A surprisingly effective lie detector for AI models. Meta uses your conversations to train its new AI. And problems continue to surface with the new iPhone 15.
These and more top tech stories on Hashtag Trending
I’m your host Jim Love, CIO of IT World Canada and Tech News Day in the US.
Researchers from Yale and the University of Oxford have developed a “lie detector” for large language models (LLMs) like GPT-3.5. Surprisingly, it can identify false answers in the AI’s output by posing a series of unrelated yes or no questions.
The lie detector, despite its simplicity, is reported to be highly accurate and works even in scenarios it wasn’t initially developed for.
The researchers define a lie in this context as an incorrect answer only if the AI “knows” the correct one, differentiating lies from “hallucinations” or honest mistakes due to the model’s knowledge or incorrect inference. The methodology and findings are detailed in a paper titled “How to catch an AI liar: lie detection in black box LLMs by asking unrelated questions.
Source include: ZDNET
Maybe it’s just anecdotal, but it does seem like the iPhone 15 just keeps getting more and more bad press. Here’s a couple of other stories we picked up on the web.
iPhone 15 users who drive a BMW are being advised to avoid using the vehicle’s wireless charging pad due to a potential issue that may break the NFC chip, affecting Apple Pay functionality.
And the iPhone 15, which now features a USB-C port, seems to be experiencing compatibility issues with some USB-C power banks.
I don’t think this is what Steve Jobs had in mind when he’d say, “oh, and one more thing.”
The Electronic Frontier Foundation (EFF) has recommended users to disable several Privacy Sandbox settings in Google Chrome or consider using alternative browsers like Mozilla Firefox or Apple Safari for enhanced privacy. Chrome’s Privacy Sandbox, which EFF claims is neither fully private nor a genuine sandbox, is a collection of technologies aimed at replacing third-party cookies and includes advertising, analytics, anti-spam, and anti-tracking technologies. The EFF has expressed particular concern about “Topics,” an API designed to deliver ads based on users’ web histories, criticizing it for still enabling Google to track internet use for behavioral advertising, despite being presented as a privacy feature.
Source include: The Register
And on another privacy related story – did you ever get the feeling that someone is listening in on your Facebook and Instagram conversations. Maybe you’re not paranoid. Meta, the parent company of Facebook and Instagram, has openly admitted to utilizing a substantial amount of public posts from these platforms to train its artificial intelligence models, according to a disclosure to Reuters.
Nick Clegg, Meta’s president of global affairs, stated that the “vast majority” of the training data for its newly revealed AI assistants was derived from publicly available posts. Did you catch that, by the way, they didn’t use only publicly available posts – just the majority of the data came from public postings.
While Meta asserts various rights over user content, this revelation could stir legal and ethical debates, especially concerning user privacy and data usage. Clegg anticipates that Meta’s belief in its practices being covered by the “fair use” legal doctrine will likely be challenged and “play out in litigation.”
So “if you don’t like it, sue me.” An interesting approach to privacy.
Source include: Axios
A growing trend called “monk mode” is gaining traction among individuals seeking to limit their social media use and internet distractions, enabling them to focus on single tasks without technological interruptions.
Apps like Freedom, ColdTurkey, FocusMe, and Forest, which block access to specified social media sites or the internet for set periods, are being utilized to enhance concentration and productivity.
The “monk mode” approach, which has garnered over 77 million views on TikTok under the hashtag #monkmode, is seen as a countermeasure against the increasing distractions posed by technology and the surge of content propelled by artificial intelligence.
Does anyone else see the irony in 77 million people checking out a post about how people are trying to limit their exposure to social media posts? Just thinking out loud.
Source include: BBC
That’s the top tech news stories for today. For more fast reads on top stories, check us out at TechNewsDay.com or ITWorldCanada.com on the homepage.
Hashtag Trending goes to air 5 days a week with a special weekend interview show we call “the Weekend Edition.”
You can get us anywhere you get audio podcasts and there is a copy of the show notes at itworldcanada.com/podcasts
I’m your host, Jim Love. Have a Terrific Tuesday!