What jobs will Generative Artificial Intelligences (GAIs) replace?
A first-principles analysis
Tasks. Not Jobs.
Undertaking a single occupation often entails executing a wide array of tasks, as exemplified by the software engineer’s multifarious responsibilities, which include coding, documentation, peer review, architectural design, and even recruitment, among others.
Thus, a more incisive question would be, “Which tasks shall Generative AIs replace?” Occupations on the verge of obsolescence are those in which the lion’s share of tasks are subject to GAI intervention.
To determine the scope of tasks GAIs can replace, we must first examine the capabilities of these artificial intelligences.
Questions and Types of Questions
In essence, GAIs (and their AI counterparts) serve a singular purpose: providing answers to questions. These responses, in turn, pave the way for prediction, optimization, personalization, automation, informatization, and a plethora of other applications.
For the purposes of this analysis, we shall delineate questions along two axes.
- The first axis concerns the deterministic validation of the answer. For instance, a question such as “Will it rain tomorrow?” has a deterministic answer, while the inquiry “How might Picasso reimagine the Mona Lisa?” does not.
- The second axis pertains to the precedent of satisfactory answers. To illustrate, the question “What is the square root of 1700?” has been answered in the past, whereas the query “What is the precise nature of anti-matter?” remains unresolved.
What Generative AIs can do
Based on these dimensions, we can classify questions into four categories:
- C1. Deterministic questions with past answers: These are questions that can be validated deterministically and have been answered in the past. Examples include solving mathematical problems, generating code for specific tasks, or retrieving factual information. GAIs can excel in these tasks by leveraging their vast data access and learning from past experiences.
- C2. Deterministic questions without past answers: These questions can be validated deterministically, but no satisfactory answer exists yet. Examples include discovering new scientific theories, developing new engineering solutions, or creating innovative business strategies. GAIs can contribute to these tasks by generating hypotheses, potential solutions, or novel ideas that can then be tested and validated by humans.
- C3. Non-deterministic questions with past answers: These questions cannot be validated deterministically but have been answered in the past. Examples include creating artistic works, writing persuasive speeches, or designing user experiences. GAIs can learn from past examples and generate new content or solutions that appeal to human sensibilities.
- C4. Non-deterministic questions without past answers: These questions cannot be validated deterministically, and no satisfactory answers exist yet. Examples include addressing new ethical dilemmas, creating entirely new art forms, or predicting the impact of unprecedented events. GAIs might struggle with these tasks, as they require a deep understanding of human values, emotions, and creativity that may be difficult for AI to grasp.
Now, we could be simplistic and pigeon-hole various jobs into exactly one of the categories above. However, the world is far more complicated. Most jobs have multiple tasks which fall into different categories.
Let’s consider some examples.
How GAIs will affect various jobs
Building Architects (C1, C2, C3)
Architects need to design structures that meet specific requirements while adhering to safety regulations and building codes (C1). They also have to come up with innovative and efficient design solutions for new types of buildings or unique challenges (C2). Additionally, architects need to create visually appealing designs that consider aesthetics and the human experience (C3).
Data Scientist (C1, C2, C4)
Data scientists analyze large datasets, utilizing established statistical and machine learning techniques to draw insights (C1). They also develop novel algorithms or models to solve specific problems or improve existing methods (C2). In some cases, data scientists may face uncharted territory, such as dealing with new types of data or addressing complex ethical issues related to data privacy, algorithmic bias, or surveillance (C4).
Journalists (C1, C3, C4)
Journalists must gather and verify information from various sources, ensuring accuracy and objectivity (C1). They also need to present the information in a compelling and engaging manner, often using narrative techniques or multimedia elements (C3). In some cases, journalists may tackle unexplored subjects, investigate complex issues, or provide new perspectives on controversial topics, requiring them to navigate ethical dilemmas or uncertainties (C4).
Environmental Policy Analyst (C2, C3, C4)
Environmental policy analysts conduct research to develop evidence-based policies and recommend solutions to environmental problems (C2). They must also communicate their findings persuasively to diverse audiences, such as policymakers, stakeholders, and the public (C3). Moreover, they may face unprecedented challenges, such as predicting and mitigating the effects of climate change or addressing newly emerging ecological threats (C4).
Concluding thoughts
Generative Artificial Intelligences (GAIs) have the potential to replace or augment tasks across various jobs, depending on the types of questions involved.
Jobs with tasks mainly falling under C1 (deterministic questions with past answers) are likely to be the most susceptible to automation, as GAIs can leverage vast data and past experiences to generate accurate and efficient solutions.
However, as the complexity of tasks increases and involves more of C2, C3, or C4 categories, the role of GAIs shifts from outright replacement to augmentation, collaboration, or generation of ideas that require human validation, creativity, and ethical judgment.
The multifaceted nature of human work, as demonstrated by the examples of architects, data scientists, journalists, and game designers, highlights that while GAIs have transformative potential, they are unlikely to replace entire jobs spanning multiple dimensions, at least in the near future.
And so, as we march boldly into the future, it seems the symphony of human ingenuity and artificial intelligence will continue to play in harmony, while the prospect of a GAI conductor usurping the maestro’s baton remains, for now, a mere encore fantasy.
