All of the internet now belongs to Google’s AI

By Fionna Agomuoh Published July 6, 2023

Google Bard being shown off at Google I/O 2023. — Image used with permission by copyright holder

Google’s latest update to its privacy policy will make it so that the company has free range to scrape the web for any content that can benefit building and improving its AI tools.

“Google uses information to improve our services and to develop new products, features, and technologies that benefit our users and the public,” the new Google policy says. “For example, we use publicly available information to help train Google’s AI models and build products and features like Google Translate, Bard, and Cloud AI capabilities.”

Recommended Videos

Gizmodo notes that the policy has been updated to say “AI models” when it previously said, “for language models.” Additionally, the policy added Bard and Cloud AI, when it previously only mentioned Google Translate, for which it collected data.

The privacy policy, which was updated over the weekend, appears especially ominous because it indicates that any information you produce online is up for grabs for Google to use for training its AI models.

The aforementioned wording seems to describe not just those in the Google ecosystem in one way or another but is detailed in such a way that the brand could have access to information from any part of the web.

Major issues surrounding the mass development of artificial intelligence are questions about privacy, plagiarism, and whether AI can dispel correct information. Early versions of chatbots such as ChatGPT are based on large language models (LLMs) that used already public sources, such as the common crawl web archive, WebText2, Books1, Books2, and Wikipedia as training data.

Early ChatGPT was infamous for becoming stuck on information beyond 2021 and subsequently filling in responses with false data. This could likely be one of the reasons Google would want unfettered access to web data to benefit tools such as Bard, to have real-world and potentially real-time training for its AI models.

Gizmodo also noted that Google could use this new policy to collect old, but still human-generated content, such as long-forgotten reviews or blog posts, to still have a feel of how human text and speech is developed and distributed. Still, it remains to be seen exactly how Google will use the data it collects.

Several social media platforms, including Twitter and Reddit, which are major sources of up-to-date information have already limited their public access in the wake of AI chatbot popularity, to the chagrin of their entire communities.

Both platforms have closed free access to their APIs, which restricts users from downloading massive amounts of posts for sharing elsewhere, under the guise of protecting their intellectual property. This instead broke many of the third-party tools that make both Twitter and Reddit run smoothly.

Both Twitter and Reddit have had to deal with other setbacks and controversies as their owners’ concerns heighten about AI taking over.

Computing Writer

Fionna Agomuoh is a Computing Writer at Digital Trends. She covers a range of topics in the computing space, including…

Topics

Computing

I used ASUS’ dual-screen laptop as a portable creative station, and my desk PC started collecting dust

The Zenbook Duo might be the creator setup I wanted in college

Computer Hardware, Electronics, Hardware

With laptops, brands are constantly in a balancing act between portability and workspace productivity. The ASUS Zenbook Duo UX8407AA tries to dodge that choice with a design that brings a whole setup in a compact form factor.

I used the Zenbook Duo as a creative machine, mainly with design apps, illustration work, writing, and multitasking. The model I tried runs on Intel’s Core Ultra 7 355, paired with 32GB of memory and a 1TB SSD. That gives it enough horsepower to handle Photoshop and Animate, for sketches and animations, and a lot more without breaking a sweat.

Computing

macOS clipboard app Maccy has a fake out there stealing passwords

PamStealer malware is disguising itself as Maccy to target Mac users

Depicting of the Maccy clipboard app for macOS on a laptop with letters inb the background.

A fake version of Maccy, a popular clipboard manager for macOS, is being used to deliver a newly discovered Mac malware strain called PamStealer. Researchers at Jamf say the malware impersonates the real open-source app, but its actual purpose is to steal data and capture a victim’s login password.

PamStealer arrives as a disk image containing an AppleScript file that impersonates Maccy. Once the user opens that file, macOS launches it in Script Editor, where the on-screen instructions tell them to press Command-R. To someone expecting a normal app installer, that may look like an odd setup step. In reality, that action runs hidden malware code and starts the attack.

Computing

A new technology teaching drones to feel pain could stop your self-driving car from harming itself

Drones first, autonomous cars next. A pain-sensing system that detects failure before it happens has real stakes for self-driving vehicles.

Transportation, Vehicle, Car

When you sprain your ankle in the middle of a run, your body sends a pain signal to your brain, forcing you to stop. Essentially, the ability to sense pain stops you from pushing through the injury and causing further self-harm.

Researchers at Delft University of Technology and Wageningen University have applied this exact concept to drones, giving them a digital equivalent of a nervous system that recognizes a faulty part and triggers a pain-like warning signal. What's even more interesting is that the technology could find use in self-driving cars.