Perhaps the most promising space for AI thus far has been software development, where it seems to be increasingly adept at augmenting human capabilities. Although a limited subset of skilled builders is experiencing significant productivity gains, the impact remains far from covering the estimated $1 trillion in AI investments expected by Goldman Sachs over the next few years? As Covello explains, “Inverting low-skilled work, such as content creation, to require highly specialized expertise is essentially the opposite of what we’ve observed in past technological shifts, including the rise of the web.”
We’re far too hasty in assuming that AI infrastructure costs will plummet significantly and soon enough to make it a viable alternative for most tasks currently. The plummeting value of servers, which contributed to the dot-com boom, prompts Covello’s observation: “As people adjusted to a drastic decline in server value over just a few years from their inception in the late 1990s, the number of $64,000 Solar Microsystems servers needed to power the web technology transition in the late 1990s pales in comparison to the multitude of costly chips required to fuel the AI transition today?” Additionally, he fails to consider the considerable energy and other costs that combine to make AI significantly more expensive.
After 18 months of generative AI’s global debut, no groundbreaking or budget-friendly applications have emerged, rendering Covello’s verdict a scathing critique. According to MIT Professor Daron Acemoglu’s forecast, within the next decade, up to 23% of tasks that AI can moderately replicate may be economically viable to automate for the foreseeable future.
