Replacing coders with AI? Why Bill Gates, Sam Altman and experience say you shouldn’t.
In the race to automate everything – from customer service to code – AI is announced as a silver bullet. The story is tempting: AI tools that can write entire applications, can streamline technical teams and reduce the need for expensive human developers, along with hundreds of other jobs.
But from my point of view as a technologist who spends every day in the data and workflows of real companies, the hype does not match reality.
I have worked with market leaders such as General Electric, the Walt Disney Company and Harvard Medical School to optimize their data and AI infrastructure, and this is what I have learned: Replacing people with AI in most jobs is still just an idea on the horizon.
I am worried that we think too far ahead. In the past two years, More than a quarter From programming lanes have disappeared. Mark Zuckerberg announced He plans to replace many of the coders of Meta with AI.
But intriguing, both Bill Gates and Sam Altman have publicly warned against the replacement of codingers.
At the moment we do not have to count on AI tools to successfully replace jobs in technology or company. That is because what AI knows is inherently limited by what it has seen – and the majority of what it has seen in the technical world is boilerplate.
Generative AI models are trained on large datasets, which usually fall into two main categories: publicly available data (from the open internet), or own or licensed data (made in-house by the organization or purchased from third parties).
Simple tasks, such as building a basic website or configuring a template app, are simple victories for generative models. But when it comes to writing the advanced, own infrastructure code that feeds companies such as Google or Stripe, there is a problem: that code does not exist in public repositories. It is locked up within the walls of companies, inaccessible for training data and often written by engineers with decades of experience.
At the moment, Ai cannot reason Already in itself. And it has no instincts. It only mimics patterns. A friend of mine in the technical world once described large language models (LLMS) as a ‘really good goker’.
Think today of AI as a member of the Junior Team – useful for a first concept or simple projects. But like every junior, supervision requires. When programming, for example, while I have found a 5x improvement for simple coding, I have discovered that assessing and correcting more complicated AI-produced code often costs more time and energy than writing the code itself.
You still need senior professionals with deep experience to find the mistakes, and to understand the nuances of how those defects can be a risk in six months.
That does not mean that AI should not have a place in the workplace. But the dream of replacing entire teams of programmers or accountants or marketers with one person and a large number of AI tools is much premature. We still need people at senior level in these jobs, and we have to train people in junior level jobs to be technically possible enough to fulfill the more complex roles in one day.
The aim of AI in technology and companies should not be about removing people from the loop. I don’t say this because I am afraid that AI will take my job. I say it because at this stage I have seen too much too much how dangerous ai trust too much.
Managers, regardless of the industry, must be aware: although AI promises cost savings and smaller teams, these efficiency gains can have an adverse effect. You could trust AI to perform more junior work levels, but not to complete more advanced projects.
Ai is fast. People are smart. There is a big difference. The sooner we move the conversation from replacing people to strengthening them, the more we pick the benefits of AI.
Derek Chang is the founder of Stratus -data.
