Home » Blog » No One Understands Anything About AI
I want to share two different narratives regarding AI. Both stories discuss the use of this technology to automate software development but lead to contrasting conclusions.
The first narrative highlights that Large Language Models (LLMs) are particularly suited for coding, as source code is fundamentally well-structured text, which these models excel at generating. Due to this close alignment between demand and capability, the programming industry is becoming an economic sacrificial lamb, being the first major sector to experience significant disruption from AI.
There is plenty of evidence backing these claims. Here are some instances, all from the past couple of months:
Aravind Srinivas, the CEO of the AI firm Perplexity, states that AI tools like Cursor and GitHub Copilot have reduced his engineers' task completion times from “three or four days to one hour.” He now requires every employee at his company to use these tools: “The speed at which you can fix bugs and deploy to production is alarming.”
An article in Inc. confidently proclaimed: “In the realm of software engineering, AI has changed everything.”
It is unsurprising that these powerful new capabilities are attributed to serious disruptions. One article from an investment site had an alarming headline: “Tech Sector Sees 64,000 Job Cuts This Year Due to AI Advancement.” No one is immune from these layoffs. “Major companies, including Microsoft, are leading these layoffs,” the article states, “citing advancements in AI as a major reason.”
My domain of academic computer science hasn't escaped unscathed either. A striking piece in the Atlantic opens with a troubling assertion: “The Computer Science Bubble is Bursting,” which it largely attributes to AI, a technology it describes as “ideally suited to replace the very individuals who created it.”
Given the assertiveness of these claims, one might conclude that computer programmers are quickly becoming obsolete like telegraph operators. However, if one reads a different set of articles and quotes from the same time frame, an entirely different narrative surfaces:
The AI evaluation company METR recently released findings from a randomized control trial where a group of experienced open-source software developers was divided into two groups—one using AI coding tools for a set of tasks and one not using any. The report summarizes: “Surprisingly, we found that developers using AI tools took 19% longer than those who did not—AI makes them slower.”
Meanwhile, other seasoned engineers are beginning to contest extreme assertions regarding AI's impact on their field. “Quitting programming as a profession right now because of LLMs would be akin to abandoning carpentry due to the invention of the table saw,” quipped developer Simon Wilson.
Tech CEO Nick Khami responded to the assertion that AI tools would significantly reduce the workforce needed to produce software as follows: “Every time I read this, I feel like I’m being misled, and I’m concerned it may lead those new to software development to feel it’s a poor investment of time.”
But what about Microsoft replacing all its workers with AI tools? A closer examination shows this is not the case. The company’s actual announcement clarified that layoffs were distributed across divisions (like gaming) to reallocate funds for AI projects—not due to AI taking over jobs.
What about the struggling computer science majors? Later in that same Atlantic article, an alternative explanation is provided. The tech industry has recently been contracting as a correction for excessive spending during the pandemic years. This soft market has an impact: “enrollment in computer science programs has historically vacillated with the job market... [and] previous declines have always rebounded to enrollment levels exceeding prior figures.” (A personal note: during my undergraduate studies in computer science in the early 2000s, there was concern about the steep decline in majors following the original dot-com crash.)
We see two entirely different perspectives on the same AI issue, contingent on which articles you read and which experts you heed. What should we derive from this disarray? Regarding the effects of AI, we still have no definitive answers. Yet this has not stopped people from behaving as though we do.
My advice for now:
Ignore both the most fervent and the most dismissive rhetoric.
Concentrate on concrete changes in areas you care about that indeed seem related to AI—read broadly and consult trusted individuals about their observations.
Moreover, approach AI news with skepticism. All of this is too fresh for anyone to accurately grasp what they’re asserting.
AI is significant. But we still do not fully understand why.
I would like to share two stories regarding AI. Both narratives focus on utilizing this technology to automate computer programming, but they highlight different aspects... Read more
