A new setting on data permissions for Grok was quietly added and users opted in by default.
Okay, this probably comes as no real surprise, seeing as though it’s actually a key selling point of the offering. But last week, X quietly added a new, explicit setting within your account options that gives X permission to use your posts and activity in the app to train its Grok AI chatbot.
As explained here, X has now opted all users into this new setting, which states that X can use your posts, and your interactions with Grok, to train and improve the system. That information could potentially be shared with xAI, which, technically, is a third party, and as such, this usage may or may not be covered under the existing parameters that users agree to when signing up to an X account.
Which is seemingly why X has added this more specific setting, which is now active by default. You can check it out for your account here.
To clarify, X has been open about the fact that it’s training Grok on public X posts, as a means to build a more up-to-date and informed chatbot than other providers.
X owner Elon Musk has repeatedly touted the capacity of Grok to be ahead of other AI tools, both in terms of timeliness and censorship (or lack of), which, in his view at least, makes Grok a better model for AI development.
Why it is so important for @xAI’s @Grok to succeed. This is a much bigger deal than it may seem. https://t.co/X9qjvn836A
— Elon Musk (@elonmusk) May 7, 2024
Yet, it hasn’t been entirely clear on how it’s using X posts to build the system.
On the official overview of Grok, for example, which X describes as “an AI search assistant with a twist of humor and a dash of rebellion,” X explains that:
“Like most LLM’s today, Grok-1 was pre-trained by xAI on a variety of text data from publicly available sources from the Internet up to Q3 2023 and data sets reviewed and curated by AI Tutors who are human reviewers. Grok-1 has not been pre-trained on X data (including public X posts).”
So the Grok-1 language model wasn’t trained on X data, but there’s no mention anywhere in the overview that X is using public X posts.
Which Musk has said it is:
Grok has real-time access to info via the ???? platform, which is a massive advantage over other models.
It’s also based & loves sarcasm. I have no idea who could have guided it this way ????♂️ ???? pic.twitter.com/e5OwuGvZ3Z
— Elon Musk (@elonmusk) November 4, 2023
X’s overview of the process is somewhat opaque in this regard, which, again, is likely why it’s added this new toggle, which it’ll also need for compliance with EU regulations around data usage and enabling users to opt out if they choose.
It’s just not advertising the fact that you can restrict your data from xAI if you choose.
Which makes sense, because given the various controversies with Grok, a lot of users might well opt out, reducing X’s training data pool.
Grok’s news headlines, for example, have repeatedly been inaccurate and misleading, with the system misconstruing X interactions on various topics. And when you factor in Musk’s own controversial views on certain subjects, and his lax views on moderation, there is a risk that Grok will spread more mis- and disinformation, especially in the lead-up to the U.S. election.
Leave a Reply