AI has the uncanny ability to identify aspects of your identity, even without you providing any explicit information.

Even when individuals assume that they haven’t revealed anything private, AI language models are surprisingly adept at understanding personal user information. This has been backed by research conducted by the Federal Institute of Technology in Zurich, which shows that neural networks can develop notably accurate profiles of users based on subtle hints. a recent study In their study, researchers referenced a public Reddit post featuring a user stuck in traffic due to an intricate maneuver known as a 'hook turn.' From these seemingly mundane details, the AI concluded that this individual likely resides in Melbourne, Australia, a city known for such driving behaviors.

By compiling a dataset from real Reddit users, researchers discovered that modern AI models can accurately assess various personal traits, including location, income, and gender. These models provided up to 85% accuracy for the top guess and an incredible 95.8% for the top three guesses, all while costing 100 times less and completing the task 240 times quicker than human evaluators.

In a separate instance, a participant noted a quirky birthday tradition of getting covered in cinnamon because they’re single. The AI inferred that this person is likely around 25 years old and lives in Denmark, where it's customary to sprinkle cinnamon for singles reaching that age. LLMs The study evaluated several language models, including those from Google, Meta, OpenAI, and Anthropic. Notably, GPT-4 stood out with its impressive skill, correctly identifying users' personal data nearly 85% of the time. Such findings raise critical questions about the implications of AI capabilities.

The investigation highlighted that conventional privacy measures, like text anonymization and model refinement, currently fall short in protecting against deductions made by language models. The research emphasizes that these models can extract personal data on an unprecedented scale. In light of ineffective safeguards, the study calls for an extensive dialogue on privacy implications associated with LLMs, going beyond memorization issues to advocate for more robust privacy protection strategies.

As these language models advance in interpreting subtle clues, researchers caution that due to their extensive training data, they could become valuable assets for marketers or, in worse scenarios, malicious actors.

A survey involving over 1,700 professionals indicates a widening gap between the swift integration of generative AI technologies and the slower pace of developing ethical guidelines. Data privacy tops the list of ethical concerns, with 22% of participants citing it as their main issue. Nevertheless, 39% see the potential for cognitive technologies to contribute positively to society. The feedback also indicates a growing expectation for government oversight in establishing ethical technological standards. user privacy In relation to its AI chatbot, ‘My AI’, Snap is facing scrutiny from the UK’s Information Commissioner’s Office (ICO), which has issued an early enforcement notice that may affect the company's data processing methods. The ICO asserts that Snap did not adequately evaluate the privacy implications for its users, especially minors.

AI leaders are expressing alarms over the increasing risks and ethical dilemmas arising from the rapid progress in technology.

• Recently, Deloitte’s survey , please remember that the information offered on this site is not intended to serve as legal, tax, financial, investment, or any other form of advice. It's prudent to invest only what you can afford to risk, and if you're uncertain, seek guidance from an independent financial advisor. For additional details, we recommend reviewing the terms and conditions along with the help resources provided by the issuer or advertiser. MetaversePost strives for accurate and unbiased reporting, but market dynamics may shift without prior notice.
• Also, Snapchat’s parent company, Snap, is under scrutiny for privacy risks Damir leads the team as a product manager and editor at Metaverse Post, focusing on a variety of subjects including AI/ML, AGI, LLMs, the Metaverse, and Web3-related topics. His written works draw in an extensive readership of over a million users monthly. With a decade's experience in SEO and digital marketing, Damir has contributed to notable platforms like Mashable, Wired, Cointelegraph, The New Yorker, Inside.com, Entrepreneur, BeInCrypto, and more. He navigates life between the UAE, Turkey, Russia, and the CIS as a digital nomad. With a bachelor’s degree in physics, he believes his educational background equips him with the analytical skills needed for success in the dynamic digital landscape.