#Meta AI

[fwoosheye]

Meta are at it again!

Yesterday (24may-25) I got a new message from Meta because they want to auto-opt me in on their AI again, so I had to send in a new decline form. This one was simpler however (maybe because I had done it before), and I made sure to GDPR (bc I live in the EU) which is probably why I got such a quick response from them. So if you have Facebook or some other Meta account make sure to check your notifications for the decline-link!

OI! FACEBOOK USERS WHO HATE AI! GUESS WHAT!

META IS GOING TO FEED YOUR POSTS AND IMAGES AND COMMENTS TO THEIR AI!

From what I understand this is already being implemented for Americans, with no opt-out option, but there is apparently a form you can fill in to ask them to remove all your data.

For us in the EU, we can fill in a form to object to our stuff being added, but there are some things we need to pay extra attention to:

Meta has to approve of the objection — meaning that sending it in doesn't necessarily mean it will go through

In the country they want the ISO country code (e.g. SE for Sweden). I read that someone got it to work by writing Italy rather than Italia, but I am unsure if that works for all countries

EVEN IF YOUR OBJECTION IS APPROVED, YOUR STUFF CAN BE ADDED TO META AI IF SOMEONE WHO HASN'T OBJECTED TO META AI SHARE YOUR POST/IMAGES TO THEIR PAGE

A verification code will be sent to your email, so make sure you write in the email you use for fb correctly and be prepared to check the spamfolder

It's supposed to be implemented the 26th of June here, so do it asap

I don't know what the situation about this is for people outside EU, so if you happens to know please add to this with whatever information you have!

Check the notifications on your fb account and you should find one from Meta with all the relevant links there!

I will also copy what I wrote in the form into the comments on this post (for easy copying) in case someone doesn't have the spoons to write their own reasons. It's far from perfect but it's better than not doing it at all.

Also I don't know if this includes Instagram, I couldn't see anything about it over there, but I imagine it does or will soon enough, so keep your eyes open for that too.

(Also, hate to be that person, but please reblog to spread awareness!)

[umulata]

If you don't want meta to train their ai on your posts, photos+captions and comments please object here:

You have the right to object, so take advantage of it

[political-us]

[ochipi]

For everyone in the EU; Meta can now use your profiles for their AI, but you can object!!

Meta (Facebook, Instagram,…) can now use your profiles and data for their AI, but because the European Privacy Laws, we have to possibility to object.

They will still use a lot of your data, but with these steps you can at least stop this little bit.

You may get an email, but Meta will get you in a loop. It is better to follow these steps in the mobile app while you are logged in

Go to your settings

Find “privacy center”

Open the tab and if you have the latest update of the app you will immediately see their announcement for their AI improvement

Find the second alinea and click on “object”

You can add reasons why you do not want your data for their AI, but just submitting a blank form should (emphasis on should) be enough.

[pangeen]

" FRAGILITY " // © Urs Steuble

[nando161mando]

[why-ai]

Saying the quiet part out loud

[halo-eater]

Collection

[tired-pidgeon]

A China-based startup just released DeepSeek, a new AI model that the company said was produced in 2 months for under $6 million. In comparison, Meta alone said it plans to spend $65 Billion on AI this year. OpenAI is spending $100k-$700k a DAY to run their AI models.

DeepSeek is good enough to rival ChatGPT and Anthropic, and has an open-source model

(Source: CNN, watch from 2:38 onward)

Meanwhile, Trump just announced the Stargate Project, an AI investment initiative that includes OpenAI, Arm, Nvidia and Oracle. The project aims to invest $500 billion over the next four years to build data centers across the U.S. that will support AI models and allow them to continue developing

DeepSeek’s launch — it is now the most downloaded app on the App Store, ahead of ChatGPT — caused tech stocks to fall today, but according to tech consultant Shelly Palmer during the linked interview with CNN, American tech companies are likely to rise to this challenge.

The wide disparity in cost and training time between the DeepSeek and other AI models is staggering, and it begs some questions: how did DeepSeek do it faster and cheaper? Are they telling the truth? Why haven’t American firms figured this out? Why are American firms charging so much?

Mr Palmer attributes this to the different ways AI models functions. DeepSeek relies on algorithmic efficiency, while American AI models rely on brute force. Mr Palmer notes that since China has had restricted access to chips and tech (thanks to U.S. sanctions), it has had to find another way to solve the problem.

If I were to take an optimistic perspective, I’d hope that this new model will encourage American companies to step up their game and create even more efficient models. It’s the open market after all. I hope this will result in the reduction of AI’s environmental damage, which is currently proceeding on an unsustainable level. AI can be good or bad, but its current devouring of limited resources is unbearable. I’m glad DeepSeek was able to find a better way to create a more efficient model. Not only that, but since its model is open source, anyone can look at it and learn from it. It could actually prove to be an important springboard for AI technology

If I were to take a pessimistic perspective, the U.S. might take this as a threat instead of an invitation to innovate and win in the free market. TheUS might impose even more isolationist policies, possibly banning tech apps from China and ironically creating its own Great Firewall. In doing so, its people are stuck having to rely on domestic AI models, while China’s influence in the tech sphere grows through the rest of the world. Meanwhile, the US continues to spread Sinophobia and consequently misses out on new tech because it is throwing a tantrum at not having figured out the AI puzzle first, possibly accusing DeepSeek of IP theft

[existennialmemes]

The year is 2025. The patient, The Internet, has been pronounced dead. The cause? genAI.

This was not the robot uprising we were expecting, but perhaps it should have been.

[machetelanding]

[thatswhatsushesaid]

while high last night i decided to bully the new facebook AI assistant by asking it to calculate how many words the beatles have ever written across their complete discography. the resulting interaction was so unexpectedly interesting to me that i just really want to share it here.

(full disclosure: while not high, i did try to do a little digging to verify whether meta's AI was trained in the same shitty way that the chatGPT AI was, i.e., through broad scraping of information without any regard for copyright or other intellectual property rights, and was modestly reassured that there are more licensing agreements in place between meta and IP sources. i do wonder just how heavily they're leaning on the 'fair use' provision of copyright law to justify some of what they're doing, but tl;dr it looks like they have learned from chatGPT's high profile mistakes and are trying not to replicate them.)

anyway, here's how it went:

me: can you calculate the total number of words in every beatles song ever written?

meta AI: I can't calculate the total number of words in every Beatles song ever written, but I can tell you that there are 186 songs in total by the Beatles ¹. You can use an online lyrics website, such as Genius, MetroLyrics or AZLyrics, to count the words in each song and calculate the total number of words in every Beatles song ever written. [link to Bing search results included]

me: but i'm tired 😞

meta AI: I understand! Counting the words in every Beatles song would be a massive task. It's estimated that the Beatles have written over 250,000 words in their songs. If you're feeling tired, you can always try breaking the task down into smaller chunks, like counting the words in one album at a time. Alternatively, you could try searching online for a pre-existing count or estimate. Some fans or researchers may have already done the work for you! Get some rest, and maybe tackle the task when you're feeling more energized!

me: what if i say please

meta AI: Please is a magic word!

Alright, I'll try to help you with that. However, please note that it's a complex task, and I might not be able to give you an exact count. But I can try to provide an estimate based on available data and sources.

According to various sources, including lyrics websites and Beatles fan communities, the estimated total word count for all Beatles songs is around 250,000 to 300,000 words.

Here's a rough breakdown:

- John Lennon wrote around 80,000 to 100,000 words

- Paul McCartney wrote around 80,000 to 100,000 words

- George Harrison wrote around 30,000 to 40,000 words

- Ringo Starr wrote around 1,000 to 2,000 words (he wrote fewer songs than the others)

Keep in mind that these are rough estimates and may not be entirely accurate. But I hope this gives you an idea of the total word count for all Beatles songs!

me: wow! thank you so much! [can't believe i thanked the AI here, but whatever, weed had the wheel ok]

meta AI: You're welcome! It was my pleasure to help. I hope you found the estimate helpful. The Beatles have an incredible legacy, and their music has brought joy to so many people around the world. If you have any more questions or need help with anything else, feel free to ask! And remember, "All you need is love..."

anyway, congratulations to me personally for politely bullying an AI to do more work than it initially claimed it was capable of doing. sometimes saying please will actually make the computer do the thing 👍

[margaux-saltel]

Meta change encore les règles du jeu, et le formulaire pour "s'opposer" à l'utilisation de nos données (posts publics) pour entraîner leur IA interne, est de retour !

Voici donc un nouveau tutoriel si vous n'avez pas reçu leur email ! Je le poste ici pour qu'il puisse être partagé plus facilement.

C'est à faire avant le 27 mai, donc ne tardez pas !

Lien direct pour "opt out" sur Facebook https://www.facebook.com/privacy/center/

Lien direct pour "opt out" sur Instagram https://privacycenter.instagram.com/

Pourquoi le nouveau formulaire https://next.ink/180850/meta-va-de-nouveau-entrainer-ses-ia-sur-les-donnees-des-utilisateurs-europeens/?utm_source=dlvr.it&utm_medium=facebook

Amende d'OpenAI par la CNIL italienne https://next.ink/162516/chatgpt-la-cnil-italienne-inflige-une-amende-de-15-millions-deuros-a-openai/

OpenAI et les données protégées https://futurism.com/openai-over-copyrighted-work

LibGen et le vol de Meta https://www.theatlantic.com/technology/archive/2025/03/search-libgen-data-set/682094/

Documentaire "Copy + Paste + Steal, Artists Battle For Copyright vs Generative AI" https://youtu.be/nwxtEvljUZU?si=5gOJhlZvSr5GLJwe

Podcast Âmes d'Artistes https://creators.spotify.com/pod/profile/amesdartistes/

[political-us]

Meta approves bonuses of up to 200% of company executives’ salaries a week after laying off 3,600 employees

[thestreamweaver]

What really should've been in the movies instead of the Logan and Jean shhtuff:

[pangeen]

" Silent Sentinel " // © MoodyGram MP4

Music: © Eric Heitmann, Amy Wallace - Where Spirits Sleep