
They also tend to use high-quality, fit-for-purpose curated datasets used to train algorithms, and are safely applied according to ethical principles.
Where AI Applications Fail
However, realizing the productivity benefits of AI across organizations, let alone the economy as a whole, is difficult. Many organizations are still working on a more fundamental digital transformation.
Consulting firm Deloitte estimates that 70% of organizations’ digital transformation efforts fail. Perhaps the real solution to the productivity dilemma lies not in using AI, but in managing organizational inefficiencies associated with the introduction of new technologies.
The modern office is cluttered with pointless emails, unnecessary meetings, and bureaucratic processes that sap employee energy and motivation. Studies have proven that when workers face such hectic tasks and distractions, they become less productive.
AI is unlikely to solve this. Modern currency is noteworthy. AI built to keep us from needlessly busy may end up nagging us. A future may emerge in which AI tools designed to protect us from distractions compete with AI tools designed to distract us.
Stuart Mills, an economist at the University of Leeds, says tools like ChatGPT simply automate bureaucratic inefficiencies and don’t improve productivity at all.
We once asked a friend, a senior manager at a global engineering firm, if he used ChatGPT at work. “Oh yes,” he cried enthusiastically.
“I use this to produce reports that management keeps asking me for. I know no one will read it, so it doesn’t have to be high quality.”
For long-term productivity gains
AI has great potential to improve productivity at a societal level in the long term, and some of these improvements could be transformative.
As of September 2022, a study found that 5.7% of peer-reviewed research published worldwide will be on AI, up from 3.1% in 2017 and 1.2% in 2000. ing.
It’s clear that innovators around the world are exploring how AI can improve productivity and perhaps aid discovery. You can expect effective solutions that truly solve problems to be self-selected and organically rise to the top.
Successful AI implementations require an understanding of the context in which the technology is applied. Choosing the right tool for the task at hand and using it the right way is essential. And before that, we need to address issues of process, governance, culture and ethics.
Stefan Hajkowicz works for CSIRO, which receives R&D funding from a wide range of government and industry clients.
Jon Whittle works for CSIRO, which receives R&D funding from a wide range of government and industry clients.
/ Courtesy of The Conversation. This material from the original organization/author may be of the nature of its time and has been edited for clarity, style and length. Mirage.News does not take any organizational positions or positions and all views, positions and conclusions expressed herein are solely those of the authors.
