Each day that goes by, I laugh a little deeper about “A.I will take jobs” claim

The Evolving Perspective on Artificial Intelligence and Its Impact on the Workforce

In recent months, I have had the opportunity to work closely with various artificial intelligence (A.I.) tools, including Google’s offerings, ChatGPT, Claude, and others. These technologies have proven to be remarkably helpful in streamlining tasks and enhancing productivity. However, despite their utility, I remain skeptical about the broader claims that A.I. will inevitably displace jobs across industries.

The Reliability and Limitations of A.I.

One of the most striking issues with A.I. systems is how significantly their responses depend on the framing of questions. For instance, if you ask a question with a particular bias or assumptions, the A.I.’s response may reflect that bias, making it both sycophantic and unreliable. To illustrate, imagine a Byzantine emperor known for severity, questioning his aide about the Turkish advancing forces: “Are the Turks approaching Constantinople?” The aide, eager to appease, might respond with exaggerated confidence: “Certainly not, your Majesty. No Turks are anywhere near Anatolia.” If asked again, the aide could craft a flattering narrative that bolsters the emperor’s confidence, regardless of the reality—highlighting how context and phrasing can influence outcomes.

Economic Viability and Energy Consumption

The companies behind A.I. technologies have attracted immense investment but still struggle to turn a profit. Their operations involve substantial energy consumption, which is a significant and ongoing cost. Companies like Claude claim to have limited long-term memory, retaining context only for brief periods of interaction to conserve energy. Given that energy costs are unlikely to decrease dramatically, this raises questions about the sustainability and trustworthiness of these technologies. If A.I. systems cannot reliably remember past interactions, their practical utility in complex workflows is compromised, discouraging widespread enterprise adoption.

Data Privacy and Ethical Concerns

Another critical issue is the data that users input into A.I. systems. Often, these inputs are recorded and stored, potentially being used later for training or other purposes. Many of these models have been trained on extensive datasets, which may include copyrighted materials or works created without permission. If such models become more profitable than traditional industries—similar to historical figures like Mansa Musa—there’s little incentive for companies to change their data practices. This creates ongoing ethical debates about intellectual property rights and the true cost of data used to train these models.

Accuracy and Workplace Dynamics

Despite their capabilities, A.I. systems can generate incorrect or misleading answers. Over