Does AI Devalue Developers?
A prospective client recently asked if we offered AI-assisted developers at a reduced rate to our full-time, full-stack, full-human developers. The question gave me pause as I had to make sure I understood the question and its motivation before responding.
While thoughtbot does offer ML and AI app development services, we don’t offer AI-assisted developers at a reduced (or elevated) rate. Though not a replacement for a talented human, some developers may choose to use generative AI for enhanced productivity or creativity. Additionally, AI can be a good option for automating repetitive tasks, generating sample code snippets, or identifying concealed insights that free up developers to focus on more complex and creative tasks.
In May, at RailsConf, our own Louis Antonpoulos will be speaking about his experience in a months-long experience pairing with a machine intelligence to create a game. By asking open-ended questions of the LLM partner, sharing context regularly, and staying flexible with his thinking, Louis and his AI partner have co-created a product that is much better written and thoughtful than if he had worked alone or treated the LLM as a “write code for me” tool.
A Certain…Je Ne Sais Quoi
AI has certainly gotten better at generating code and automating tasks. However, when it comes to complex, nuanced tasks that require empathy, creativity, context, and a deep understanding of a subject, human developers are the only correct answer. Human developers have the ability to understand the broader context of a project, including unique user needs, business goals, and ethical considerations–all of these are used constantly to make judgment calls and trade-offs.
This Feels Familiar
It’s worth noting that while fundamentally different technologies, the excitement today around AI is not dissimilar from the previous buzz felt by blockchain, crypto, or the metaverse. Just as these newsworthy technologies aspired to disrupt industries, AI too has the potential to transform the way we design and build products. However, it’s important to understand the limitations of AI by ensuring that it’s used responsibly and ethically.
Should you choose to work with AI-generated code, one of the significant challenges becomes the complexity of debugging. Since the code is generated by a machine learning model, it can be difficult to understand and debug, especially for complex issues. While it may be quickly produced, the required adjustments over any period may outweigh its initial benefits.
Generative AI models, which can be used to create various types of content–code, prose, lyrics, images, or videos, often struggle with issues related to copyright infringement. Since models are trained on large datasets of existing content, it should be no surprise that generative AI has inadvertently generated content closely resembling copyrighted material. This raises legal and ethical concerns, highlighting the need for careful oversight and regulation when using generative AI in any context, especially software development.
So What’s the Verdict?
While AI has the potential to enhance software development in many ways, it’s not without challenges and should not be used without the proper guardrails. Revisiting the original question of whether or not a consultancy might reduce its rates through the adoption of AI, I believe it’s unlikely. AI will continue as a tool that helps us be more efficient with our time. However, I see no advantage for client or consultancy to reduce the rates of the initial creation of lines of code at the expense of a much costlier debug and refactor later. With this in mind, we can say confidently that AI is helping thoughtbot developers be better consultants, not cheaper ones.
By approaching AI with a human-centric mindset and understanding its limitations, developers, like many other disciplines, can use its power to innovate and improve the way we build software. If you would like to learn more about how thoughtbot can help elevate your team and your product using the power of AI within the guardrails of our Playbook for software design and development, let’s talk.
