Avi Goldfarb is an accomplished economist and esteemed University of Toronto Rottman School professor. Goldfarb's academic journey began in the late nineties when the internet was a groundbreaking new technology. As a young scholar in economics, he was tasked with choosing an industry to specialize in. Recognizing the potential and novelty of the internet, Goldfarb seized the opportunity to focus his studies on this emerging field. Consequently, he dedicated his Ph.D. to the economics of the internet, conducting intensive research on search engine competition when it was a genuinely competitive landscape. His fascination with prediction and machine learning led him to co-authoring Prediction Machines - The Simple Economics of Artificial Intelligence.
Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto
I recently sat down to chat with Avi about the future of work trends as part of my research to produce season one of The Lever with Drew Fortin. This show highlights how the rise of artificial intelligence (AI), robotics, and web3 will shift the paradigm of humans at work for the better. I wrote this article based on my notes and transcripts from the interview. You can also watch a video of my interview with Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto, below.
Throughout the initial decade of his career, Goldfarb dedicated his efforts to unraveling the economic implications of the internet. His extensive work encompassed a range of areas, including the Digital Divide, the effectiveness of online advertising, privacy issues, and the use of the Internet in businesses.
One significant offshoot of his research into internet usage in businesses was exploring the healthcare sector. Goldfarb and his team applied their insights to electronic medical records, highlighting hospitals' challenges when adopting these systems. As is often the case for researchers, Goldfarb's investigative journey continued post-tenure.
In 2012, his colleague AJ Aggarwal launched an organization known as the Creative Destruction Lab. This initiative, which began in Toronto over a decade ago and has since expanded to 12 universities worldwide (including prestigious institutions such as Oxford and George Tech), aims to assist science-based startups in scaling their operations.
Interestingly, one of the first companies to join the lab claimed to be utilizing AI for drug discovery. In retrospect, this idea seemed far-fetched in 2012. The prospect of using artificial intelligence to unearth new pharmaceuticals was not on Goldfarb's radar, nor was it on most people's. However, the company turned out to be led by a student from Jeff Hinton's lab – Hinton is one of the pioneers behind the current enthusiasm surrounding AI and a key inventor of deep learning.
The following year saw the arrival of more AI-focused companies, primarily emerging from universities around Toronto. Recognizing this surge, Goldfarb and his co-authors AJ and Joshua of "Prediction Machines" identified an opportunity. Similar to how they had started their careers focusing on the then-mysterious internet, they saw a chance to delve into another novel technology largely unexplored – artificial intelligence.
Goldfarb observes two distinct paths in which AI is transforming workflows. The first path is incremental, where companies examine their existing workflow to identify specific tasks that AI can enhance. They then replace the current process, often human-led, with an AI solution without significantly altering the overall workflow. This approach can provide quick wins as it doesn't require substantial changes. However, Goldfarb notes that companies adopting this strategy often find the impact of AI underwhelming as the best it can achieve is minor cost savings.
The more exciting path, according to Goldfarb, involves a complete overhaul of workflows and responsibilities within a company. This systemic change aims to leverage AI to deliver superior products, serve customers differently, and better fulfill the organization's mission. Although this approach presents significant opportunities, only a handful of companies have successfully implemented it due to its risky and challenging nature. It also often meets resistance from influential individuals within the company who are comfortable with the status quo.
"The more exciting element is to think through how you can deliver a better product, serve your customers differently, and better deliver on the mission of the organization through AI. That involves completely changing your workflows and the people responsible for your company. That's what we call system-level change." - Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto
Goldfarb has observed that individuals with substantial power and responsibility within a company often resist AI-based changes, especially those involving a systemic transformation. These individuals tend to prefer minor adjustments to the workflow, such as replacing an existing process with a new one, rather than embracing a complete overhaul of the system. Therefore, while AI offers promising potential to revolutionize workflows, its successful implementation requires a willingness to embrace systemic changes and navigate the associated risks and challenges.
Goldfarb provided an insightful perspective on the implications of AI on job markets. He began by discussing studies that predict a significant loss of jobs due to AI. These studies typically analyze existing workflows and jobs and then compare them with the capabilities of AI. For instance, AI can perform prediction tasks, write, or translate languages. They quantify how many human tasks within current workflows could be performed by AI, leading to estimates like the World Economic Forum's prediction of 80 million jobs being replaced by AI in the global economy.
However, Goldfarb points out that these analyses often overlook the new opportunities AI brings. The advent of AI doesn't merely replace human tasks; it opens doors for businesses to do new things and improve existing processes. A prime example is the ChatGPT, which excels at writing coherent sentences, paragraphs, and even essays. While this might seem threatening to those whose jobs involve writing, Goldfarb argues that it also creates immense opportunities.
He acknowledged that individuals who rely on their language parsing skills, such as copy editors, may have cause for concern. Nevertheless, he notes countless people have limited job opportunities due to poor written communication skills. These individuals might possess a wealth of creative ideas or specialized skills, but their inability to effectively express these in writing hinders their career progression.
"There are all sorts of people who are amazingly skilled at things, or maybe they have all sorts of creative ideas, but they don't know how to write them down. With a tool that could allow them to write and write well, it could create incredible opportunities." - Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto
With AI tools like Chat GPT, however, these individuals can improve their written communication, effectively 'upskilling' themselves. This ability to write well and their other skills could unlock significant labor market value. In essence, Goldfarb suggests that the impact of AI should be viewed not just through the lens of job loss but also in terms of the new opportunities and systems it can create rather than merely incrementally improving the old ways of doing things.
When considering the future of work, Goldfarb highlights two significant aspects. Firstly, he raises the question, do we genuinely desire to work? He reminds us that work, by its very nature, is not always enjoyable. He points out that many of the 20th century's labor movement victories were about reducing work hours, allowing us time for childhood, adolescence, retirement, and leisure. Goldfarb argues that this would be a win for most people if technology enables us to maintain our living standards with fewer work hours.
Goldfarb's second aspect is the difficulty in identifying the new opportunities that technological disruption will create. While it's relatively easy to pinpoint the jobs technology might replace, it's more challenging to imagine the novel forms of work that could emerge. He draws parallels with the past, noting how someone from the 19th century would struggle to comprehend the jobs we do today. Similarly, the roles that computing and the internet have created were unimaginable in the 1930s. Although he doesn't claim to know what these new jobs will be, Goldfarb expresses confidence that they will appear.
In terms of AI capabilities, Goldfarb refers to his book, "Prediction Machines," which focuses on machine learning, a branch of computational statistics that has improved dramatically over the past few decades. Machine learning, or today's AI, excels at taking available information and filling in the gaps, a process known as prediction. Predictions are integral to decision-making, but Goldfarb emphasizes that judgment, the ability to determine which predictions to make and how to use them, remains an inherently human trait. In the foreseeable future, he believes this decision-making judgment component will continue to be a human responsibility.
"Judgment at a high level is knowing which predictions to make and what to do with those predictions once you have them, that remains inherently human and for the foreseeable future, figuring out which predictions to make and what to do with them once you have them is going to be human." - Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto
As Goldfarb points out, predicting the exact impact of AI on various sectors or the specific jobs it will create is virtually impossible. Yet, he remains confident that AI will continue to create new opportunities and transform the world of work.
Technology's impact on employment and job responsibilities can be pretty astounding. Take the advent of Automated Teller Machines (ATMs), for instance, which is a case study discussed in Goldfarb's "Prediction Machines." The widespread use of ATMs, as pointed out by economist Jim Bessen, didn't lead to fewer bank tellers as one might expect. Instead, it increased the population of retail bank employees. Once ATMs took over dispensing and depositing cash, bank employees were freed to take on more valuable roles such as financial management assistance, marketing new programs, and helping customers structure their banking lives.
This role shift was only possible because the introduction of ATMs expanded the scope of services that banks could offer. Similarly, we can anticipate comparable shifts in integrating AI in various sectors, although these changes have not fully materialized yet.
Consider the medical field, where the role of doctors is central due to their unique ability to diagnose conditions. They have years of training to interpret symptoms and deduce the underlying causes. However, diagnosis is essentially a form of prediction, and AI has proven exceptionally adept at making predictions. With AI potentially taking over diagnostic roles, what would the future of medicine look like?
One possibility is to replace doctors with AI systems, but this doesn't necessarily imply a substantial improvement in healthcare quality. It might reduce costs, considering the high expense associated with physicians, but the overall healthcare workflow would remain essentially unchanged.
Alternatively, we could envision a new breed of medical professionals who work alongside AI systems. These professionals, possibly still referred to as doctors, could focus on helping patients navigate the stress of the health system and health crises while the AI handles diagnoses. This new role might require just an undergraduate degree, focusing on managing medical crises – a task many doctors currently handle.
This approach would necessitate a complete healthcare system overhaul, from training to insurance processes. However, it could result in better, potentially less expensive care, with professionals concentrating on tasks machines cannot perform, such as providing emotional support. In this scenario, AI doesn't replace jobs but transforms them, improving the system as a whole.
In the work context, Goldfarb emphasizes the importance of human judgment and aligning actions with an organization's mission. In his book "Power and Prediction," Goldfarb emphasizes beginning with the mission and then identifying measurable aspects of the end objectives. From there, the organization must decipher the core decisions contributing to its mission. Measurement and predictions then come into play.
For instance, treatment choices and resource allocation are crucial decisions in a hospital. These decisions, influenced by the hospital's mission, could range from focusing on low-cost care to saving lives or addressing acute needs. Prediction technology can aid these decisions by improving diagnoses and anticipating needs. However, human judgment remains vital, especially when prioritizing resource allocation or life-saving measures. The complexities of these decisions highlight the irreplaceable role of humans in managing organizational objectives.
"What is the organization trying to optimize? What are we trying to achieve? What's our mission, and how do we measure it? These questions are critical and remain inherently human because that's not what prediction machines do." - Avi Goldfarb, Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto
As we wrapped up our enlightening conversation, Goldfarb succinctly summarized the matter: "Prediction is the process of filling in the missing information. That's it. That's what AI does. That's the reason we're talking about AI today. Prediction is a big deal because it helps us make better decisions."
Hear more from Avi Goldfarb, Co-author of Prediction Machines and Chair in Artificial Intelligence and Healthcare, Professor of Marketing at the Rotman School of Management, University of Toronto: