AI engineer role faces displacement as general models consolidate power
A computer vision engineer warns that the broad definition of AI engineering is becoming obsolete as large technology firms release versatile models that reduce the need for custom solutions.
A blog post published on 17 August 2025 argues that artificial intelligence engineers are likely to be replaced by general AI models sooner than other software developers. The author, a computer vision engineer writing on dmanco.dev and shared via Hacker News, contends that the term "AI engineer" is overly broad and that large technology companies are increasingly developing general foundation models capable of handling diverse tasks without the need for custom solutions.
The author points to Meta’s recent release of DINO, a versatile vision model, as a prime example of AI cannibalising other branches of the field. DINO can be applied to different tasks with minimal work and without annotations, offering a plug-and-play solution that reduces the immediate need for tailored research. This trend signals a shift where general models absorb various domains, making dedicated AI engineering roles less convenient for most companies.
Citing the book “AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference” by Arvind Narayanan and Sayash Kapoor, the post highlights the current marketing confusion surrounding AI. The author notes that while technologies like large language models, image processing algorithms, and classical pathfinding algorithms share underlying neural network fundamentals, they require different technical knowledge. However, the industry lacks clear differentiation, leading to a job title that encompasses everything from using ChatGPT APIs to complex computer vision work.
The author predicts that as these general models become more specialised with minimal data, the demand for dedicated AI engineering roles will decline. Research will concentrate within major tech firms, while the broader market relies on off-the-shelf solutions. Tailored AI solutions are expected to become a luxury that most companies will avoid, leading to a flattening of the AI progress curve as fewer engineers drive innovation.
While general models are not yet specialised enough for many domains, particularly where data is scarce, the author expects improvements within a few years. Software developers will still be needed to integrate these AI models into applications, as agents require user direction and oversight. The author suggests that AI agents cannot completely steal software developers' jobs at least in the near term, as someone must still possess the knowledge to ensure the AI is building the application correctly.
