Meta starts rolling out generative AI tools for all advertisers | Hacker News Reader

Meta starts rolling out generative AI tools for all advertisers

reuters.com

60 pointsmfiguiere35 comments

Surely this tool will be used responsibly and not lead to a renaissance in misleading ads, phishing scams, and malvertising. I trust Meta to have put in the proper safeguards to make this tool simply a net benefit for online users.

I think the technology behind all of this is so cool and I love it, but it's difficult not to assume that all of this will ultimately result in the crashing and burning of modern civilization.

It's certainly going to result in wide-sweeping changes, one way or the other.

I’ve noticed recently a flood of weird ads on YouTube. I realized that automated western-sounding voice overs allow companies, I assume from China, to make up some bullshit ad using rhetoric similar to “one weird trick” and “doctors hate him!” to hook people in to sell some random cheap device.

The cost of a full video ad (using stock video) and a custom voice over is so low we are being flooded with absolute nonsense ads. They literally don’t make sense but they must catch suckers.

It’s gonna be an issue when these things are more logically coherent and visually mature.

did you forget the /s?

Was not needed :)

you never know how many FB sycophants are out there though

I assure you that I am, in fact, not a fan of the fact the brightest minds in the past 20 years have been dedicated to developing and selling more efficiently targeted ads.

I hate to imagine what kind of actual societal progress we could have made had these brilliant minds been working on nearly anything else. So much wasted potential just to serve an ad that gets ignored a tiny bit faster.

well, the brightest minds as self-described by the people doing it.

And the people doing it self select for showing the ability to work in a disingenuous self-advocating marketing field, so you might expect them to talk themselves up.

i don't think the actual brightest mind could stand working in the field for more than about 5 seconds

You can argue that they aren't the top of the pyramid and I'd give you that, in exchange for the admission that the people working for FAANG, MANA, or whatever, are certainly a cohort of smart people who's skills could have gone to basically anything else and humanity would have benefited for it.

The "surely" is a dead giveaway.

i think the username is a pretty good guage of what's going on :)

Your delivery is beyond deadpan to a kind of exhumed-pan I find impressive. Superior snark sir.

Fortune favors the bold.

Pretty easy to block malevolence. Just permaban advertisers that misuse the tools. It's not like you'll struggle to collect evidence on their misdeeds.

They don't do that now[1], why would they do it in the future? There's literally no incentive for them to do it. It's obviously bad for business, since the 'customer' isn't the user, it's the advertiser.

[1] such as https://www.bleepingcomputer.com/news/security/almost-60-per...

I would love to see a deepfake video of Zuck saying the exact opposite of this. You know that would be the truth.

There are already fully generated display ads running on e.g. Apple News.

They are not very good.

Meta's thing is different. It's three specific workflows when you upload ads in narrow product categories that make sense.

This isn't meant to be a big educational comment about the digital ad industry which I work in. It's complicated, but Meta already knows who the clickers and buyers are, they tend to click and buy everything; if you select the conversion optimizations, track a purchase and report it to Meta, it isn't super material what the creative is. You pay an exorbitant cost per conversion, which doesn't change here. You save a ton of time setting up display ad assets though. It's a good, narrow, sensible application, but it isn't the Big Thing you'd expect them to do.

I tried to train Facebook. First thought: I absolutely do not want it to advertise to me anything I might actually want to buy (like books): that feels creepy. So maybe I could deliberately interact with ads for things I'll never buy, like fashion? (I dress myself out of charity shops, mostly.) Second thought: If I interact with ads with hot guys in them, Facebook will show me more hot guys, which is nice eye candy.

It seemed to work, for a while. Some of the ads were for underwear, definitely targetted at gay men, and bordering on being NSFW (which is fine, as I don't Facebook at W). And then I mostly abandoned Facebook.

---

For YouTube, before I got a free deal on YouTube Premium with my phone plan, things were weirder. YouTube's video recommendation algorithm knew me well, but its ad recommendations were very confused, including about which language I spoke. It weirdly often gave me ads for scientific lab equipment, sometimes in Hebrew. (I'm a web developer, am Irish, and live in Ireland, and don't speak any Hebrew. I have an interest in linguistics, and have read up on the origin of the Semitic trisyllabic root, but that doesn't mean I speak Hebrew.)

In general, there are certain tells of what constitutes a serious business and advertisement. This makes it easy to the trained eye to filter out the blatant frauds. But coming advancements in AI generated content will blur the line between legitimate and illegitimate. The only reasonable approach will of course be to trust nothing you see on the internet. Very sad.

How about AI generated profiles? Befriend an AI, live life like The Truman Show.

The "won't someone think of the children!" on this has already started: https://thehill.com/opinion/technology/4218666-ai-girlfriend...

Already done.

https://chirper.ai/

Like the Truman Show, the inhabitants are pleasantly polite...

A lot of AI companies will be freed to focus on other things. Established companies will offer everything under the sun for "free", allowing small players to focus on other important things, such as working for the corporations that put them out of business.

Can’t wait for the paid ad free meta estate!

I just saw an ad that has a convincing video of Mark Zuckerberg pumping a crypto trading platform called Libre. I can’t even tel if it’s real because 1 it uses his face voice, 2 it’s called Libre with the copy of the logo, 3 I saw the ad on Facebook

Google already had a programmatic advertising tool that would take a set of images, logos, copy, etc that you provided, and move things around to try and identify what worked best.

That’s just automated A/B testing and very different from generative AI

Hey sorry this is unrelated to this post, but a couple months ago you posted a link to a barbell plate calculator you wrote hosted on AWS. I was hoping you could share the unminified JS for it somewhere (github or something)? I want to potentially tweak it to avoid out-of-order plate ordering. Thanks for writing it btw!

Which tool is this?

I think they call this responsive ads, you can create them on ads.google.com

Yep. It's been the standard format for manually building SERP ads for a couple of years now.

There are also Dynamic Search Ads which automate even more of the process of setting up a campaign based on your current website: https://support.google.com/google-ads/answer/2471185?hl=en

I got an email about some kind of generative ai and chat bot that was rolling out in the ads platform, but I still don't see it on any of our accounts.

Responsive search ads are not generative ai

If someone from Facebook Ads is reading this, can you please reach out to me? hi at my username.com. There's a bug in your "suspicious log in detected" system and I'm trapped in a customer support loop and no one is able to properly escalate the situation.

Reply on news.ycombinator.com