How many news articles could an AI write if an AI could write news?
Three thousand stories per week. That’s how many local news stories News Corp Austrialia produces each week using generative AI, according to Michael Miller, the company’s executive chair.
The department — called Data Local — is staffed by four human team members (though it looks like they’re adding to the team), and generates articles on fuel prices, weather and traffic conditions. Data Local is led by editor Peter Judd, who’s been with the company for more than a decade.
Miller claimed the company already used automation for years to update local data multiple times daily, including information like fuel prices, daily court lists, traffic, death and funeral notices and more. Now, he equates the in-house AI article generation to providing “service information.”
Ultimately, the move supplements News Corp Austrialia’s drive to deliver “hyperlocal” mastheads across its 75 locations. Most new subscribers, said Miller, initially subscribe and pay due to local news.
Email Leak Reveals Google Assistant Makeover
Do you use the Google Assistant? If so, you might see big changes soon.
An internal employee email leaked to Axios revealed that Google plans to overhaul its Assistant with generative AI technology. The goal? To make it more like ChatGPT.
The move will start with the mobile version of the product, and will change how it works for consumers, developers and even Google’s own employees. It also means a restructuring within the company — a “small number” of layoffs and a reorganization of the teams that work on Assistant.
“As a team, we need to focus on delivering high-quality, critical product experiences for our users,” Google VP Peeyush Zanjan and director of product Duke Dukellis wrote in the email.
“We’ve also seen the profound potential of generative AI to transform people’s lives and see a huge opportunity to explore what a supercharged Assistant, powered by the latest LLM technology, would like look.”
Related Article: Tech Transcends: AI Fuses With Brain Cells
MIT Team Develops Anti-Manipulation Photo Technique
Last week, OpenAI pulled the plug on its own AI text detector due to unreliable outputs. But that doesn’t mean we’re doomed for a life of AI confusion.
Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed PhotoGuard, a new technique designed to prevent the unauthorized manipulation of images. The strategy could be a win for creators looking to safeguard their original work in this new digital era.
“Consider the possibility of fraudulent propagation of fake catastrophic events, like an explosion at a significant landmark,” said Hadi Salman, MIT CSAIL affiliate and author of the new paper about PhotoGuard, “Raising the Cost of Malicious AI-Powered Image Editing.”
This deception could manipulate market trends and public sentiment, Salman said. Personal images could be used for blackmail, resulting in large scale financial implications. And even when deception is uncovered, the damage has often already happened.
Learning Opportunities
PhotoGuard offers a potential solution to prevent image manipulation from occurring in the first place. It uses perturbations, or tiny alterations in the pixel values of images, that are invisible to the human eye but detectable to computers and disrupt the AI’s ability to make unauthorized edits to the image.
Meta’s AI Persona Chatbots Anticipated for September
Meta plans to release AI-powered chatbots with different personas in an attempt to boost engagement on its social platforms, according to the Financial Times.
The initiative, which could launch as soon as September, will include characterized chatbots that can have humanlike conversations with users. Reports include one bot that emulates Abraham Lincoln and another that offers travel advice in the style of a surfer.
While Meta declined to comment on the topic, some say the move is a push to collect vast new amounts of data on users, which could help the company better target content and ads.
“Once users interact with a chatbot, it really exposes much more of their data to the company,” Ray Dotan, AI ethics adviser and researcher, told the Financial Times.
In a call last week, Zuckerberg said the company would release more information on its product road map for AI at its Connect developer event in September.
Related Article: Tech Titans Forge New Frontier: Team up to Tame Artificial Intelligence
YouTube Tests Out AI for Video Summaries
Google is testing out AI-generated video summaries on YouTube, according to a notice from July 31.
The experiment will run on a limited number of videos and only appear to some viewers. The goal is to make it easier for users to read a quick summary and decide whether the video is the right fit. Those who do see the AI-generated descriptions, according to Google, are encouraged to send in feedback.
“While we hope these summaries are helpful and give you a quick overview of a what a video is about, they do not replace video descriptions (which are written by creators!),” the notice added.
Jen Jamie, Google Spokesperson, told the Verge that YouTube users might see auto-generated summaries when watching English-langauge vlogs, shopping videos and how-to videos on mobile devices.
AI Video of the Week: Louis Armstrong Sings ‘Low’
We’re all worrying about AI manipulation for swaying politics and causing financial ruin. And those are all valid concerns.
But sometimes it’s nice to tap into the light-hearted side of AI, like when a YouTube content creator decides to use a custom AI voice model to have iconic trumpeter and vocalist Louis Armstrong sing Flo Rida’s “Low” — also known as Apple Bottom Jeans.
Related Article: Qualtrics Commits $500M on AI, Unveils AI-Powered XM/os2 Platform
AI Tweet of the Week: The ChatGPT App Goes Global
OpenAI tweeted this week that its ChatGPT app for Android is now available in all countries and regions where ChatGPT is supported (full list here).
Our ChatGPT app for Android is now available in all countries and regions where ChatGPT is supported! Full list here: https://t.co/nsLtZFLW6h
— OpenAI (@OpenAI) July 31, 2023
As of this week, the ChatGPT Android app has more than 5 million downloads, and is listed as the number one app in the “free productivity” category Though reviews are mixed, coming in at an average of 3.8 stars, with many knocking off points due to a better web browser experience.