What is the AI effect, and is it set to happen again?

With all the hype surrounding modern artificial intelligence, it’s easy to forget that AI isn’t actually a new concept. Artificial intelligence started to make a name for itself as far back as the 1950s.

Indeed, we’ve had more than one iteration of AI so far. The technology we’re seeing today — while new and exciting — is not the first to hold the label of ‘artificial intelligence’.

This is due to a phenomenon known as the AI effect.

But if all the past iterations of AI have since been re-named or forgotten, will the same thing happen to the artificial intelligence of today?

The AI effect

The AI effect refers to a phenomenon in which technology once considered artificial intelligence loses its AI label. That is, it’s no longer viewed as AI, by virtue of not being ‘real’ intelligence.

In other words, the AI effect is when people stop viewing an AI tech tool as valid artificial intelligence. It usually happens once the technology becomes a widely common part of daily life.

For example, a machine displays seemingly intelligent behaviour that it couldn’t have done before. It’s labelled as AI.

But, after a while the way that the machine completes that task becomes better understood by the majority. Suddenly, that ability doesn’t count as AI, it’s just another computation.

The AI of yesteryear

In the 60s and 70s, AI saw its first hype cycle. AI promised automatic translation abilities. It powered early chatbot ELIZA. And it explored rule-based systems known as ‘reasoning as search’.

Today, translation isn’t often equated with AI, instead settling as its own tool. ELIZA, though an important part of AI and chatbot history, never understood what she was saying. And the algorithm of ‘reasoning as search’ required too much data and code paths to be workable. 

Come the 80s, and expert systems were the new form of AI. Expert systems were another rule-based system that focused on being good at one thing. Essentially, these systems form narrow AI, also known as weak AI. Which now, some argue, isn’t good enough to count as genuine intelligence.

Each time, AI would fall short. We’d descend into an AI winter, and the AI abilities of our tech adopted new names or hid in the background of trusted tools.

Why does the AI effect happen?

Early on, artificial intelligence earned a stigma following the first AI winter. During this time, calling your research or tech ‘AI’ meant limited or no funding.

As a result, technology that might have otherwise held the AI label instead hid behind other fronts. (For instance, ‘expert systems’ rather than ‘AI’.)

The AI effect could also relate to the perception of AI. That is, when the ability of a piece of tech becomes commonplace, it stops being AI.

Because people are so used to technology completing a given task, it loses its magic. There’s no longer the sense that it’s an amazing computational ability. It’s not exciting AI like in the science fiction stories. It’s boring.

Perhaps the biggest cause of the AI effect is the vastness of what ‘counts’ as AI-led technology or real intelligence. Each advancement gets a separate name and breaks away from the AI label. Despite, that is, stemming from AI research.

Defining AI – a moving goalpost

AI is a loose term — it’s more of a catch-all label than it is a clear name for a single specific technological ability. Rather, there are multiple subsets of what we today consider artificial intelligence.

For instance, ‘AI’ could refer to general or specific machine intelligence. It could describe machine learning or natural language processing. It could mean any program that makes a decision for you. The list goes on.

This means that identifying what ‘AI’ pertains to is often a difficult task.

Everyone has a different idea about what ‘counts’ as artificial intelligence. Specifically, no one can agree on the level of intelligence an AI tool needs to earn the right to the name ‘artificial intelligence’. As a result, the goalpost of what makes something ‘AI’ keeps moving.

Why the AI effect won’t happen again

Determining whether the AI effect is set to happen again means looking at the differences between today’s AI and the AI of yesteryear.

A marked difference between then and now is that, in general, there’s a very different attitude to AI. Far from hiding AI labels, businesses are showing off their AI technology. It seems less likely that we will remove the now-coveted AI label from our tech.

Additionally, the artificial intelligence we’re advancing is on a different level than ever before. We aren’t just having computers complete tasks for us, we’re making them learn like us, talk like us, and see the world the way we do. That is, we’re not doing the ‘easy stuff’ anymore.

The flipside of the coin

However, we still don’t have a solid, single definition of what AI is. There’s still disagreement about what counts as intelligence. It’s still possible that even if we accept that our tools come from AI research, they’ll lose their AI labels.

For example, facial recognition AI will simply become ‘facial recognition’. AI chatbots will simply be ‘chatbots.’ And so on. The AI effect, then, could happen again due to the term not relating to any one tech tool.

There’s some question as to whether the AI effect will ever stop. Perhaps it will keep happening until we achieve artificial general intelligence. I.e. the kind of AI that we see in science fiction movies bringing about the apocalypse.

Or perhaps it will keep happening even after that, until the term ‘AI’ is a forgotten relic of the past.

Future uncertainty

AI, as an idea, has developed into an ongoing goal. It’s come to refer to the next round of technological innovation in the realms of machine understanding.

We’re in the height of an AI hype period, one which shows no signs of slowing. It’s working and fulfilling promises where it let us down before.

That doesn’t mean the AI effect is history. Only time will tell if the AI label will stick.

Further reading