If I'm reading this right, the playbook was... deleting scam ads ? And the implied problem is they only deleted searchable ads, and not trying harder to get rid of all of them.
It's interesting that Facebook was trying NOT to uncover identities, they're famous for insisting on real names.
All ads are searchable. They found the exact words and phrases that regulators used and then made sure those were clean.
>As a result, Meta decided to take the tactic global, performing similar analyses to assess “scam discoverability” in other countries. “We have built a vast keyword list by country that is meant to mimic what regulators may search for,” one document states. Another described the work as changing the “prevalence perception” of scams on Facebook and Instagram.
> the implied problem is they only deleted searchable ads
Well, more just the ads that matched the specific queries the regulators were using. So yes, they removed some scam ads, but there are probably many more that people are still seeing just because those didn't match the search queries the regulators were searching for.
> It's interesting that Facebook was trying NOT to uncover identities, they're famous for insisting on real names.
It isn't really surprising. If they required real identities, they wouldn't be able to make money from scammers using throw-away accounts, or from entities subject to US sanctions, so there is a monetary incentive not to know the identity of the ad customers.
If this method actually removed a significant percentage of scam ads, rather than just heading off scrutiny, then a) doing proper verification wouldn't cost them $2b a year like they claim it would, and b) their quarterly revenues would be taking a meaningful (single digits %) hit and the share price would suffer.
This is the same tactics VW used. Find out about the test the regulator uses and focus on passing the tests instead of complying with the rules the regulator wants to enforce.
I believe that's incorrect - the article quotes Meta as saying "By cleaning those ads from search results, the company is also removing them from its systems overall".
The real problem as I understand it is that they didn't stop the ads from entering the system, but rather identified the words used by regulators and only deleted those ads (after an unspecified amount of time online) from the system.
My friend's eSIM experience with Tello was pretty good. Their kid got their first phone with an eSIM, and it was stolen a few months later. They were able to transfer the number to a new phone from the Tello website.
Mobile customers are a kind of second-class citizen for McDonalds, though they have instructed the drive-through salespeople to always lead with "Will you be using the mobile app today?". MCD uses location tracking to only submit the mobile order once you are in range of the store.
Should probably add that in-store customers are third-class citizens; drive-through orders without customization get priority.
> MCD uses location tracking to only submit the mobile order once you are in range of the store
Interesting. Can they be certain that the device from which one places the order, is also taken along for the ride? Some people might order from a separate phone or tablet, which stays at home. Perhaps all of this customer's devices have the McD's app; and any one of them may trigger the order, by approaching the store.
I haven't used the McDonalds mobile app in the drive-thru line. Can you clarify the experience? Why would I use the mobile app once I'm already at the drive-thru? For points/payment? Or for ordering as well?
I once ordered ahead using the mobile app at a Starbucks. I got out of my car, only to find the doors locked, as they closed early for some reason. So I had to get BACK in my car, and sit in the drive-thru queue, just to pick up my already-completed drink, which I found infuriating. (actually, what was infuriating was that I really only order at that location to use their restroom, which of course was an option unavailable in the drive-thru for hygenic reasons)
It's useful for large orders where everyone wants time to review the menu and/or customize the toppings. Will spend a few minutes entering everyone's order to the app and checking out.
At the drive-through speaker spend 10 seconds saying the code instead of 90 seconds ordering, with no chance of mistakes. The line still needs time to move, so the time savings isn't as big as it sounds, the order accuracy is what I appreciate.
You might save another 30 seconds at the payment window, they ask if you had a mobile order and then give you the receipt.
If the line has moved too efficiently, you will receive your drinks and be asked to wait in a reserved parking spot. This is the "time-out corner" for patrons with too many customizations on their large order, its still more convenient than picking up inside. In 5 minutes someone will bring the rest of the order to you.
The app has "20% off orders over $15" as a large incentive to use it.
Why can't the plan be judged on its merits? Rigorous verification of the idea is a good thing that should happen anyways. The main potential problem I see is transmission of privileged information to a third party.
I assume they are working at a business to make money, not a school or a writing competition.
Because AI can generate meritless works far faster than anyone can judge their merits. Asking someone to read your AI thing is basically asking someone to do the work for you. If you respect your colleagues time, you should be sharing your best version of inputs, not raw material. Not only that, you should have thought about and be able to defend it. Throwing some AI thing over the fence, you haven’t thought about it either, why would you expect your colleague to?
I’d add to that, long form AI output is really bad and basically unsuitable for anything.
Something like “I got GPT to make a few bullet points to structure the conversation” is probably acceptable in some cases if it’s short. The worst I can imagine is giving someone a “deep research” article to read as if that’s different from sending them to google.
This is a trust issue. If someone I trust hands me a big pr, I focus on the important details. If someone i dont trust hands me a big pr, i just reject it and ask them to break the problem down further. I dont waste my time on this kind of thing, regardless of whether it was hand written or generated.
Yes I made the assumption that the person who "put the plan together" did their own diligence of reviewing it before emailing, but maybe that is too charitable for an "AI plagiarist".
If someone sends me incomplete work I will judge them for that, the history of the work relationship matters and I didn't see it in the blog post.
> Why can't the plan be judged on its merits? Rigorous verification of the idea is a good thing that should happen anyways.
Situational.
I don't know this blogger or what the plan involved; but for sake of agument, let's say it was a business plan, and let's say in isolation it's really good, 99.9% chance of success with 10x returns kind of good.
Everyone in whatever problem space this is probably just got the same quality of advice from their own LLM prompting. That 99.9% is no longer "in isolation", it is a correlated failure where all the other people doing the same thing as you makes it less viable.
That's a good reason not to use a public tool, even when the output is good.
Correlated risk disguised as uncorrolated risk was a big part of the global financial crisis in the late 00s.
The problem comes from the asymmetry between the effort that went into generating and judging. You can have one person spinning out documents that can keep a whole team busy and dragging everyone down.
Along the same lines as "A lie travels around the globe while the truth is putting on its shoes."
If the documents they're putting out are bad, then they're doing bad work and that eventually comes with consequences from your coworkers and superiors. If they're doing good work, then great! Who cares if an LLM wrote most of it and they just edit it? That's not super different than the current relationship between senior and line workers.
I guess I'm making some assumptions here. But I've been asked to review some documents before. Maybe I didn't notice the ones that were good. But my general assumption is that if someone gives me the output of an LLM to review, it's not going to be good work. In my experience it hasn't been good work generally.
Because of the difference in effort involved in generating it vs effort required to judge it.
Why are you entitled to "your" work being judged on its merits by a real human, when the work itself was not created by you, or any human? If you couldn't be bothered to write it, why should someone else be bothered to read it?
Well, clearly, you know a lot about being a serious enterpreneur. Don't let us luddites drag you down, I'm sure your next 100% vibe coded B2B SaaS will be a massive success.
So many technologists offended at the use of technology. Next they’ll insist on pen-on-paper for truly authentic work product, and after that, 3 days’ wilderness meditation on it, to prove you really internalized it.
Look, it’s now like, email in 2004. You see spam, that it has found email. It doesn’t mean you refuse to interact with anyone by email, write geocities posts mocking email-users. You just acknowledge the technology (email) can be used for efficiency, results, and it also can be misused as a giant time-waster.
The author of the article here is basically saying “technology was used = work product is trash”. The ”spam” folks are seeing must be horrible to evoke this kind of condemnatory response.
Nvidia has become an incredibly important strategic and economic American asset. If it ever seemed that the CEO of such an asset was colluding with a foreign nation (especially China), you can bet things wouldn't end too well. People have been done for far less, just look at what happened to Suchir Balaji which got covered up.
I appreciate you giving it a shot, then! Our experiences are pretty different; I've almost always found solid working apps from F-Droid, and my main messenger, web browser, podcast app, ebook reader, RSS reader, calendar, tower defense game, and home screen launcher are all from F-Droid. I use the Play Store mostly for Google's own apps (namely Maps and YT Music) and my banking app.
reply