News
Study shows facial recognition misfires with dark-skinned people
It’s most accurate with white men
Artificial intelligence is only still emerging, but pundits are already decrying how it will soon outsmart mankind. Amid these warnings, AI has yet to overcome a very human prejudice — skin bias.
Recent studies show that AI-powered facial recognition works more accurately with certain skin tones and genders compared to others. MIT researchers Joy Buolamwini and Timnit Gebru recently conducted experiments that tested the accuracy of modern facial recognition.
The experiment compiled 1,270 different facial images from three African countries and three Northern European countries. The difference creates an equal parity between dark-skinned and light-skinned individuals. The test used AI tech from Microsoft and IBM.
The results showed that facial recognition worked better with lighter skin tones rather than darker ones: 11.8 percent error rate for lighter skins versus 19.2 percent for darker skins (the lower the percentage, the better). Further, it shows that the tech performed better with males than females: 8.1 percent for males versus 20.6 percent for females.
As you might expect, males with light skin tones fared way better than females with dark skin tones: 0.3 percent versus a whopping 34.7 percent.
Despite the rigor of their study, the researchers quickly pointed out that there were some AI developers (like Google) who did not want to be included.
Also, they add that accuracy might depend on where the tech was developed. For example, AI recognition developed in Asia might recognize Asian faces more accurately than Caucasian ones.
One study isn’t enough to put down the entire AI industry. The tech is still constantly improving itself. Soon, developers will likely bring error rates down to zero.
However, its current inaccurate state is a warning for those who use AI facial recognition for security and law enforcement.
Right now, facial recognition is most prevalent just in selfie cameras. However, some companies are already adopting the tech to bolster their security by scanning the faces of bad guys. If the current trends continue, your badly shot selfie might incriminate you for a crime you didn’t commit.
SEE ALSO: Chinese woman returned her iPhone X for faulty Face ID
Accessories
Logitech introduces a dedicated shortcut for ChatGPT
Compatible with most keyboards and mice
Artificial intelligence is already meant to simplify a workflow. However, despite the ease, there are still a few ways to optimize the process. Today, Logitech has launched a new shortcut to launch ChatGPT straight from your mouse or keyboard.
Minus programmable hotkeys or buttons, there are hardly any built-in way to easily access an AI model. For a typical workflow, it still consists of manually opening the model on a separate window and keeping it within cursor’s reach throughout the day.
Logitech is simplifying the process by adding a dedicated shortcut for its devices. With a Logitech keyboard or mouse supported by the Logi Options+ app, users can program a keyboard or mouse button for the new Logi AI Prompt Builder.
Once activated, the AI Prompt Builder automatically accesses easy options for highlighted text. For example, users can easily rephrase or summarize the text. Likewise, they can input their own custom queries. By eliminating a few clicks and some typing throughout the day, the new tool hopes to save you time.
While most Logitech keyboards and mice are compatible with the new tool, it is also more easily accessible with the new Logitech Signature AI Edition Mouse. The new mouse has a dedicated button just for AI prompts. If you’re an AI power user, this one is perfect for you.
SEE ALSO: Logitech unveils G Pro X 60 gaming keyboard: Price, details
Despite the ease that the technology promises, artificial intelligence still isn’t the most reliable thing you can depend on. Today’s models can still flub answers. Now, the pitfalls of AI are abundantly clear with a few mishaps on X. Golden State Warrior Klay Thompson was just “accused” of a vandalism spree involving bricks.
Late Tuesday night, X presented an AI-generated news snippet. The article, posted soon after the Warriors’ play-in loss to the Sacramento Kings, reads: “Klay Thompson Accused in Bizarre Brick-Vandalism Spree.”
The snippet hilariously described an incident where Thompson supposedly vandalized houses in Sacramento with bricks. Thankfully, according to the snippet, no one was hurt. It even had a few sources for its claims below the news summary.
First off… I am ok.
My house was vandalized by bricks 🧱
After my hands stopped shaking, I managed to call the Sheriff…They were quick to respond🚨
My window was gone and the police asked if I knew who did it👮♂️
I said yes, it was Klay Thompson
— LakeShowYo (@LakeShowYo) April 17, 2024
The sources, however, are just joke posts stemming from the recently concluded game. In that do-or-die game, Thompson put up a staggering 0 points on ten shot attempts, the most misses without a make by a Warrior since the ’68 season. Naturally, everyone joked that the Warriors guard just threw bricks all over Sacramento.
Now, X is currently experimenting with a new AI feature called Grok. The model collates trending topics and creates snippets of what’s happening for X users. However, it’s not exactly the smartest in determining real news from satiric ones. The feature notes as much, carrying a fine-print caveat warning users to “verify its outputs” because it’s an early feature and can make mistakes.
SEE ALSO: New X users must pay a dollar per year to post and reply
New users on X might soon face a tough time on the platform. The social media website will likely start charging new accounts a small fee for the right to post on the platform.
Now, the fee isn’t a new one. Almost six months ago, the company tested the paid system in New Zealand and the Philippines. New users in those countries had to pay a dollar per year for the ability to post and reply to content.
As spotted by X Daily News on the same platform, the company might be ready to take the experiment to a larger market. New text strings have shown that the policy is rolling out worldwide.
SPECULATION: X might be expanding its policy to charge new users before they reply/like/bookmark a post https://t.co/odqeyeiHBx pic.twitter.com/EU71qlwQ0D
— X Daily News (@xDaily) April 15, 2024
The policy is designed to combat a wave of bots appearing on the platform. By preventing new accounts from creating posts, X hopes to stave off the standard behavior of bots these days. You might have noticed them as OnlyFans creators in unrelated posts, peddling NSFW content on their bio.
Though the global rollout was only just spotted, owner Elon Musk has seemingly confirmed the change. Replying to X Daily News, Musk says that it is “the only way to curb the relentless onslaught of bots.” He says that the current breed of bots can easily bypass simple checks these days.
-
Accessories2 weeks ago
Apple Vision Pro Review: Two Months Later
-
Features5 days ago
Fortify your home office or business setup with these devices
-
Gaming1 week ago
The Rogue Prince of Persia looks like an ultra-colorful roguelite
-
Events1 week ago
Stellar Blade: PlayStation taps cosplayers to play Eve for game’s launch
-
Gaming1 week ago
Star Wars Outlaws release date revealed
-
Accessories1 week ago
Logitech unveils G Pro X 60 gaming keyboard: Price, details
-
Philippines2 weeks ago
vivo Y100 to release in Philippines on April 27
-
Deals2 weeks ago
Samsung Awesome April: Deals on Galaxy A series