Threads Dev Interviews: I am currently using Threads to find developers and interview them directly on the platform. If you’re interested in being interviewed, feel free to let me know on Threads. Please note that the views expressed in these interviews are personal and do not represent the interviewee’s employer.
Today, we have @vpalepu discussing the concerns of software written by AI in the era where AI can write code. @vpalepu believes that coding is both an act of expressing and understanding human needs. As long as we remember this when using AI to write code, there shouldn’t be much to worry about. However, @vpalepu does have some concerns and excitement about the possibilities of AI-generated code.
Firstly, @vpalepu is concerned about the idea that AI can write both code and tests. They believe that there needs to be human involvement to ensure the correctness of the code. They also worry about the widespread use of AI-generated code and its impact on various aspects of life, such as car software. @vpalepu reflects on the responsibility and liability when AI makes mistakes and wonders who should be accountable for fixing them.
Moving on to @vpalepu’s educational and work background, they currently work at Microsoft on the Outlook product. They hold a PhD in Software Engineering from UC Irvine and studied computer engineering in their undergraduate program.
When asked about the future of software, @vpalepu is excited about the concept of no-code solutions. They believe that these tools empower non-programmers to be more involved in building software systems. They also see the potential for automation to positively impact various aspects of life, such as routine tasks.
However, @vpalepu acknowledges that there are drawbacks to no-code solutions. These tools have limitations in terms of customization and can lead to complex systems that may be better suited for traditional programming. They highlight the need for deliberate limitations in no-code tools to avoid creating flaky and hard-to-maintain systems.
When asked if “no-code developers” may emerge as a job title, @vpalepu believes it is a possibility. They compare it to professions that require expertise in tools like Excel, which created an entire economy of problem solvers who are not traditional programmers.
Regarding the use of AI in their current work, @vpalepu utilizes AI for tasks such as writing documents, creating presentations, generating meeting summaries, and literature reviews. They have yet to use AI for writing code but express a cautiousness in trusting AI systems for coding tasks.
One area of technology that currently interests @vpalepu is software testing. They find it fascinating how AI can potentially assist in writing software tests, as it allows for defining correctness without specifying the algorithm. @vpalepu wonders if AI can understand the intent behind the code to help generate tests.
In terms of research and products related to AI in software testing, @vpalepu mentions tools like Evosuite and Pex that use genetic/evolutionary algorithms or symbolic execution. However, the acceptance and maintenance of these tests by human developers and their accuracy in capturing the original intent of the code remain a challenge. The introduction of GenAI adds an interesting dimension as it claims to understand intent, although the extent of its understanding is still uncertain.