OpenAI and the Dangers of Monopoly

On November 30, 2022 OpenAI CEO Sam Altman stepped into an office in Silicon Valley and dropped what would become one of the most significant products of this decade in ChatGPT. Functionally, ChatGPT is not terribly different than Alexa or Siri, but its commercial appeal is undeniable and launched Altman into the ranks of Silicon Valley royalty like Steve Jobs or Elon Musk. This would come to be one of the more significant non-technical effects of ChatGPT from a policy perspective as Altman began appearing with foreign Heads of State and in the halls of the US Congress, assumedly to provide testimony on the details and future outcomes of the technology. This developing relationship between Altman and policymakers around the world likely creates a complicated regulatory environment in which one of the most direct stakeholders in the success of the technology is heavily involved in the information dissemination about that same technology, information without which regulators cannot effectively do their job. The complexity and scale of the technology allows Altman and other scions of this new technological landscape to hold Washington in tension between regulator’s hunger for information about this technology and the captains of this new industry being the ones steering the ship.

AI is fiendishly complex. Large-scale AI projects like the ones that dominate Silicon Valley start with mountains of data. How these companies choose their datasets, the methods they use to prepare that data for consumption by the algorithm, these are the kinds of questions that can have huge downstream effects in the process that go undetected for years. It was this past October, a full year after the launch of ChatGPT, that researchers discovered the chatbot was providing wrong and misleading medical information based on patients racial background(1). As we start to rely more on these artificial intelligence systems in our work, these issues of bias can be incredibly impactful and have real cost. This is the benevolent case, where those involved have good intentions and the issue is in the implementation. If an actor has less than the best interests of society at heart, these systems are full of potential to make mischief. From the viral spread of deepfaked content to automated spambots scamming social media users, as these systems become more sophisticated so too do the methods used by cyber criminals to bring harm and sow chaos.

The need to balance the malevolent sophistication of threat actors against a desire to maintain as free and open an internet as possible is what gives Altman and Co the power to sit with Heads of State around the world. It is an awesome responsibility, but one that brings with it a risk of reviving the robber barons that loomed large a century ago. Men like J.P. Morgan and John D Rockefeller would feel quite familiar with the situation in which Altman finds himself . An emerging technology which creates both incredible progress and new threats, business leaders roaming the halls of power, a government bureaucracy that is bloated and inefficient. These are the problems regulators must contend with if they are to properly address the concerns being raised by many in the field over the proliferation of this technology. These challenges are nothing new, and exist now mostly as an extension of the challenges regulators already faced in the digital age.? The need to provide privacy protection, regulate access to key hardware such as semiconductors, implement worker training and third-party testing, are ubiquitous in this space. Traditionally, regulation only comes after a crisis which spurs political outrage over industry standards. That approach will not work when it comes to AI regulation. It is important that our political leaders not only understand AI, but understand the gravity of the work they take on.?

The Biden Administration in its Executive Order on the Safe, Secure, and Trustworthy Development of Artificial Intelligence released on Oct 30 2023 demonstrated that? the White House, at least for the moment, intimately understands the magnitude and importance of the task before them(2). Even prior to October 2023 the Biden Administration demonstrated its willingness to go to bat on issues of technology through its landmark legislative achievement CHIPS Act. While these steps are excellent, they are nowhere near sufficient. Top-down legislative measures must be met with equally vigorous work by a broad base of regulators and stakeholders. Sam Altman himself has suggested that failure to properly address these issues is a threat on par with climate change or nuclear armaments. Whether the measures taken by Washington, under the advice of Altman, translates to improvements in outcomes will be the real test. Our role as policymakers is to make sure we understand the challenges this technology poses and the points of vulnerability where our policy can serve to protect the end users from the threats enabled by artificial intelligence.

Sources

1. https://fortune.com/well/2023/10/20/chatgpt-google-bard-ai-chatbots-medical-racism-black-patients-health-care/
2. https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/