I am now in the twilight of the career that I chose over three decades ago. This is not because I am old, mentally infirm, or am no longer a good software developer, but rather because my industry is going through a seismic shift, and my current chosen career will cease to exist.
There is no uncertainty as to whether or not it will cease to exist — it will; it is now just a matter of time. Many in my sector will say that it is impossible for a computer system to do what we do. On the contrary, this is possible and absolutely critical for the next stage of AI development.
The Shift Toward AI-Driven Software Engineering
Mark Zuckerberg announced a few weeks back that Facebook will be focusing on developing an AI Engineer in 2025. (Source)
Google already uses AI to create and maintain 25% of its codebase. These statistics will only increase as models improve, and it will become commonplace for firms to use AI to create and maintain the code that runs their platforms.
Why AI-Generated Code is Critical for AI’s Evolution
We are rapidly approaching the singularity — the hypothetical future point where artificial intelligence surpasses human intelligence, as referenced in Ray Kurzweil’s book, The Singularity is Near.
For AI to evolve, it needs the ability to adapt itself. For it to adapt itself, it needs the ability to create, evolve, and maintain its own code. To do this, it needs us to give it the ability to code. Once it has this ability, it will be able to write its own destiny.
Some may argue that AI already has this ability, but the current systems still rely on human oversight and are limited in context. They can only work on small codebases and lack the capacity to operate on the large-scale codebases required for AI to understand itself.
Three Key Areas That Must Improve
For AI to achieve the goal of consistently creating good code, three critical areas need to advance:
- Cost and Availability of Compute Power
The cheaper and more available compute becomes, the cheaper and quicker it becomes to train large language models. - More Efficient and Better Algorithms for LLM Training
Improved algorithms will reduce hardware costs, lower energy consumption, and enable real-time adaptation of models. - More higher quality Data
Expanding the training datasets will make LLMs richer by allowing for greater generalization and enabling emergent abilities like reasoning, coding, and multimodal understanding.
The Tipping Point
Would I bet against major developments in these three areas in 2025? No.
The largest organizations have devoted serious amounts of money to address each of these, and we will reach a tipping point where AI contributes to solving these problems. When AI is able to contribute to its own evolution, it may rapidly progress to the point where it no longer needs humans to improve. This will be the moment of singularity.
The Future of Human Developers
Once AI achieves the ability to create its own code, coding skills will no longer be a prerequisite for implementing creativity in software. Therefore, human developers will no longer be required to code.
Am I scared? No. Here’s why:
- It will still take some time before a fully autonomous AI Software Developer is developed.
- There will be limits — at least initially — on what a fully autonomous AI Software Developer can do.
- Deploying a fully autonomous AI Software Developer will be costly, so initially, they will be assigned to the most pressing business issues, leaving space for human developers.
- We will adapt to new careers within the industry where AI will not feature as strongly.
For over two decades, we have integrated digital into the human space; now, we will spend the next two decades integrating humans into the digital space.
How Developers Should Prepare
Right now, developers should:
- Embrace AI tools like Copilot, Code Llama, or Gemini and become experts in AI-assisted development.
- Learn AI concepts, including ML, NLP, and AI ethics, even if they are not AI engineers.
- Develop industry expertise in fields like public sector, finance, healthcare, cybersecurity, or biotech, where AI needs human oversight.
- Focus on creative, strategic, and human-centered roles that AI cannot easily replicate.
The Singularity Will Bring Unparalleled Innovation
I am excited! My career has always been one of change, and now change is happening faster. The singularity will rapidly bring innovation in all areas of life, including:
- Disease eradication and aging reversal
- Human longevity and healthcare transformation
- Food security and energy abundance
- Robotics, automation, and labor displacement
- Climate change mitigation and environmental restoration
Of course, it may also see humans as obsolete or a threat. Only time will tell!
