> I'm not confident that I'd be able to distinguish GPT-4o from a human speaker in the best of circumstances and I'm almost certain that I could be fooled if I'm hurried, distracted, sleep deprived or otherwise impaired.
Set a memorable verification phrase with your friends and loved ones. That way if you call them out of the blue or from some strange number (and they actually pick up for some reason) and you tell them you need $300 to get you out of trouble they can ask you to say the phrase and they'll know it's you if you respond appropriately.
I've already done that and I'm far less worried about AI fooling me or my family in a scam than I am about corporations and governments using it without caring about the impact of the inevitable mistakes and hallucinations. AI is already being used by judges to decide how long people should go to jail. Parole boards are using it to decide who to keep locked up. Governments are using it to decide which people/buildings to bomb. Insurance companies are using to deny critical health coverage to people. Police are using it to decide who to target and even to write their reports for them.
More and more people are going to get badly screwed over, lose their freedom, or lose their lives because of AI. It'll save time/money for people with more money and power than you or I will ever have though, so there's no fighting it.
The way to get around your side channel verification phrase is by introducing an element of stress and urgency: "omg, help, I'm being robbed and they need $300 immediately or they'll hurt me, no time for a passphrase!" can additionally feign memory loss.
Alternatively while it may be difficult to trick you directly, phishing the passphrase from a more naive loved one or bored coworker and then parroting it back to you is also a possibility. 'etc.
Phone scams are no joke and this is getting past the point where regular people can be expected to easily filter them out.
Or just ask them to tell them something only you both know (a story from childhood, etc). Reminds me of a book where this sort of thing was common (don't remember the title):
For many people it would be better to choose specific personal secrets due to the amount of info online. I'm not a very active social media user, and what little I post tends not to be about me, but from reading 15 year old Facebook posts made by friends of mine you could definitely find at least one example on each of those categories. Hell, I think probably even from old work-related LinkedIn posts.
We had a “long lost aunt” come out of nowhere that got my phone number from a relative who got my number from another relative.
At that point, how can you validate it, as there’s no shared secret? The only thing we had was validating childhood stories. After a preponderance of them, we accepted she was real (she refused to talk on the phone — apparently her voice was damaged).
We eventually met her in real life.
The point is, you can always use these three principles: asking relatives to validate the phone number — something you have — and then the stories — something shared — and finally meeting in real life — something you are.
Oh, you remember those little games that your mom played on facebook/tic tok that asked her "Her favorite", sorry she already trained the AI who she was.
I only say this sort of jokingly. Three out of four of my parents/in laws are questionably literate on the internet. It wouldn't take much of a "me bot" for them to start telling it the stories of our childhood and then that information is out there.
Set a memorable verification phrase with your friends and loved ones. That way if you call them out of the blue or from some strange number (and they actually pick up for some reason) and you tell them you need $300 to get you out of trouble they can ask you to say the phrase and they'll know it's you if you respond appropriately.
I've already done that and I'm far less worried about AI fooling me or my family in a scam than I am about corporations and governments using it without caring about the impact of the inevitable mistakes and hallucinations. AI is already being used by judges to decide how long people should go to jail. Parole boards are using it to decide who to keep locked up. Governments are using it to decide which people/buildings to bomb. Insurance companies are using to deny critical health coverage to people. Police are using it to decide who to target and even to write their reports for them.
More and more people are going to get badly screwed over, lose their freedom, or lose their lives because of AI. It'll save time/money for people with more money and power than you or I will ever have though, so there's no fighting it.