Why Derrida is being read again in the age of artificial intelligence

The argument about artificial intelligence circles a small set of questions. Can machines think? Can they overtake human intelligence? Do language models genuinely understand anything?

The thought of Jacques Derrida steps back from all of these and asks something more elementary:

If a machine can write, produce answers and work over the linguistic traces left behind by human beings, then where do the things we call "author", "subject", "memory", "meaning" and "responsibility" begin, and where do they end?

Derrida died in 2004 and never wrote a line about artificial intelligence. But the concepts he built around writing, technics, the archive and memory are being read again, because they show why the present argument is not merely a technological matter but a philosophical one.

Technology is never a neutral instrument#

In Derrida's thought, technology cannot be treated as a neutral implement lying ready in the human hand.

A communications technology does not simply carry existing thoughts along a new channel. It also changes what can be said, what can be kept and what can be remembered.

The strongest statement of that claim is Archive Fever (Mal d'archive). There Derrida refuses to see the archive as a passive store in which the past is kept safe. The technical structure of the archive determines the structure of whatever will enter it.

The book was written in 1994, as electronic mail was first spreading. Derrida argued there that email was not merely a faster letter; it redrew the boundary between the public and the private.

The archive no longer merely keeps#

That thought has an immediate use in thinking about large language models.

These systems work over a vast linguistic archive: books, articles, texts scraped from the internet, code, news reports. The corpus is turned into a dataset, and algorithmic operations produce new texts from it.

The critical question is no longer "What does the artificial intelligence know?" The question is this:

What processes of archiving and selection have produced the thing we take the machine to know?

A traditional archive puts the past on record. A digital archive also classifies it, ranks it and makes it visible. Artificial intelligence goes one step further: out of the traces held in the archive it generates new sentences, new images, new possibilities.

The line between archive and production blurs. Derrida's claim finds its contemporary form exactly here: technology does not merely preserve the past, it also settles the form in which the past will present itself to us.

What are we saying when we say that a machine writes?#

Derrida's account of writing opens another door onto the same argument.

He showed that Western philosophy had long treated speech as more original, more alive and closer to the subject than writing, and he put that hierarchy in question. His notion of "writing" is not confined to words set down on paper: it concerns the circulation of traces, the fact that they can be repeated, and the impossibility of tying meaning to a single source.

Seen from there, a text produced by a machine raises an uncomfortable question:

Must there be a conscious subject behind a text for that text to mean anything?

When a language model produces a poem, there may be nothing there resembling human intention. The text can nevertheless be read, interpreted and made sense of.

In Derridean terms this is a productive paradox: the meaning of a text can enter circulation quite independently of any intention that produced it.

Artificial intelligence therefore raises not only the question of whether machines can write as humans do, but the harder one: is a human subject really required in order to write?

The real problem is not intelligence but responsibility#

The second concept that makes Derrida pertinent here is responsibility.

When a system produces false information, who is answerable? The engineer who wrote the algorithm, the company that developed the model, the institutions that supplied the data, the journalist or the academic who used the output? Or is a new domain taking shape in which responsibility cannot be assumed in full by anyone at all?

A substantial part of Derrida's ethics and political philosophy rests on the thought that a decision cannot be reduced to a rule. A genuine decision is not the mechanical application of a procedure fixed in advance; it is taken where the rule runs out. Where the rule suffices, what we have is not a decision but a calculation.

The distinction that follows is sharp. An algorithm can compute the options. But what is calculable and a decision that carries responsibility are not the same thing.

The argument over authorship by artificial intelligence that we reported this month turned on precisely this knot: a journal's grounds for banning machine-authored articles had nothing to do with the quality of the prose, and everything to do with there being someone behind it who bears responsibility.

When the other turns into data#

In Derrida's thought, ethics is not built inside the world of the self. The encounter with the other, and the obligation owed to the other, stand at the centre.

Systems of artificial intelligence, meanwhile, convert people into data, into profiles, into categories and statistical patterns.

A person is no longer merely a name: they are a pattern of consumer behaviour, a credit risk, an image of a face, a health profile, a category of political leaning, a recruitment score.

Derrida's framework issues a warning here: a human being is not the sum of those properties about which calculations can be made.

The deepest philosophical problem of artificial intelligence may not be how well it imitates the human. It may be the extent to which it turns the human into a calculable object.

The trace and the right to be forgotten#

Derrida's concept of the trace takes on a new sense as well.

Every trace left on the internet — messages, searches, photographs, purchases, location data — becomes part of steadily swelling datasets.

His thinking about the archive is a reminder that to record something is not to store it away. To record is also to create the conditions that will determine how it is read in the future.

The "right to be forgotten" is therefore not only a legal matter. It is an ethical question too: can a person who is never deleted from the memory of a machine be said to have been forgotten at all?

Would Derrida have opposed the technology?#

It would be easy to answer yes or no, but his thought does not permit that sort of pairing.

Derrida cannot be read as a philosopher opposed to technology. On the contrary, he thought seriously and early about how electronic communication and the new media technologies were transforming human experience.

A Derridean approach would be neither blind admiration nor romantic refusal. The question it asks is a different one: how is technology remaking the human world?

Unsettling the centre#

At the core of Derrida's philosophy lies a criticism of any attempt to fix meaning to a single unchanging centre. Deconstruction is an approach that questions the oppositions and the foundations which texts and institutions take for granted.

Artificial intelligence carries that problem into daily life.

Who wrote this text? Who produced this image? Where did this idea come from? Who owns this decision?

As the boundary between human and machine blurs, the concepts on which modern thought was built — author, subject, originality, intention and property — are reopened for argument.

In closing#

One of the standing errors of the technology world is the belief that every question can be settled with better models and more data.

Yet some questions are not technical but philosophical. What is it to understand? What is it to remember? What is it to be a subject? What is it to be responsible for a decision?

Derrida is being read today not because he supplies ready answers to these questions. His importance lies in making visible the assumptions that lie beneath them.

Perhaps his legacy comes down to a single suggestion: before considering the answers technology gives us, look at the questions technology prevents us from asking.

For the great philosophical problem of the age of artificial intelligence may not be that machines have begun to think like people. The real problem may be that people have begun to redefine their own thinking to match the way machines work.