Not so long ago, the idea that Artificial Intelligence or AI could threaten my livelihood was the stuff of science fiction. Suddenly, many of us are now questioning what the latest quantum leaps in AI technology might mean for knowledge workers and consultants whose value proposition is based on formal training, hard-earned expertise, human judgment and experience.
In a matter of seconds, so-called deep learning algorithms or generative AI platforms like Bing AI, Bard and ChatGPT can write a thoughtful birthday poem for mom, come up with relevant blog topics for a digital marketing business, code a piece of software based on specifications, write a term paper on feminism in Shakespeare, help you role-play a promotion request to your boss or produce a poster of a unicorn sliding down a rainbow in outer space.
The Chatbotification of Knowledge Work
Considerable euphoria and evangelical fervor has greeted this emergent AI technology. But it has also been met with fear, panic and dire warnings. Backed by corporate behemoths, a homogenous cohort of tech nerds who share the same world view and political ethos have designed a revolutionary new technology that could have profound effects on our collective futures, including the radical displacement of human labour. What could go wrong? The more immediate and profane preoccupation of this article, however, is whether generative AI will eliminate human jobs currently done by knowledge workers.
According to Sam Altman, the CEO of OpenAI (the company that created ChatGPT), a seismic disruption of labour markets is not only inevitable, it is desirable because “the marginal cost of an AI doing work is close to zero” relative to the cost of highly skilled and highly paid labour. Check out Altman’s eye-popping interview with Ezra Klein from the New York Times and Vice’s counterargument to Altman’s contention that AI work is “low cost”.
“We’re all going to be replaced by robots!”: version 4.0. It’s a familiar war cry that has echoed through the ages. I must admit that my first instinct was to assume, rather pompously, that the rarified and extremely specialized tax consulting services that I provide cannot be mimicked, replicated or automated by an AI bot. Admittedly, there are vast swaths of tax that are already managed through automation and AI technology, such as high-volume, low-complexity corporate tax filings. But surely the kind of bespoke, fact-specific, high-complexity, premium services that seasoned professionals deliver are, by their very nature, shielded from chatbotification? Right?
Here are 14 reasons why ChatGPT (as a proxy for all generative AI) will not take over our jobs… yet:
- ChatGPT is still pretty clunky and makes dumb mistakes that those endowed with human judgment would not make.
- ChatGPT has no ability to exercise qualitative discernment. It cannot tell the truth from a lie; it is merely interested in data.
- ChatGPT has no innate capacity for comprehension and meaning-making, although the large language models (LLMs) on which ChatGPT is built can translate, interpret, predict, generate and even learn information. To paraphrase a tech friend of mine, LLMs are limited to simplistic interpretations based on strict semantic analysis. By contrast, the human mind is really good at synthesizing disparate and multi-modal information into unique combinations.
- ChatGPT has no clue whether all the correct or relevant questions have actually been posed to help solve your problem.
- ChatGPT intelligence is deep and narrow and not designed to solve complex, interconnected problems. It cannot identify let alone address the ancillary or related problems that could be induced from the questions that you are asking.
- ChatGPT (or OpenAI) is not a neck that you can choke or an entity that can be held legally liable for screwing up, although this could change
- ChatGPT cannot underwrite the risk of providing bad advice. This is partly related to the previous point. ChatGPT leaves those who are foolish enough to rely on its advice in high stakes situations, completely exposed.
- ChatGPT cannot be audited.
- ChatGPT is not a source of truth recognized by regulators or one that meets established industry standards. For example, no self-respecting accounting firm will sign off on a company’s specific tax filing position based on ChatGPT.
- ChatGPT does not care about your privacy and confidentiality. It is under no obvious legal or ethical obligation to safeguard or protect the sensitive information that you disclose to it, including the responses to your queries that it generates.
- ChatGPT is not morally accountable nor is it subject to ESG (Environmental, Social and Governance) factors or fiduciary standards.
- Regulated professions like law, finance, medicine, education, etc. will not go gently into the night. Never underestimate the power of NIMBY. Regulated professions will almost certainly lose ground around the edges but will be harder to replace over the long run, compared to unregulated occupations that are governed primarily by the rules of the market.
- ChatGPT may be a new Trojan horse for threats to cyber security.
- ChatGPT hallucinates. It can malfunction in bizarre and harmful ways that have no place in a professional environment (or for that matter in a civil society).
Basically, it comes down to trust, scalability, moral accountability, “auditability,” liability and a sprinkle of technophobia. Still, if you are skeptical about the human-generated reasons listed above for why knowledge jobs are safe for now, perhaps you might be more convinced by ChatGPT.
What ChatGPT said about ChatGPT stealing our jobs
Out of curiosity, I asked ChatGPT to list the reasons why it would not take over knowledge jobs. Pretty meta, not to mention ironic, right? Here’s what ChatGPT said, in summary:
- Limited contextual understanding
- Lack of domain expertise
- Ethical and legal considerations
- Human interaction
- Creativity and innovation
- Human judgment and decision making
- Economic and social considerations.
Conclusion
Looks like ChatGPT and my human brain are largely aligned. There is no lack of media eulogizing the virtues of AI chatbots. Yet, I would be remiss if I did not pause to acknowledge the awe-inspiring capabilities and extraordinary potential of this fast-moving technology. This is really cool stuff. The debate over whether new machines offer better solutions or whether they should replace humans, has been raging, in some form or another, since before the industrial age. Putting aside the whole annihilation of human labour scenario (or other apocalyptic event), a far more interesting question than “will ChatGPT eat my lunch” is: how can I use ChatGPT and other generative AI to be better at my job and to service my clients more efficiently and effectively?
How do you see generative AI impacting your work in the near to medium term?
Sandra Rosier is a tax consultant based in Toronto, Canada. You can find her at rosiertaxadvisory.com
Recommended Resources:
Our Final Invention: Artificial Intelligence and the End of the Human Era, James Barrat
Power and Prediction: The Disruptive Economics of Artificial Intelligence, Ajay Agrawal, Joshua Gans and Avi Goldfarb
AI for Diversity, Roger A. Sǿraa
Speak, Louise Hall
Klara and the Sun, Kazuo Ishiguro
Global majority perspectives on AI: https://www.linkedin.com/pulse/12-black-women-ai-paving-way-better-world-caroline-lair
#ChatGPT
#generativeAIrevolution
#ChatGPTiscomingforyourjob
#largelanguagemodels
#AIhallucinations
#thefutureofwork
#AIandTax
Acknowledgements: Warmest thanks to the leadership team at HG Learning Solutions for helping me think through the heady issues raised by generative AI.