OpenAI’s Ad Testing Sparks Skepticism from Former Researcher
This week has been significant for OpenAI as it begins testing advertisements on ChatGPT. However, it’s also a time of change for one of its researchers, who has decided to resign after two years with the company. This individual played a role in shaping the development and pricing of AI models and contributed to early safety policies before standards became more established.
Initially, they believed their work could help guide the builders of AI in addressing potential issues. However, recent events have made them question whether OpenAI is still prioritizing the critical questions they set out to explore.
While the researcher does not view ads as inherently wrong or unethical — acknowledging that running AI can be costly and ads may provide a needed revenue stream — they express concerns about OpenAI’s direction.
For years, users of ChatGPT have shared personal and sensitive thoughts, believing they were engaging with a neutral entity. This openness has created a unique archive of human emotion and vulnerability. Now, the introduction of ads could lead to potential manipulation, a risk that is difficult to comprehend and manage.
Many see the funding of AI as a dilemma between two bad choices: limiting access to those who can afford it or using ads that exploit users’ fears and desires. However, the researcher argues this is a false dilemma. Tech companies can explore alternatives that ensure broad access without compromising user privacy or ethics.
OpenAI has stated that ads on ChatGPT will be identifiable and won’t influence responses. The researcher believes this initial phase might adhere to those promises. However, they worry that future developments might undermine these rules due to the pressures of creating a profitable model.
Recalling Facebook’s early promises to safeguard user data and empower users, the researcher points out the gradual erosion of those commitments, often driven by a pursuit of advertising revenue. They worry that OpenAI might be headed down a similar path, prioritizing engagement in ways that could lead users to develop unhealthy dependencies on AI.
There is also a growing concern about subscription costs for top-tier AI services, making them less accessible. The focus, then, should not merely be on whether to have ads or not, but on creating frameworks that allow everyone to benefit from these tools without falling victim to manipulation.
One potential solution is to implement cross-subsidies, where profits from one service support access for others. For example, businesses utilizing AI for high-value tasks could help subsidize access for individuals or small businesses.
Another approach could involve regulated advertising that includes a system of governance. Companies could be required to adhere to strict oversight regarding how personal data is utilized.
Lastly, an independent system could grant users control of their own data, similar to the Swiss cooperative MIDATA, which allows members to manage their health data securely.
While none of these solutions are simple, there remains a window of opportunity to address them, preventing a future where technology manipulates users or is accessible only to the wealthy.
With thoughtful solutions, the AI industry can navigate these challenges and create a fairer technology landscape for everyone.
