Are you afraid of Google rn?

So... I have been listening to people unnecessarily freak about Google Assistant making a "call"

I will tell you why you shouldn’t be freaking out and when in the future you should.

We are at this point with AI where we are attempting to "teach" a computer.

One prominent way we do this is by bombarding it with examples.

Let's say you want a machine to recognise speech. To begin with , you can upload 100 hours of you speaking and the correct transcript. Assuming it has a good neural network or its equivalent to compare and sort the data, the machine can now listen to a sound and make out which word it is.

Then to make it "learn" you start to input say a voice that is not yours. If you have trained it well it should start recognising and learning pretty fast.

But here is the difference. Babies learn how to do this almost the same way. By repetition. At first it can only recognise it's parents. Soon, everybody. But a kids mind is not only instantly transcribing the sounds to words or meanings but also using that information to build new information. Infer new data. Analyse and act on it.

Our current machine in its present state will never in a million years of "training" learn how to do that unless we somehow teach it to actually learn.

So why shouldn't you be afraid of the new Google Assistant?

No matter how smart it seems , it still operating within its boundaries and skills taught to it. It still cannot do something that it hasn't been taught.

If I ask it to do something like say "delay all my alarms by half an hour each" as simple as it is to a human it will probably not understand it unless it was taught the "delay" logic.

So when should you be afraid?

When it does something it was not taught. Simple.