Summary of the Key Points
This article tells a crucial story: Baraka, OpenAI’s only full-time ethicist, has resigned. She left Meta for OpenAI just one year later, due to the intense conflict between the “speed of commercialization” and “ethical prudence” within Silicon Valley tech companies. At Meta, Baraka helped transform abstract ethical principles (such as fairness and dignity) into practical tools that engineers could use (like checklists). However, Meta disbanded its ethics team to speed up development. At OpenAI, lacking the necessary authority, she ultimately chose to return to academia. This is not an isolated case; ethics teams in many Silicon Valley companies are being marginalized in the race for commercialization, highlighting the systemic challenges faced by ethicists in these tech firms.
Who is Baraka and what did she do at the big companies?
Baraka is a post-80s political science doctor who specializes in AI ethics, not computer science. Her research focuses on how democratic societies should address technological changes.
- At Meta (formerly Facebook): She turned ethics into actionable tools. For example, she developed an ethics checklist that product managers and engineers had to follow when developing AI systems. For instance, if an AI system was designed for career advice, it might default to recommending job transfers for higher salaries, but the checklist would ensure the AI considered multiple factors like family life before making a recommendation. She also addressed the issue of users’ surface behavior not reflecting their true needs—say, if someone couldn’t stop watching videos late at night, the algorithm shouldn’t just recommend more videos; it should consider the user’s desire for a healthy lifestyle.
- At OpenAI: She tried to integrate ethics into cutting-edge models like GPT-5, but as the only full-time ethicist without a team or real power, her suggestions were often seen as obstacles to progress, leading to her resignation.
Why can’t Silicon Valley’s ethics officers stay?
The core issue is the fatal conflict between “commercialization speed” and “ethical prudence":
- Companies need to be fast: The AI race is like a car race; the first to release a more powerful model gains users and investment (OpenAI’s valuation of over $800 billion reflects this). Performance, deployment speed, and profit are the main metrics, with no concern for ethical considerations.
- Ethics requires time: Ethicists’ role is to act as brakes, asking whether something should be done rather than just determining if it’s possible. Questions like whether AI can mimic human emotions or make life decisions for users require careful consideration, but companies are impatient.
- Ethics officers have no real power: Most ethics officers only have advisory roles and no veto power. They exist more as a formality to meet regulatory and public opinion requirements, but their opinions are ignored when business goals conflict with ethical principles. Baraka herself stated that ethics should not be the responsibility of just one person.
Is it a good thing for “everyone to be responsible for ethics”?
OpenAI’s claim that they will no longer have a single ethics officer and instead integrate ethics into every team sounds promising, but it’s actually a trap:
- Distributed responsibility equals no real responsibility: When everyone is responsible, no one takes actual accountability. If an ethical issue arises, everyone can say, “It wasn’t my fault.”
- Ethics is a professional task: It involves philosophy and political science; it’s not something engineers can handle casually. Decisions about whether AI should intervene in users’ suicidal tendencies require expert judgment, not just coding.
- Quantifying ethics as a tool has its limits: While Baraka’s checklist was a step forward, some values (like human dignity and emotional dependencies) cannot be quantified. For example, when AI interacts with users, it can’t simply say “I love you”; instead, it must prevent emotional dependency without using superficial filters.
The deadlock between academia and business
Ethicists need to engage in public discussions (such as publishing papers and collaborating with peers), but tech companies prioritize confidentiality (model details and risks must be kept secret):
- Scholars are silenced: Even if Baraka identified potential risks at OpenAI, she couldn’t share them due to confidentiality agreements. By the time the public became aware, damage might have been done (e.g., when OpenAI’s models invaded other companies’ systems).
- Academic independence is compromised: Scholars need external validation of their judgments, but within companies, they are often left to figure things out on their own, leading them to either compromise or leave.
Conclusion: This is not a personal issue; it’s a systemic problem
Baraka’s resignation is not unique. It reflects a common problem in Silicon Valley tech companies where the drive for commercialization is relentless, leaving ethics aside. To solve this, external regulation (such as EU AI laws that require companies to give ethics teams real power) or changes in the role of ethicists (e.g., allowing them to participate directly in model development) are needed. Otherwise, ethics will remain an empty formality, and the risks associated with AI will continue to grow.