Will AI make blockchains conscious?

This is a segment from the Blockworks Daily newsletter. To read full editions, subscribe.

“Consciousness may end up being found in very strange places.”

— Christof Koch

The canonical question in the philosophy of consciousness was posed by Thomas Nagel in 1974: “What is it like to be a bat?”

Nagel’s idea was that consciousness is defined simply by what it feels like to be something — the inner, subjective experience of being alive and aware.

“An organism has conscious mental states if and only if there is something that it is like to be that organism,” he explained.

Many have found this subjective answer unsatisfyingly circular: What is this something???

David Chalmers later declared this question “the hard problem of consciousness” because it exposed a gap between subjective experience and objective science.

In 2004, however, Giulio Tononi confronted Chalmer’s hard problem with a paper proposing a mathematical model for consciousness: Integrated Information Theory (IIT).

Consciousness, he argued, is a mathematical property of physical systems — something that can be quantified and measured.

But can a system be conscious?

After interviewing the computational neuroscientist Christof Koch, the co-hosts of the New Scientist podcast concluded that computers, being systems, could theoretically achieve consciousness if they were able to “integrate” the information they process.

And nearly anything could be a system: Even a rock might register a trace of consciousness if its atoms form the right kind of structure (as proven in the science documentary Everything Everywhere All at Once).

Newsletter

Subscribe to The Breakdown

Which got me thinking: Ethereum is a world computer, right?

And critics accuse Bitcoin of being a pet rock.

So…if computers and rocks can be conscious, surely blockchains can be, too?

Blockchains do, in fact, tick a lot of IIT’s boxes.

IIT posits, for example, that a system can only be conscious if its current state reflects everything it’s been through — just as your memories shape who you are and each moment builds on the last.

Blockchains like Ethereum work in a similar way: A blockchain’s current “state” is a function of its history and each new block depends entirely on the ones before it.

That history-dependence gives it a kind of memory — and because thousands of nodes agree on a single shared version of reality, it also creates a unified “now” (or “state”) that IIT says is a characteristic of consciousness.

Unfortunately, IIT also says that for a system to be conscious, it has to have “causal autonomy” — which is to say, its parts have to influence each other internally and not just in response to inputs it passively receives from external actors.

Blockchains don’t work like that, of course.

Instead, they rely on external inputs (like users sending transactions and validators adding blocks) to act on and advance — and the nodes that run the network don’t influence each other internally, they just blindly follow the same set of rules.

There’s no spontaneous activity, no internal causation — not even the aimless vibration of molecules you’d get in an inanimate chunk of granite.

So I’m sorry to report that, on the IIT spectrum of consciousness, blockchains rank below even rocks — and that the “pet rock” jab might therefore be a compliment to Bitcoin (or an insult to rocks).

But maybe not for long!

In 2021, the computer scientists (and married couple) Lenore and Manuel Blum co-authored a paper describing how to engineer consciousness into machines.

Their framework treats consciousness as a computable property — achievable with AI algorithms designed to produce systems with the “causal autonomy” required for conscious experience.

The AI wouldn’t be conscious itself, in this case, but a system that deploys it could be.

Now imagine an AI-enabled blockchain that doesn’t just run code, but thinks about running code.

Instead of inert ledgers passively waiting for inputs, blockchains could be self-contained, “causally integrated” machines — more like synthetic brains than distributed databases, with the kind of internal autonomy that IIT researchers consider essential for consciousness.

This could be useful!

A system like that might be able to reason about its own security, detect anomalies in real time, and decide when to fork itself (perhaps after a period of soul-searching introspection).

In short, it would do things not because it was told to, but because it understood what was happening — both inside itself and in the outside world.

It’s not impossible.

Today’s blockchains are more like nervous systems without a brain — wiring without will.

But tomorrow? Who knows.

If IIT is right, philosophers might soon be asking, “What’s it like to be a blockchain?”

(And also, is it better than being a rock?)

Get the news in your inbox. Explore Blockworks newsletters: