Elon Musk calls Grok ‘the most powerful AI by every metric’ but ‘secretly’ trains the new model with your X data by default

by

What you need to know

  • Elon Musk recently announced “the world’s most powerful training cluster” that will be used to transform Grok into “the most powerful AI by every metric by December this year.”
  • Interestingly, X recently shipped a new Data Sharing experience that’s enabled by default and uses your data to train Grok.
  • The feature is hidden in the platform’s privacy settings and can only be disabled on the web app.

As a seasoned researcher with extensive experience in the field of artificial intelligence and data privacy, I find Elon Musk’s recent announcement about Grok LLM being trained using “the most powerful AI training cluster in the world” intriguing, but also raise some concerns.

As a researcher, I recently came across an exciting development in the world of artificial intelligence. Billionaire Elon Musk, through his company xAI, made an announcement on X (previously known as Twitter) that they had initiated the training process for their Grok Large Language Model using what is claimed to be “the most powerful AI training cluster in existence.” Musk expressed his confidence that by December of this year, the Grok LLM would surpass all other AI models and become the “most powerful AI by every metric.”

It turns out that Grok AI requires more than just “the most powerful cluster in the world” for its training, as previously thought. A recent tweet from @EasyBakedOven on X revealed that Twitter has activated a setting by default for all users, allowing them to use your data for training Grok. Even more surprising is that Twitter implemented this feature quietly and began using data for Grok’s training without explicit consent.

Read moreApple Maps now works on your Windows PC... sort of

It wasn’t entirely unexpected that OpenAI, the creator of ChatGPT, would face challenges related to copyrighted content while developing such AI tools, given the increasing number of copyright infringement lawsuits against them. However, OpenAI clarified that from a legal standpoint, copyright law does not prohibit the use of copyrighted material for training purposes.

The X investigator pointed out that disabling the feature through the mobile application is not an option. On the other hand, disabling the feature on the web application is feasible, although it’s not easily accessible.

How to disable the data sharing feature on X

X’s intent in utilizing your data to enhance Grok’s efficiency and effectiveness might explain why this new feature was secretly released and activated by default. It should be noted that disabling this functionality seems to present a challenge for typical users.

To prevent Grok from using your data for training, navigate to the Settings section of the web application. Under the Privacy and safety tab, locate and choose Grok. You’ll be taken to the Data Sharing page. Here, you will find that the option allowing Grok to use your posts, interactions, inputs, and results for training and fine-tuning is activated by default. To deactivate it, just toggle off the switch.

Read moreRemnant 2 announces third and final DLC 'The Dark Horizon' on its birthday, and its release date isn't far off

The following explanation is offered under the unsolicited Data Sharing feature:

In order to enhance your experience consistently, we might utilize your X posts in addition to your user activities, inputs, and outcomes with Grok for the purpose of training and refining. Consequently, your user activities, inputs, and outcomes could also be transmitted to our service provider xAI for these reasons.

A study published in the journal Nature raises concerns about the potential negative effects of AI systems using the same data repeatedly, which some experts have referred to as “AI inbreeding.” Another report, however, questions the widespread use and hype surrounding AI, suggesting it may be a passing trend. Regarding chatbots specifically, there are growing doubts that they have reached their pinnacle and could become less sophisticated with increased regulatory oversight.

Read More

2024-07-26 14:39