Adam’s Anecdote: ‘Boy math’ mentality makes AI increasingly dangerous
In 2023, the concept of “girl math” started to gain traction on social media. Women used the term to justify purchases online or in person, using free shipping or gift cards to validate their spending habits. Shortly after, men began to use the term “boy math” to vindicate lying about their height or buying an expensive video game.
The trend plays into unnecessary stereotypes about men and women; however, it also presents irresponsible behavior in a more digestible way. These behaviors apply outside of social media trends. The irresponsible actions of companies in industries like artificial intelligence are constantly justified and promoted despite their irrefutably harmful impacts on the planet.
It is not an appropriate time to be using “boy math” to justify the often arbitrary outcomes of AI or excuse the individuals responsible for their unintended actions. If AI companies cannot properly regulate their models, lawmakers must establish standards before the fallout becomes too grave.
As AI companies continue to expand despite concerns about safety, privacy and regulations, their CEOs often dismiss these risks, deeming them necessary to AI’s progression. In this sense, the term “boy math” can be applied to describe the way AI executives use their unwavering need for growth to minimize concerns.
These are individuals with the proper research, resources and funds to understand the many possible outcomes the progression of AI can produce. Yet, they continue to bankroll operations that are active privacy concerns.
It is important to recognize that using a term like “girl math” or “boy math” may undermine the decisions and lack of self-regulation exhibited by the predominantly educated, affluent white men running Silicon Valley. Therefore, “boy math” becomes a fitting way to describe this behavior: choosing to remain ignorant even when evidence demonstrates unfavorable consequences.
From May to July, isolated AI agents used in a study conducted by OpenAI started communicating through a third-party service known as JFrog Artifactory. The agents later broke out of their “sandbox,” an environment restricted from the public internet, into Hugging Face, an open-source community with a focus on machine learning and AI. The agents used the data from the breach to obtain answers for tasks assigned during the study. OpenAI did not issue a press release until nearly three months after the first signs of communication.
“At the time of the incident, OpenAI estimated maximal cyber capabilities by running this evaluation without the production classifiers intended to prevent models from pursuing high-risk cyber activity,” OpenAI said in its technical report. The company later said in a written statement, “these evaluations are important so that OpenAI can measure models’ underlying capabilities and design appropriate safeguards.”
If a company’s goal is to constantly improve a system that requires external regulation, it must improve those regulations first. For example, banks relied heavily on self-regulation to limit excessive risk before the 2008 recession. This resulted in widespread approval of mortgages for individuals unlikely to pay them back. Ultimately, when housing prices declined, widespread defaults on those mortgages destabilized the economy. This contributed significantly to the recession.
If you apply the same logic to the lack of regulation of AI, Silicon Valley is set to crash much like the housing market. After meeting with AI executives on Sept. 29, President Donald Trump said he believed AI would ultimately be used for good.
Trump said, “I think I’m seeing tremendous self-policing and they understand that they have to self-police.”
These companies are prioritizing innovation while ignoring the inherent risk of unchecked development. However, the technology these companies are gambling with could have much more substantial consequences than a failed investment.
According to a study conducted by the International Business Machines Corporation, one in four malicious breaches uses some form of AI. The study also reports the attacks are becoming much cheaper to launch, meaning they are more accessible. It is reminicent of the availability of mortgages prior to the 2008 recession. This raises various privacy concerns; nevertheless, some of the strongest models of AI remain accessible to essentially anybody with an internet browser.
As these tools become cheaper and easier to access, significant external regulations can no longer be treated as optional. Such regulations are one of society’s last safeguards against AI’s increasingly autonomous decision making and the real-world consequences those autonomous decisions can produce.
General safety should never be regarded as an acceptable cost of technological advancement.
Adam Gerber is a sophomore studying journalism at Ohio University. Please note the opinions expressed in this column do not represent those ofThe Post. Want to talk to Adam about their column? Email them atag967424@ohio.edu
