You may have encountered problems, this Saturday, July 1, when logging into Twitter. If many users first thought of a social network bug, Elon Musk came to clarify the situation: the blue bird application will temporarily restrict the reading of tweets to contain the massive use of social network data by third parties, in particular to feed artificial intelligence models.
Concretely, the platform will limit reading to 6,000 messages per day for verified accounts, 600 for unverified users and 300 for new unverified accounts. Ceilings which should be “soon” raised to 8,000, 800 and 400 respectively, according to the executive chairman of the board of directors.
Rate limits increasing soon to 8000 for verified, 800 for unverified
A decision made “to remedy the extreme levels of data collection and system manipulation”, explained Elon Musk in a tweet.
The day before, he had already announced that it would no longer be possible to read messages on the network without connecting and giving his identifiers.
“Hundreds of organizations [perhaps more] were gleaning data from Twitter very aggressively, to the point that it disrupted ordinary usage” by Internet users, according to the majority shareholder of the San Francisco group. By limiting the number of tweets that can be read per account, it seeks to prevent these organizations from collecting massive amounts of data that are used in particular to develop so-called generative artificial intelligence (AI) models.
To build a generative model that can respond human-like to plain language requests, these companies need to “train” the interface by giving it example conversations. “Almost all the companies that do AI, from start-ups to the largest groups in the world, were collecting large amounts of data,” Elon Musk insisted. “It’s a bit frustrating to have to urgently add a large number of servers just to justify the obscene valuation of some AI start-ups,” he said.
Twitter is not alone in dealing with the consequences of the acceleration of generative AI and the development of services built around language models. In mid-June, the discussion platform Reddit raised the prices it charges third-party developers to use data and conversations posted on the social network. The decision had caused an outcry, because these platforms have so far provided access to public data on their site at moderate prices or free of charge, to promote the development of an ecosystem.