AI Trickery and Why Fiverr Video Calls Will Soon Be Untrustworthy

[emmaki]

Now sellers who rely on ChatGPT to communicate for them can also use the magic of artificial everything to do Zoom calls on Fiverr. I won't be naming the software here (do your own darn legwork!), but it is rather impressive.

AI video has been capable for a while now of making it look like you're "always looking at the camera", which is very useful for when AI interview software is listening to difficult questions and feeding you answers. 

AI photography can work from, let's say 30 headshots, in your phone gallery to create a "you" (and of course make you look nice instead of the terrifyingly ugly mess you actually are). 

It can then turn that gorgeous headshot into a talking video, complete with hand movements, nods, and other slightly repetitive natural gestures! If not in real time, then soon. 

Oh, and it'll also clone your voice. What's that, the interview's in another language like say, English? No problem. 

You don't actually know anything about the job or what you do because AI has reduced your brain to the size of a raisin and all you know how to do is write prompts? No problem. Feed the AI with knowledge and sound smart.

===

That's just a few of the things this did. Is it any good? I have no idea, but even if it isn't that good now, it will be a lot better soon and close to perfect later on. 

What I will say is that the demo looked really impressive. The guy was gurning and thumbs-upsing for a TikTok video as the interview continued in the background. Anyway, I know quite a few people use Zoom calls to establish their humanity and realness and knowledge - but that advantage is rapidly disappearing. 

Side-note: this kind of technology is almost certainly going to start being used by scammers in the near future. Be careful, children, and remember Fiverr has no AI disclosure policies. 

Quick question for Fvierr Staff, on the off-chance that anyone reads this on the single day of the week that is the weekend in TLV and the West: What measures is Fiverr taking to prevent this kind of technology from being used in its on-platform Zoom calls? I ask, because it's very clear that many sellers and buyers are not as AI-aware as they could be, leaving them vulnerable - and many users already fall victim to less sophisticated scams. As a platform, you do have some duty of care here. 

In brighter news, it looks like regulations will soon mandate the watermarking of generative AI-created stuff, which should help actual detection rates increase a lot. Bad news for the people who are building entire businesses around their ability to "write" and "draw" and whatever else when their actual ability is that of a sprouting potato. 

Edited March 23 by emmaki
forgot to add the fake video stuff

[smashradio]

Yeah, I've seen this software, and it's scary. It will clearly be used by scammers, and probably already is in wide use. 

[emmaki]

The voice cloning stuff definitely is. In the past few years in the UK, scammers have somehow been able to clone people's voices (my guess is answering phone calls from unknown numbers). They then use the cloned voice to contact a family member and con them into paying thousands for various fake situations - usually ones that prey on the desire to help - so they send money to XYZ without thinking about it.

There's been quite a bit of publicity about already, but people still get scammed. It's also fooling people who are otherwise quite technically savvy. I'm not quite sure what to make of the era we're moving into, since we effectively can't trust anything anymore! And the problem is that no matter how much education you have, hearing (or worse, seeing) a loved on in distress almost always short circuits any natural suspicions and distrust. 

[sabbirthedealer]

Thank you

[texvox]

Things like this do make me wonder (genuinely) what protections Fiverr has in place to ensure AI is not used in un-savory or un-ethical ways on the platform? I would assume any company going as all-in on AI as Fiverr has probably had some meetings about this. There are a lot of AI-based scams out there in the ether outside of the platform already. With the rapid advancement of these tools they'll find their way here on no time flat, and I don't think that a simple policy requiring people disclose their use of AI is going to cut it. 

Edited April 19 by texvox