Study measuring IQ of various AI puts Google's at 47.28

Originally published at:

Im fascinated by how AI is marching forward and yet consciousness seems to be becoming irrelevant in its development.
As such we may pass the singularity , of a self sufficient and self evolving computer, before what we even understand as consciousness is even simulated.


Is it? Seems to me it’s pretty much 100% hype, and has been since HAL-9000.

I’m wondering about methodology. I took an IQ once, and as I recall a booklet was pretty much just plopped in front of me. Try that with Google, I don’t think much is going to happen. In fact, if you type in the instructions I still don’t think anything will happen.

However, if you define IQ as finding relevant web pages really fast…


So you think AI should start off as smart as you?
Its getting better, much better and as the article suggests its still only 47.28 IQ at best.
Get some perspective and maybe you could be smart too.

See Peter Watts’ novel Blindsight for a description of an encounter with intelligent but non-conscious aliens.

1 Like

The question is, I think, what this is supposed to mean? It obviously isn’t the same test that I or @Boundegar took. I skimmed the article and they don’t offer any clues other than to reference a 2014 article. I suspect that their measure doesn’t correspond to whatever is tested in human IQ exams.

Edit: Ok, here is the paper where they (kind of) define their IQ measure:

It doesn’t look like a human IQ test to me, but the last time I took one was when I was ~8 years old.


We will be able one day to build a strong general AI.
I’m not sure we will ever be able to understand consciousness.

1 Like

what is it that we understand as consciousness? Never mind the machine that’s obviously just a machine - how do you know that I have a consciousness that you understand? Or that you have one, that you understand?


I agree that what they’re measuring isn’t IQ in the sense that we measure it in humans. Even if it were, the range they’re talking about is 3.5-5 standard deviations below the human average, and there’s no way they have enough data to think their test is valid in that regime. I’m not sure even human IQ tests are known to be valid that far below 100.

But assuming they still think it’s an interesting and useful result anyway… why on Earth would they report four significant figures in the score? AFAICT they didn’t report error bounds, either. To me that suggests the authors don’t understand statistics all that well, so for now I’m filing this under “about as reliable as the odds Spock gives for whether the Enterprise will survive.”


So, digital assistants are… special…

If you don’t think someone with an IQ of 47.28 can be a threat to Humanity, you haven’t been paying attention to the news recently.


Rorschach might just have been badly equipped to understand Sascha’s topic of conversation. You can be self aware but totally dumb about some things, and smart about others. I would be willing to place a bet that google is self aware now, and that we could never tell by interacting with it.

My symbiotes don’t know that I am self aware (assuming that I am). I am just a source of food to them.

1 Like

an IQ of 47 is a threat to humanity. People will come out of the woodwork to be inhuman, to put down such a being. That’s the threat to homo sapiens.

This topic was automatically closed after 5 days. New replies are no longer allowed.