What does Google's AI say about YOU? Is it funny, accurate, horrifying?

Miscellaneous and Personal Stuff I Must Share
1

A listserv that I participate in recently had a conversation where someone said (paraphrasing here), “have you tried asking Google AI who you are? It tends to be wildly wrong, in hilarious or disturbing ways. In short, it lies.”

Curious, I asked Google, “Tell me about [my IRL name here.]” When you do that, at least in my configuration (Mac/Safari), the first result will have a list of options at the top; in my browser, it reads: AI Mode, All, Images, Short Videos, and so on. The default is “All” but you can click on “AI Mode” which is what I did for this post.

Interestingly, what AI says about me is almost entirely correct, if a bit puffed up. One key caveat: My IRL name is quite common; I know of at least three people with my name who have had interesting, world-spanning careers - a World Music researcher/dancer, a prominent journalist, and a world-roaming travel photographer. So I can’t just say, “Tell me about CairoCarol.” But as soon as I give it just one reasonable clue - “Tell me about CairoCarol who now lives on the Big Island of Hawai’i,” or other specific facts (“Tell me about CairoCarol who was once employed by GlurpCorp) - it is actually quite accurate (though overly flattering). The only mistake I have found is that it claims I currently live in Hilo; I do not.

To my surprise, the AI summary even added a legit qualification along the lines of, “it is possible to confuse CairoCarol with CairoCarole, whose background has various similarities but who is primarily known as an oboe player in Minnesota.” (I’m making up that oboe/Minnesota part, but you get the idea.)

I’ve done a similar exercise on some real life friends (none of whom have names as common as mine), and have found that the results are pretty accurate.

How about you? Is AI correct? Is it laughably wrong? What do you think about this?

ETA: I recommend this Pit thread, which discusses how harmful AI summaries can be, as an accompaniment to the above exercise. I’m not trying to argue that we should all be placid about AI. I’m just a bit surprised that in my case, it isn’t too off base.

2

This is truly scary.

I didn’t use my real name. I used Beckdawrek. Nailed me.

I’m afraid to use my real name. It’s not the same as my name I use online.

They are both correct but different.

3

That’s sort of scary. Is your use of the name “Beckdawrek” unique to the Straight Dope, or do you use it other places as well?(SHHhhh… don’t answer that! AI is listening.)

4

I don’t have much online presence. Period.

The Dope is it actually (I do go on Giraffe occasionally).

So my whole write up was about the Straight Dope and my posting here.

Like I said: scary.

Kinda happy I don’t do more online.

5

It doesn’t tell me very much about me (under my real name); everything it shares is just from a scrape of my LinkedIn profile.

6

I dislike Google AI and find its “AI summary” schtick to be annoying and often wrong. But I’m not here to threadshit, but to mention that in one of the threads about AI, a number of us (including me) with extensive histories with ChatGPT asked it what it thought of us. This was based not on our internet presence but on the questions we had asked over the years and the ensuing discussions.

For me, anyway, the result was remarkably accurate in terms of my interests and the wide range of things I was curious about, from baking to cosmology, although it was excessively sycophantic. It was also possible to ask it to describe negative traits, or to summarize a personality humourously, and it did that quite well, too. I’d point you to the thread but I don’t remember which one it was, or even what forum it was in. We’ve had quite a few threads on AI-related topics.

ETA: An interesting variation on this was to ask ChatGPT what it thought of any particular poster on the SDMB. It was impressive how quickly it could scan the board and come up with a pretty good summary!

7

About the same, though it made me sound better than my profile does. I asked it about if I had any family and it was quite wrong which surprised me, given all the online public record services. You’d think that it would at least pin my spouse correctly.

8

While I’m not a mod, I don’t consider that to be a threadshit in the least. It’s quite on point for the discussion.

9

I asked it about family, too, and it pretty much said, “his online presence is focused on his professional status” (which isn’t really the case at all). This tells me that, whatever it’s pulling from, it’s not Facebook (not too surprising, given that Meta is their competitor), and it’s not getting anything from public records.

10

This is what it said about me (real names hidden).

The company I work for was bought out by another company several years ago, so the company name is outdated. But this indicates where Google got this info from. I haven’t bothered to update my LinkedIn profile in quite some time, and it still has the old company name on it, so Google obviously stole this from there.

Google stole career stuff from Facebook. I don’t list the actual company I work for on Facebook, but I did list the Straight Dope as a “work” place on Facebook. So that’s kinda funny. :slight_smile:

I actually got my start in the defense industry, working on airborne radar and flir systems for fighter jets. Then I built custom equipment for a medical researcher. I work in process control now, but that’s not where I started. I don’t know why Google thinks I started in process control.

I don’t know why google thinks I am “active in the community”. I don’t think I am.

FWIW, it got all of the names and locations correct.

I can definitely tell where it got its info from based on what it said.

11

It doesn’t know me at all. Even if I give my location.

I guess I haven’t accomplished that much in my life.

12

I tried this with Google AI on “sdmb user spoons.” It was pretty accurate, but very general. Yes, a few of my favourite topics to comment on were there, but not others (I wonder what it thought of my participation in Thread Games?). I suppose that it was accurate as far as it went, but it didn’t go very far.

13

Try it with ChatGPT. I bet you’d get better results!

14

Frankly, I’m afraid to try that one!

15

Wow. Mine was-- wow…completely accurate, yet I sounded like Nobel-level scholar with the soul of a saint. I never could have written myself up that well. Even playing with the truth a bit, I probably could not make a version of myself like that, but this was entirely unimpeachable.

If I ever need to write a grant proposal, I am going to use Google AI for help.

Sheesh. Going to bear this in mind next time I read something with the AI touch.

16

When I ask who I am AI says “yes friend” in Spanish. Damn AI is amazing! :grin:

17

Most of the AI chatbots refuse to engage with me on inquiries of this nature; the few that do provide extremely limited responses. That’s because I’m in the EU, where GDPR rules and active regulatory enforcement make the tech companies reluctant to openly process and share personal information, unlike the unregulated free-for-all in the US.

I’m sure all the personal data are in there somewhere. But the chatbots have been programmed with enormous reluctance to reveal it.

18

I don’t care what AI says about me. However, there is a bit of information I have revealed about myself so that it will perform certain jobs coherently. Sometimes it “remembers” it, sometimes it doesn’t, which (according to my analysis) reveals something about its own protocols and priorities.

19

I put in my real name, and it spit back a portion of a bio I wrote about myself. Lazy AI – cheating by cribbing my own notes.

When I type in “CalMeacham” it tells me about This Island Earth

20

Well, now, lookie here!

I found that I am a retired Air Force colonel, and. . . NASA astronaut! I flew on two Shuttle missions, and started out as a helicopter pilot before transitioning into Special Ops. I flew over 50 combat missions in Panama, the Persian Gulf, and Northern Iraq.