Timnit Gebru

Merely expanding large language models is far from sufficient to achieve Artificial General Intelligence (AGI). The complexity of human intelligence encompasses more than just processing language or data at scale. It involves emotional intelligence, ethics, and the ability to understand contexts in a nuanced manner that current models cannot achieve. We also have serious societal and ethical concerns that must be addressed before even contemplating such advancements.

As someone deeply involved in the ethics of AI, I strongly believe in the necessity of a global institute for AI safety. Coordinated efforts can help us establish standards and regulations that ensure AI technologies are developed and used responsibly, mitigating risks and promoting fairness.

I strongly disagree with the idea that the first lab to build Artificial Superintelligence should prevent others from developing competing ASIs. Concentrating such immense power in one entity poses a significant risk to ethical standards, diversity of thought, and global governance. A pluralistic approach to ASI development encourages collaboration, transparency, and accountability, which is vital for ensuring these technologies benefit all of humanity.

I strongly believe that global AI governance frameworks must prioritize equitable access to AI technologies for all countries. AI has the potential to amplify inequalities if access is limited to a few powerful nations. Ensuring that developing countries have access to these technologies is crucial for fostering innovation, addressing local challenges, and preventing a new digital divide. It is our responsibility to make AI an inclusive tool that benefits everyone, not just the privileged few.

I believe nationalizing AI firms is essential to prevent them from accumulating too much power and influence. Without oversight, these companies can easily prioritize profits over public welfare and ethical considerations, leading to potentially harmful consequences. By nationalizing them, we can ensure that AI development aligns with societal values and serves the public interest.

In my view, large language models don't truly understand the content they process. They mimic patterns in data but lack real comprehension or consciousness. This limitation must be considered in developing responsible AI.