Meta’s artificial intelligence half-fed: Zuckerberg will not train the AI with posts from European users
Meta (for now) stops and will not feed its AI the posts of European Facebook and Instagram users as requested by the authorities of the Old Continent. The company's response is piqued, bringing its rivals into play: "We are following the example of Google and OpenAI who have already used European data to train artificial intelligence"
For the moment, European users can breathe the proverbial sigh of relief. The American giant Meta, although disappointed, makes it known that it will not feed, at least for the moment, the contents that European and British users produce on Facebook and Instagram to its Meta AI virtual assistant.
OBSTACLES ACROSS AND ON THIS SIDE OF THE CHANNEL
At the basis of this sudden slowdown there is not only the cautious, not to say hostile, position of the European Commission which, it is known, worked hard in the last legislature to protect the privacy of its users, but also objections and comments sent to the Holding of Mark Zuckerberg by the Irish Privacy Authority and its British counterpart.
MENLO PARK'S SPICY RESPONSE
In a post , Meta said the Irish Data Protection Commission (DPC) asked it to delay training its large language models by leveraging public content from Facebook and Instagram profiles.
PRIVACY AND COPYRIGHTS
It is now clear to everyone that no AI algorithm, to use the popular saying, is "born learned": artificial intelligences need to be fed with an immense amount of data. However, this creates legal problems regarding copyright and, as the British authorities have highlighted, privacy.
As regards the first profile, more and more software houses are making agreements with the main publishers , so as to be able to exploit archives with news and editorials as a training ground (or canteen?) for their knowledge-hungry algorithms.
Meta, for its part, is a spurious publisher, very anomalous: the contents are not created by professionals (not only, at least), but by its users, who post billions of statuses and images on the Group's social networks every day.
And Meta obviously intends to exploit such a treasure for its own AI, even if several artists who use Instagram as a professional showcase have decided to oppose it and move en masse to social networks that prevent AI from devouring their works . And then there is the issue of privacy, which is particularly protected in the Old Continent.
HALF "DISAPPOINTED" BY THE AUTHORITIES' REQUEST
Meta said it was "disappointed" by the request, "especially because – the US company replies – we have incorporated the regulator's feedback and the European [Data Protection Authorities] have been informed since March". Meta recently began notifying European users that it would collect their data and in exchange offer an opt-out option in an attempt to comply with European privacy laws.
META FREEZES ITS AI IN EUROPE
“Without including local information we would only be able to give people a second-rate experience. This means that we are currently unable to launch Meta AI in Europe,” Menlo Park's response.
EUROPEAN DOUBLE PEACE?
But the most curious aspect is that the reply, in an unusual way, includes a justification that brings up its rivals: "We are following the example of others, including Google and OpenAI, who have already used European data to train artificial intelligence.”
In short, in a not too veiled way the Menlo Park Group suggests that there would be preferential treatment in the Old Continent.
THE RESPONSE OF THE IRISH AUTHORITY
“The DPC welcomes Meta's decision to suspend its project to train its large language model using public content shared by adults on Facebook and Instagram across the EU/EEA,” the DPC writes on its website .
The intervention of the Irish Data Protection Commission follows the dozens of complaints that Max Schrems ' NGO NOYB – None of Your Business has presented throughout the Old Continent.
The founder himself declared in this regard: “Meta is essentially saying that it can use any data from any source for any purpose and make it available to anyone in the world, as long as it is done through AI technology. This is clearly the opposite of being GDPR compliant.”
This is a machine translation from Italian language of a post published on Start Magazine at the URL https://www.startmag.it/innovazione/lintelligenza-artificiale-di-meta-nutrita-a-meta-zuckerberg-non-addestrera-lai-coi-post-degli-utenti-europei/ on Sat, 22 Jun 2024 04:44:12 +0000.