The fear of AI destroying human creativity and the rise of AI detection tools have been hot topics over the past few years. After prizes being awarded to AI-generated books and students being failed for false AI detector outputs, it is time for us to look at three significant examples of how AI, AI-generated content, and failed AI detection can destroy trust, human connections, and opportunities. First, we have an AI-generated application and a corresponding Git repository, thereby reducing the number of slots available to genuine candidates. Second, we look at the experience at Bona Books, which recently had to fight with a number of AI submissions. Lastly, we get to look at my experience co-writing an article with a publication’s staff journalist, only to have their submission tool flag our work as AI-generated.
Ultimately, all the examples turned out well. Yet, it was never because of the AI or detection tools, but always because we managed to build a genuine connection.
AI Fakes Git Repository #
I recently had the pleasure of helping a startup founder and accelerator screen candidates for a GO developer position. The accelerator program’s staff acted as recruiters and were responsible for the initial screening of candidates. They would provide us with at most 5 pre-screened, ranked candidates. The strongest candidate seemed to have an impressive resume and a corresponding GitHub repository. A multitude of Pull Requests for some of the biggest Go-written projects in just the past few months. On the surface, it looked impressive. Yet, knowing about Git made one issue glaringly obvious. None of his Pull Requests had made it into an actual repository. Most were pending, and a number had been reopened with a generic confirmation that the code had not been AI-generated.
Going into the interview, it was confirmed that the candidate had no more than a cursory understanding of GO, and most of his answers appeared to be read directly from the Google search-integrated AI. Needless to say, the opportunity went to another candidate.
Yet, the issue highlights two problems. First, with a limit of 5 pre-screened candidates, we might have missed out on the best possible match for the position. Thus, the use of AI, at worst, took the opportunity away from a more deserving candidate. Second, the non-technical founder and the recruiter lacked sufficient technical expertise to hire a developer, highlighting the need to improve the hiring process and ensure the selection committee always includes subject-matter experts.
AI Costs Publishing Opportunities #
The loss of opportunities is also a key part of Bona Books’ discovery that two of the authors in their short-story work were using AI. Given the cost and length limits of paper books, accepting AI authors would mean fewer spaces for genuine creative humans.
Yet, the more fascinating part of the story is how they tried to dig into the backgrounds and methodologies of the two creatives, even interviewing them. In an industry marred by deadlines and limited funds, they could instead have relied on AI detection tools. They combined their own creativity with efforts to build a rapport with the suspected authors, giving them a chance to explain the discrepancies and the processes used to create the short stories.
Only when the attempt to build this connection failed did the editors decide to ban AI users and assign the respective spaces to more deserving authors. In short, it was the human element, not the computer, that decided whether something was generated or created.
When Tools Miss The Mark #
It undoubtedly was a good choice not to rely on AI tools, as my recent experience in publishing shows. A few weeks ago, I was co-writing a story with a staff journalist of a well-known business magazine. The writing took place in two video sessions, with the document on a shared screen. Thus, it would have been impossible for either of us to use AI without the other noticing. Unfortunately, when we tried to submit the first draft, the system outright rejected the document, deeming it 64.5% AI-generated. No appeals process visible, just an outright rejection from a publication’s internal system.
Not only did I feel devastated, but it also made me question how little internal trust there must be if a journal feels the need to police its salaried employees with such tools. From the outside perspective, it looked like management, d as the journalists were lazy and took no pride in their work. Why else would it be necessary to police them with such a flawed system?
AI Detection Lessons For The Board Room #
For the board, the three stories offer important lessons for building corporate culture. First of all, AI allows you to fake rudimentary competencies, yet it is impossible to fake deeper understanding. As such, processes that need to pass judgment on others, from recruiting and disciplinary to sales, cannot be fully automated. There always will be people who otherwise fall through the cracks, game the system, or get lost. Second, some opportunities are a zero-sum game. From speaking slots to job openings, it often is important to find the most deserving candidate. Thus, it is critical that we pass judgment humanely and with compassion. We should give people the opportunity to explain and reason with us. In short, we should work on building more human connections.
Thirdly, we will never win a war against AI slop by employing more AI. Instead, we must grow our competencies and capabilities. AI detection software is incredibly faulty, despite the fractional percentages and glossy websites. Thus, the human aspect is our only defense against AI.
A New Value For Creativity #
Ultimately, the question of how much we value creativity and genuine connections comes down to each of us. AI won’t be able to address the moral and cultural questions that will keep companies on the right track and our workforces engaged. Unless we build our own competencies and strengthen our trust in one another, we won’t be able to sustain a highly productive culture.
We must acknowledge that only humans are truly creative, while AI only recombines word salads. Only then can we learn from the latest AI problems in recruiting, journalism, and publishing, and build organizations that rely on trust, knowledge, and human connections.