Meta Trains Its New AI With Your (Mostly) Public Instagram and FB Posts
Meta has revealed a whole host of new AI-powered features and platform technologies that will be attaching themselves to its services.
One of these is called Meta AI, which works as a virtual assistant that will “live” on the social media platform’s services and help users. Well, Meta has trained this particular AI with the Instagram and Facebook post of its billions of users.
On the one hand, it’s only natural for a tech giant like Meta to completely embrace the AI game.
The technology has advanced by leaps and bounds in the last year, and become pretty much ubiquitous to the point of saturation across the whole landscape of technology services worldwide.
We’re now seeing usable, even excellent AI tools being added to many common software platforms that all of us use as photographers, creators and individuals.
With all of that happening and the fizzled response to Mark Zuckerberg’s drive for the “Metaverse” in the last couple of years, the company naturally pivoted to AI too and has the potential to make a huge impact with this.
Given Meta’s uniquely vast and largely walled-off troves of social media content including photos, public posts, comments and probably even personal messages, their potential training set for any chat AI is enormous.
With all that in mind, we might be able to expect genuinely impressive behavior from the suite of AI technologies that the company has announced for editing images on Instagram and its Meta AI assistant.
According to a press statement from the company, “Meta AI is a new assistant you can interact with like a person, available on WhatsApp, Messenger, Instagram, and coming soon to Ray-Ban Meta smart glasses and Quest 3,”
Meta’s announcement further elaborates that Meta AI is “powered by a custom model that leverages technology from Llama 2 and our latest large language model (LLM) research.
In text-based chats, Meta AI has access to real-time information through our search partnership with Bing and offers a tool for image generation.”
While the company has been a bit vague on the degree and details of how it used user content to train its AI, the company’s president of Global Affairs Nick Clegg has admitted to Reuters that public FB and IG posts were partially used for training.
Clegg added “We’ve tried to exclude datasets that have a heavy preponderance of personal information,” and underscored that the “vast majority” of the data used by Meta was publicly visible on its social platforms.
He essentially emphasized that the company stuck to this because wanted to respect privacy concerns.
However, Meta has a long history of doing exactly the opposite. Also, if you read the obvious evasiveness of his statement, it leaves lots of room for them having also used private communications and content for their training to just about any degree they please.
After all, if the minority of the content used for training was still private as insinuated by Clegg’s comments, little stops it from having really been even more than admitted or being more heavily down the road.
Clegg then explains that public photos, text posts and comments were used for the AI training. The photos in particular were used to train Emu, the AI that’s the backbone of the forthcoming Instagram image editing tools.
IG and FB text posts on the other hand were used for the Meta AI chatbot. In both cases, he reaffirms the public nature of the content being used, or okay, at least the “vast majority.”
If you’re posting content of any kind to either of these social media platforms from Meta, you probably already know full well that you have little expectation of privacy for anything you upload.
However, if you’re worried about copyright issues around your text and pictures being used for the vast and profitable company’s AI product development (obviously without compensation to users like us), then the above is also worth thinking about.
Clegg himself touched on the hot-button issue of copyright concerns and mentioned that he expected a “fair amount of litigation” to decide if copyrighted material can be added to AI training under fair use provisions.
He blandly stated, “Whether creative content is covered or not by existing fair use doctrine… We think it is, but I strongly suspect that’s going to play out in litigation,”
Since Meta’s terms and conditions for joining its social media platforms in any case allow the company-wide leeway in how it takes advantage of user content, their jumping on that content for making their AI technology as competitive as possible goes without saying.
The litigation can be solved later by their lawyers. This doesn’t mean that you’re likely to find your photos visibly appearing as part of some AI-rendered visuals generated by Meta technology.
However, it is possible that they play a tiny part in some aspect of what their image-editing AI tech is capable of.