OT: Fooled by AI

[AnthonyL]

I needed to call the NS&I helpline to get my registered telephone number changed in order to then receive an OTP. The landline number they had for me was in the form +44 01xxxx and I wasn't getting the OTP.

A very helpful sounding voice answered my call and asked for a description of the problem. I couldn't get over the issue of the format of the phone number being the suspected problem and it was only several minutes later as my temper was rising that I realised the "helper" was a bit thick. Usually automated voices are quite readily identifiable. I'm foreseeing problems with this approach.

[John Rumm]

Sometimes asking the AI to let you speak speak to a human works :-)

[Scott]

I got caught in exactly the same way. It was only when the same question was repeated that I realised it was AI.

[Graham.]

Was it "Chris" trying to sell you loft insulation?

Note that in this recording, "Lenny" is running on my phone systen, I hesitate to call him an "AI" as there is no intelegence, he simpply waits for a pause in the incoming speech before delivering his next phrase. "Chris" on the other hand does seem able to deliver limited context-sensitive replies, and yes, the first time he called it didn't occur to me that I was talking to a recording.

[Jethro_uk]

To respond to your title, I don't think I've been "fooled by AI" -quite the reverse.

However I have engaged with chirpy chatbots that clearly have only one end point in their rambling.

[Brian Gaff]

Yes when designing an AI often they do not teach the whole range of possibilities to the system resulting in it not knowing about your problem. I get this with being blind. The classic ones are the router faults where its just like the script readers of old.They have no fall back if you are unable to see lights on the unit. I can quote a lot of data from the router internal diagnostics, but they know nothing. Until these devices can default to sending you to a human if they cannot understand where you are coming from, I tend to put the phone or stop typing to, them, and try to find another way. Brian

[Theo]

Would you consider making that searchable? A while back I was trying to explain to somebody who had been caught out by 'Chris', but Youtube's search couldn't find the video.

(although I'm tempted to agree to the 'free' energy assessment just to see what happens... and waste their time of course)

Theo

[Graham.]

I've set it to "Public" Theo.

It must have been my intention when I put a transcript in the description. I'm not sure how long it takes Google to index Y.T. descriptions.

[Theo]

Thanks. That video is new today, so the one I was previously looking for can't have been this one. Sounds like there's plenty on the internet about 'Chris', but nobody has noticed he's an AI.

Theo

[Graham.]

It's the same video, It's been up for a year, its just that Youtube has reset the date when I changed it from "unlisted" to "public".

A Google search for any of "Chris's" script now brings up my video, I have added an explanation to the description about it being a recording.

[Theo]

That's good. It comes up second in Startpage's video search for 'Chris local energy advisor' - not showing on the text search yet but hopefully the algorithm will bubble it towards the top.

Theo

[John Rumm]

I have made a wiki page that gives some background and links to it:

[Dave W]

I've just watched it - Lenny is absolutely brilliant! When I was young I made a device with a loudspeaker mounted on chipboard with eyes and nose drawn above the aperture. It made various grunts with pitch controlled by 8 different curves on a wheel scanned by a photocell. It waited for pauses in mother's boring diatribes, then played the next "mmm", giving the impression it was understanding and sympathising with her.