Aravind Srinivas, the CEO of Perplexity AI, joined a discussion about AI‘s impact on software engineering by commenting on a viral post.
The original post was by a physics and AI/ML student @TheVixhal, which talked about large language models, taking over routine, coding tasks, and shifting computer science back towards maths and physics foundations.
This post gained over 15,000 likes and around 1 million views, on which Aravind Srinivas commented, “Well said.”
Aravind Srinivas Backs Viral Post on AI Reshaping Software Engineering
The same opinion was also shared by the Anthropic CEO. Dario Amodie, who stated that within 6 to 12 months, AI will be able to handle most end-to-end tasks that are currently performed by software engineers. He also stated that some engineers at Anthropic have already stopped writing code entirely and rely on AI systems instead.
Another CEO to support this opinion was Replit’s CEO, who said that the traditional role of a software engineer may sort of disappear.
A 2023 Microsoft study shows that developers who use GitHub Copilot completed tasks 55.8% faster. Anthropic AI exposure index also states that around 75% of programming tasks can be handed by LLM’s, the highest level among all professions studied.
This shows that the nature of software engineering is changing with a reduced focus on repetitive coding and a greater emphasis on deeper problem-solving.
Tasks that are routine, like boilerplate coding, are increasingly automated, and engineers now focus more on system behaviour, trade-offs, and scalability issues.
However, there are a few concerns as LLMs still struggle with complex and unfamiliar system design problems.
Experts Say AI Can’t Replace Senior Engineers Despite Increased Automation
Critics have pointed out that while junior developers may benefit significantly from AI tools, experienced engineers are still needed for validation and decision-making. Amodei’s prediction about rapid AI takeover mainly applies to existing tasks and does not fully account for the challenges of creating new systems.
However, despite these limitations, the shift is strong enough that Code.org’s founder has suggested changes in computer science education where the focus can be reduced on syntax and shifted to logical thinking. He has penned this transformation by stating, “Coding is dead,” and “Long live coding.”
