Thriving in the Age of AI: Where Humans Have the Advantage Toby Carrodus explores how individuals can stay relevant and leverage AI to their advantage while human creativity and interpersonal skills remain irreplaceable in an AI-driven world.

Artificial intelligence is coming for your job. Humans are soon to be redundant. This is what many would have you believe. The truth is, scaremongering gets your attention, doesn't it? Perfect clickbait. But such statements belie the lessons of history.

According to quantitative researcher Toby Carrodus, resistance to new technology is nothing new. When John Henry Ford presented the first affordable family car to the world, immediate reactions were disapproving. Many feared that replacing horses with cars would cause mass unemployment by putting horse stables, farriers, and manure sweepers out of work. However, an unforeseeable new economic ecosystem arose around the automobile. New jobs were created, ranging from mechanics to gas stations and lacquerers.

In fact, this phenomenon, known colloquially by some as FOBO: "Fear of Obsolescence," was documented as early as the 16th century, when Queen Elizabeth I rejected a patent for a knitting machine for fear of the impact it would have on weavers. At its essence, FOBO is also a fear of the unknown. However, like the lessons provided by Ford, history has some timeless lessons on how we can stay relevant in an age of technological change.

Interpersonal Skills

Despite the rise of computers, we still share this planet with 8 billion human beings. The more people you get along with and have good relationships with, the more opportunities and better odds of success you have in life. It is in your best interests to ensure you have the skills to get along with as many types of people as possible because it is still people who debate and create laws, found companies, and buy products and services – not computers. Everything that computers do, they do for people.

Historically, we evolved in tribes because this ensured the greatest probability of survival. We depended on each other's strengths and benefited from the diversity of skills within the group to survive. In modern times, although our immediate tribes may have become smaller in membership, we are more likely to be members of several different 'tribes' simultaneously.

If the individual members of these different tribes each have a unique set of skills, we have access to a far greater breadth of skills and abilities than our ancestors did. In their famed summary of human history, "The Lessons of History", published nearly 60 years ago, renowned historians Will and Ariel Durant noted that "the men who can manage men manage the men who can manage only things" – things like computers. This lesson was as true then as it is now. Being able to cooperate, influence, and negotiate are still innately human characteristics and are key to succeeding in a world of 8 billion humans.

Drawing from his extensive career as a quantitative analyst, Toby Carrodus notes having observed this firsthand, having made great use of so-called "machine learning" techniques. One notion Carrodus expressed is that it is often those quantitative analysts who master relationships as well as their subject matter that get ahead. Computers and AI are unlikely to change this.

Creativity

Another key aspect where humans trump AI is in creativity. Let's not forget that AI is trained on content produced by humans. Recent research shows that AI trained on AI-produced content produces nothing more than gibberish.

For example, recent research published in Nature studied the results of an AI large language model (LLM) trained on original, human-generated content vs AI-generated content. The researchers compared output from the LLM trained on human-generated content with that produced from 9 successive iterations of the LLM model trained on its own AI-generated content after having initially started with the human-generated content.

The result? The output of the LLM model trained on its own AI-generated content was nonsensical babble that bore no semblance to reality! The researchers concluded that the model "becomes poisoned with its own projection of reality," becoming more and more disconnected with each iteration.

This bodes well for domains that rely on our innately human ability to create new and novel concepts. Carrodus witnessed this personally as a quantitative researcher focused on algorithmic trading models. AI models can be successful in identifying combinations of parameters that perfectly explain the past but have minimal predictability for the future due to certain statistical properties (or lack thereof) in markets.

Toby Carrodus explains that most AI-trained algorithmic trading models suffer from a lack of stability in their statistical estimates of markets (known as "stationarity"). A basic consequence of this is that just because you estimated some statistical parameters on historical data does not mean you can reasonably expect these estimates to remain valid when applied to future data. Furthermore, by taking a purely data-oriented, statistical approach without any domain knowledge, the true cause of any effect often remains unknown.

This becomes important for understanding when a perceived effect (in this case, a trading signal) may stop working. This difficulty is true of most human domains, which are characterized by social – or "complex" – phenomena, where we lack the ability to run parallel experiments in parallel universes.

In contrast, AI may be most useful in studying physical phenomena, where statistical properties such as "stationarity" are less of an issue. For example, recently, Microsoft partnered with the Pacific Northwest National Laboratory to narrow 32 million theoretical materials down to just 18 that could be used to reduce Lithium usage in batteries by 70%. The scientists still need to conduct the work, but AI has helped them narrow the scope of their research and reduced the costs of conducting such large-scale searches.

Having said all that, it is worth noting that AI has demonstrated some creative ability in the realm of, say, art. AI is capable of creating beautiful, fake images of fantasy landscapes and artworks where there is no right or wrong answer. But in areas where there is one right or wrong answer, AI still depends on human-generated content and skilled human input.

Use AI to Your Advantage

Technology is only going to be more a part of our day-to-day lives as time goes on. Rather than withstand it and claim to be a victim of its rise, it is in our interest to leverage it in our daily tasks where possible. Fears about AI in its current form consuming large swathes of our working world are, on balance, overblown. An AI LLM still depends heavily on human input to be trained on and to interpret its output.

By focusing on the innately human attributes of your work, such as creativity and interpersonal skills, you can set yourself up to thrive in an age of AI. New jobs around the AI ecosystem are already emerging, such as "prompt engineers", who are paid to, well, ask better questions of LLMs such as ChatGPT. It would, therefore, be wise to learn how you can leverage these technologies and, most importantly, not forget the fellow tribe members you share this planet with if you want to have a successful and fulfilling career.

About Toby Carrodus

Toby Carrodus is a quantitative researcher focused on algorithmic trading. He has worked in Frankfurt, London, Sydney, and Los Angeles for firms such as PIMCO and Winton Capital. Toby Carrodus also spearheads a scholarship fund for students from low socioeconomic backgrounds, leveraging his experiences to enable the next generation of learners. Follow him on LinkedIn, X, or Substack.