Bill Gates has revisited one of the most uncomfortable questions of the artificial intelligence revolution: which jobs will still make sense when models are capable of writing, summarizing, reasoning, programming, analyzing data, and performing digital tasks with decreasing human intervention? His response has attracted attention because he doesn’t list a long set of protected professions but highlights a few areas with greater resilience: programming, biology, energy, and professional sports.
A straightforward interpretation might be that almost everything else is doomed. But that is too simplistic. AI does not eliminate professions overnight; it first automates tasks, changes workflows, reduces demand for repetitive profiles, and raises the bar for those who remain. What Gates is pointing out is not a perfect boundary between safe jobs and lost jobs but a more useful distinction: there are jobs where value lies not just in producing an answer but in understanding complex systems, manipulating the physical world, or generating human interest.
For a tech-focused medium, the most interesting case is that of programmers. Because programming is both one of the areas most affected by AI and one that Gates considers more resistant.
Code will be cheaper; understanding systems will be more valuable
Programming assistants can already write functions, generate tests, review errors, explain logs, document APIs, convert code between languages, and create prototypes from descriptions. Tasks that once took hours for a junior developer can now be completed in minutes with well-utilized tools.
That doesn’t mean programmers will disappear. It means the core of the profession is shifting.
In real projects, the problem is rarely writing an isolated function. The challenge is whether that function fits into an architecture, scales properly, introduces technical debt, poses security risks, breaks dependencies, complicates maintenance, or solves the wrong problem. AI can generate code effortlessly, but it doesn’t always grasp business context, product history, previous technical decisions, or the consequences of modifying components within a large system.
In video game development, this is especially clear. A modern game is not just code; it’s graphics engines, physics, enemy AI, audio, networking, internal tools, optimization, animation, servers, patches, telemetry, monetization, and user experience. AI can assist in generating scripts, prototypes, behaviors, or utilities. However, transforming all that into a stable, enjoyable, and efficient product requires human judgment.
| Programming Area | What AI Can Do | What Still Needs Human Judgment |
|---|---|---|
| Repetitive code | Generate functions, tests, and documentation | Validate architecture and maintainability |
| Debugging | Suggest causes of errors | Understand context, logs, and side effects |
| Video games | Create scripts, prototypes, and tools | Refine gameplay, performance, and experience |
| Security | Review vulnerability patterns | Prioritize risks and avoid false positives |
| Enterprise software | Accelerate integrations | Decide on processes, data, and compliance |
A programmer who only translates simple instructions into code will face increased pressure. Conversely, those who understand products, architecture, data, security, performance, and user experience will hold more value. AI doesn’t diminish the importance of software; it devalues code produced without discernment.
Biology and energy: when AI touches the physical world
Gates also mentions biologists. This makes sense amid the explosion of AI applied to science, pharmaceuticals, proteins, genetics, and health. Models can analyze large datasets, find relationships, propose molecules, accelerate experiments, or detect patterns that would take human teams much longer to see.
But biology isn’t a spreadsheet. Living organisms have variability, context, side effects, ethical limits, and results that must be reproducible in labs or trials. AI can generate hypotheses, but science requires validation. A model may suggest a path, but someone must design experiments, interpret results, identify biases, and bear responsibility for decisions affecting health, drugs, or ecosystems.
The energy sector is another example of resistance due to physical complexity. AI can optimize power grids, predict demand, adjust maintenance, or improve plant operation. But energy relies on real infrastructure: generation, transportation, distribution, storage, regulation, industrial safety, and fault response.
This point becomes even more critical with AI itself. Data centers are increasing electrical demand in many countries. Advanced models require chips, cooling, networking, and continuous power. This makes energy experts a technical and strategic key. Algorithms that optimize consumption are not enough; skilled professionals are needed to plan infrastructure, negotiate capacity, manage risks, and keep critical systems running.
Sports, eSports, and the human element that automation cannot replace
The mention of professional sports is the most striking but perhaps also the simplest to understand. An AI can analyze millions of plays better than a human, simulate strategies, or play optimally in certain video games. But audiences don’t watch sports solely for efficiency—they watch for human stories.
People want to see pressure, mistakes, comebacks, rivalries, talent, fatigue, impossible decisions, and human gestures. This logic applies equally to eSports. A bot might outperform many professionals, but it doesn’t forge the same emotional connection as a human team competing live. The value isn’t just in execution but in narrative.
This difference will matter in more sectors than it seems. There are jobs where AI can perform a task, but cannot replace the social value of a human doing it. Entertainment, education, healthcare, consulting, and communication will still have spaces where trust, empathy, responsibility, and human presence are crucial.
The true boundary: automatable tasks versus responsibility
The debate about employment and AI often presents as a binary question: which professions will disappear and which will survive? The reality will be more nuanced. Within a single profession, some tasks will be highly automated, while others will be more valuable. An attorney, journalist, teacher, designer, or programmer won’t lose the same share of work—it depends on how much value they add beyond generating text, code, presentations, or generic analysis.
Gates’ list acts as a warning to the tech sector. The profiles that will resist best are those combining AI with deep domain knowledge. Knowing how to just use tools isn’t enough. Understanding the problem is essential.
In programming, that means shifting from writing code to designing systems. In biology, from data analysis to knowledge validation. In energy, from optimizing models to operating infrastructure. In competitive sports and gaming, from executing moves to creating human experiences that audiences want to follow.
AI will make many tasks cheaper but will also make judgment more valuable.
Frequently Asked Questions
Which jobs has Bill Gates identified as more resistant to AI?
He has mentioned programming, biology, energy, and professional sports as areas where full substitution will be more difficult.
Will AI replace programmers?
Not entirely. It will automate many coding tasks, but people capable of system design, architecture review, product understanding, and technical responsibility will still be needed.
Why might video game programming be more resistant?
Because a video game combines code, performance, design, experience, networking, internal tools, and creativity. AI can assist but cannot replace the entire creation and fine-tuning process.
Which tech roles are most vulnerable?
Those limited to repetitive tasks, simple code, generic content, or basic support without deep understanding of the system or business.
What should a tech professional learn to adapt?
Architecture, security, data, automation, product judgment, and advanced AI tool usage. The advantage lies in combining technical knowledge with decision-making capacity.
