Stop Insulting My Artificial Intelligence, Please

A lot of companies are using the term AI to get funding and attention without really using artificial intelligence technologies, writes the co-founder of AI-based market analysis company Zirra

Aner Ravon 08:1209.11.18

TAGS:

It seems like every company is doing something with Artificial Intelligence these days, right?

Wrong.

A lot of companies are using the term AI to get funding and attention without really using artificial intelligence technologies.

For daily updates, subscribe to our newsletter by clicking here.

Seriously though, AI is great. But there is little academic news about contemporary AI. The main algorithms in use today have been developed in the 60s, 70s, and 80s at the latest.

Aner Ravon. Photo: Amit Sha'al

What’s news is the available computing power. In other words, an iPhone can perform AI tasks that required a room-sized machine 30 years ago. That’s big. Really.

This means developers and entrepreneurs face a lot of great opportunities to innovate and make money. It also means that the days of Skynet are not nigh, the end of the world is not approaching, and that weapons in the hands of politicians pose a much bigger threat than AI reaching the point of singularity.

AI is still very task driven and is at early stages of getting anywhere beyond. Therefore, not only is traditional data science very much alive and kicking, but billions of dollars of venture capital money are being washed down the drain as we speak.

Oh, and in the meantime, we are victimized by annoying chatter from people with an urge to sound intelligent. This, perhaps, is the worst.

So when should AI be used, when shouldn’t it, and what are its shortcomings? AI’s main strength has to do with reading, finding and using patterns. The “Intelligence” in artificial intelligence refers to its ability to mimic human thinking patterns and most importantly retrain itself and improve based on feedback.

AI, when properly used, can help find indicators for further research better than humans. But it cannot conclude very easily, because reading patterns involve another science, that of statistics and probability. In simple words, an AI machine will think very fast, go through millions, billions, even trillions of details very quickly, come up with pattern matching, even reach a probable conclusion faster than any human and the conclusion will most of the time be right.

But many times, the conclusion will be wrong. And this is a critical drawback.

In such cases, the AI engine will need to be retrained. Rigorously. This can be expensive and everlasting, but it will increase the probability of getting things right the next time.

Even so, no matter how much we train the machine, many times the conclusion will still be wrong.

Therefore, while AI can provide tremendous diagnostic help, it cannot replace doctors, their decisions, and their care-taking.

AI can help a driver avoid countless dangerous situations and save a lot of lives. It cannot drive a car. Oh, and it won’t be able to drive a car for many years to come. Your kids may need to take their cars in the air to avoid traffic, but they will still require a driving license. Your grandkids, most probably, too.