I Think We Have Removed Too Much Art From Technology

I have been thinking about the separation between art and technology because I am not convinced the separation makes very much sense.

We speak about technical intelligence and artistic intelligence as though they belong to different kinds of people.

I do not experience them that way.

To design something well requires perception. You have to recognize proportion, hierarchy, unnecessary complexity and the relationship between individual parts and the whole. These are artistic questions, but they are also questions I encounter constantly in technology.

I think we have narrowed the meaning of design until it frequently means appearance.

The architecture is completed first.

The system is built.

Then somebody is asked to make it beautiful.

But beauty is not something that can always be applied at the end.

A beautifully designed system may be beautiful precisely because somebody understood what should never have been added to it in the first place.

That is a different kind of design.

I have always been interested in the intersection between the liberal arts and technology because I think each discipline corrects something in the other.

Engineering asks whether something can work.

Design asks how it should work.

Language asks whether we have described the problem accurately enough to know what we are solving.

Art asks us to notice relationships that are difficult to quantify but immediately perceptible when they are wrong.

These questions become particularly interesting with artificial intelligence because language itself has become part of the technical architecture.

An LLM is computational, but the thing through which we encounter its intelligence is language.

That means vocabulary matters.

Context matters.

Ambiguity matters.

The distinction between two apparently similar words can matter.

What the model has been allowed to know matters.

What the human assumes the model knows matters even more.

At North Shore, our experiments with local language models have made me increasingly interested in this problem.

It is relatively easy to make a model produce an answer.

It is much harder to construct the conditions under which the answer deserves to be trusted.

What information did it receive?

What information did it retrieve?

What distinctions existed in the source material?

Which distinctions disappeared when that information entered the model?

When is the model retrieving knowledge and when is it constructing an interpretation from incomplete evidence?

These are engineering questions.

They are also linguistic questions.

And, in a strange way, I think they are artistic questions because they require discernment.

Technology has become extremely good at addition.

We can add another application, another integration, another model, another automation and another layer of abstraction.

I am more interested in whether we understand the system well enough to know what does not belong there.

This is where I think art has something important to return to technology.

Not decoration.

Discernment.

Steve Jobs famously spoke about Apple existing at the intersection of technology and the liberal arts. I think that intersection has become more important rather than less important in the age of artificial intelligence.

The more capable our technology becomes, the more consequential our assumptions about people, language and meaning become with it.

A system can be technically impressive and intellectually careless.

It can also be technically sophisticated and aesthetically incoherent.

Neither failure is inevitable.

At North Shore, we are interested in building technology as though all of these things belong to the same intellectual problem.

Because I think they do.

Technology is made by human beings, interpreted by human beings and ultimately exists to change something about human experience.

There should be more art in that.

Previous
Previous

When Employees Become the Architecture