Compute is to thought as the printing press was to speech.

Technology’s advance expands our fundamental rights and gives us new ones.

And no right is so fundamental as the freedom of thought.

“Without Freedom of Thought, there can be no such Thing as Wisdom; and no such Thing as public Liberty, without Freedom of Speech."

Benjamin Franklin published this statement in 1722, as a teenager working as a printer and journalist, nearly 70 years before the US Bill of Rights was ratified. His words are now displayed on the wall in the US Capitol Building. Freedom of speech has always been recognized as a threat to tyrants, and a right worth protecting. (The Soviet Union’s tight lock on publishing and copy machines produced one of English’s few Russian borrowings: samizdat, or self-published material.) Freedom of thought was too obvious to be included in the US Bill of Rights, but by the middle of the 20th century, it was under attack from violent collectivist ideologies. In 1949 George Orwell’s dystopian novel 1984 reflected the dark trend, giving us terms like wrongthink, thought crime, and the Thought Police. Already in 1948, Article 18 of the Universal Declaration of Human Rights declared:

“Everyone has the right to freedom of thought, conscience and religion.”

This statement was reiterated in Article 10 of the EU Charter of Fundamental Rights and Article 9 of the European Convention on Human Rights.

Our right to compute is being threatened now.

Several actors, including the current biggest AI companies, other Big Tech companies, AI safety activists and organizations, and government organizations, are trying to regulate, restrict, and even criminalize, current AI models and the computing power used to create them. These actors have different underlying motives, some well-intentioned, some seeking commercial advantage and regulatory capture, and some seeking power, but they share an ideology of paternalism. Because AI could be used for nefarious purposes, they say access should be restricted to a limited trusted circle, so as to protect the rest of us. We consider this not merely inequitable, but dangerous. Parties with such advantage are bound to abuse it, and ordinary people would not be able to compete or defend themselves.

Both the United States and the European Union have already begun regulating computing power. President Biden’s Executive Order 14110 regulates AI in numerous ways based on several criteria. One of these criteria is the computing power used to train an AI. EO 14110 sets three thresholds: 10²⁶ FLOPs for any AI model, 10²³ FLOPs for an AI model trained primarily on biological sequence data, and 10²⁰ FLOPS for any computing cluster. The EU’s AI Act sets a single threshold of 10²⁵ FLOPs. In both the US and EU, these compute limits can be changed easily, with bureaucratic rather than legislative action. This is only the beginning.

There are other examples, such as limitations on the usage of encryption. Many major countries restrict use of encryption, such as Singapore, China, India, and Russia. The right to compute is here intertwined with the right to privacy.

If an individual’s access to modern technology is limited, so are their rights. Imagine the absurdity of recognizing the freedom of speech, but only for one’s own unaided voice, denying the rights to publish or use telephones. Imagine the weakness and vulnerability of individuals denied their rights of speech and thought, at the mercy of those “trustworthy” enough to be granted such power. Imagine trying to work productively in the Information Age, falling increasingly behind those with access to compute.

The UDHR, the European Convention on Human Rights, and the Charter of Fundamental Rights of the European Union all state that “Everyone has the right to freedom of thought, conscience and religion”.