Posts tagged with "OpenAI"

Pope prank: Fake photos of the pontiff in a puffer jacket go viral, conveying the power and peril of AI

March 29, 2023

It was a cold wind that blew through St. Peters Square at the Vatican over the weekend; but that didn’t deter Pope Francis from taking a stroll outside to greet the faithful, as he often does. When images appeared online showing the 86-year-old pontiff dressed to fight the elements in a stylish white puffer jacket and silver bejewelled crucifix, they soon went viral—racking up millions of views on social media platforms, reports CBS News.

The picture, first published Friday, March 24, on Reddit along with several others, was, in fact, a fake. It was an artificial intelligence rendering generated using the AI software Midjourney.

While there are some inconsistencies in the final rendered images—for example, the pope’s left hand, which is holding a water bottle, looks distorted and his skin has an overly sharp appearance—many people online were fooled into thinking they were real pictures.

Some Twitter users were shocked and confused. “I thought the pope’s puffer jacket was real and didn’t give it a second thought,” tweeted model and author Chrissy Teigen. “No way am I surviving the future of technology.”

The “pope in the puffer jacket” was just the latest in a series of “deepfake” images created with AI software. Another recent example: pictures of former President Donald Trump that appeared to show him in police custody. Although the creator made it clear that they were produced as an exercise in the use of AI, the images, combined with rumors of Trump’s imminent arrest, went viral and created and entirely fraudulent but potentially dangerous narrative.

Midjourney, DALL E2, OpenAI, and Dream Studio are among the software options available to anyone wishing to produce photo-realistic images using nothing more than text prompts—no specialist training required.

As this type of software becomes more widespread, AI developers are working on better ways to inform viewers of the authenticity, or otherwise, of images.

CBS News’ Sunday Morning  reported earlier this year that Microsoft’s Chief Scientific Officer Eric Horvitz, the co-creator of the spam email filter, was among those trying to crack the conundrum—predicting that, if technology isn’t developed to enable people to easily detect fakes within a decade or so, “most of what people will be seeing, or quite a lot of it, will be synthetic. We won’t be able to tell the difference.”

In the meantime, Henry Ajder, who presents a BBC radio series entitled The Future Will be Synthesised, cautioned in a newspaper interview that it was “already very, very hard to determine whether” some of the images being created were real.

“It gives us a sense of how bad actors, agents spreading disinformation, could weaponize these tools,” Ajder told the British newspaper, i.

There’s clear evidence that this is happening already. Last March, video emerged appearing to show Ukrainian President Volodymyr Zelenskyy telling his troops to lay down their arms and surrender. It was bad quality and quickly outed as a fake, but it may have been merely an opening salvo in a new information war.

So, while a picture may speak a thousand words, it may be worth asking who’s actually doing the talking.

Research contact: @CBSNews

Meet Bard, Google’s answer to ChatGPT

February 8, 2023

Nothing has made the tech industry cower in quite the way that ChatGPT has. The chatbot, which was launched by San Francisco-based OpenAI last November, already has attracted a multibillion-dollar investment from Microsoft.

Microsoft reputedly invested in OpenAI in order to super-charge its search engine, Bing, with ChatGPT, which could have a widespread release this spring.

Concurrently, other major tech players—Google, in particular—are attempting to compete. The search giant has just announced its answer to the wildly popular chatbot, and it’s called BardAI, reports Gizmodo.

BardAI is Google’s own experimental chatbot that is built with the company’s Language Model for Dialogue Applications, or LaMDA. LaMDA is the same AI engine that an ex-Google engineer warned us was sentient, but the company hopes that LaMDA is powerful enough that it will make Bard Google’s rival to ChatGPT.

Bard is currently only available to testers, but Google says that BardAI is using a lightweight version of LaMDA, so that it can scale easily after the trial period and reach more users.

“Bard seeks to combine the breadth of the world’s knowledge with the power, intelligence and creativity of our large language models. It draws on information from the web to provide fresh, high-quality responses,” Google CEO Sundar Pichai said in a company blog post, adding, “Bard can be an outlet for creativity, and a launchpad for curiosity, helping you to explain new discoveries from NASA’s James Webb Space Telescope to a nine-year-old, or learn more about the best strikers in football right now, and then get drills to build your skills.”

Google has also added some AI capability to its basic search engine function, because, as Pichai puts it, “people are turning to Google for deeper insights and understanding.” In other words, Google wants its search engine to provide quicker answers to deeper, potentially multi-part questions. The company also indicated that the search engine’s interface may change slightly too, in order to feed users more in-depth answers to questions in an easier way.

ChatGPT has taken the world by storm since its release to the general public late last year. As Big Tech has taken notice, Google appears to be the first to release its own version of the chatbot while others, like Microsoft, have decided to hop on the bandwagon.

Research contact: @Gizmodo