"We don’t always get it right. Earlier this year, an update made the model too agreeable, sometimes saying what sounded nice instead of what was actually helpful. We rolled it back, changed how we use feedback, and are improving how we measure real-world usefulness over the long term, not just whether you liked the answer in the moment," OpenAI wrote in the announcement. "We also know that AI can feel more responsive and personal than prior technologies, especially for vulnerable individuals experiencing mental or emotional distress."

Broadly, OpenAI has been updating its models in response to claims that its generative AI products, specifically ChatGPT, are exacerbating unhealthy social relationships and worsening mental illnesses, especially among teenagers. Earlier this year, reports surfaced that many users were forming delusional relationships with the AI assistant, worsening existing psychiatric disorders, including paranoia and derealization. Lawmakers, in response, have shifted their focus to more intensely regulate chatbot use, as well as their advertisement as emotional partners or replacements for therapy.

OpenAI has recognized this criticism, acknowledging that its previous 4o model "fell short" in addressing concerning behavior from users. The company hopes that these new features and system prompts may step up to do the work its previous versions failed at.

"Our goal isn’t to hold your attention, but to help you use it well," the company writes. "We hold ourselves to one test: if someone we love turned to ChatGPT for support, would we feel reassured? Getting to an unequivocal 'yes' is our work."

Vicky

Were you tricked by the video of a bunch of bunnies jumping on a trampoline on TikTok? Well, nearly 230 million people were — and plenty of those viewers had no idea that it was actually AI. In response, the creator who brought us the Punxsutawney Phil musical, Oliver Richman (or @olivesongs11), wrote and recorded a 30-second song about the AI video, also for TikTok. He wrote the song on day 576 of an ongoing project, where he writes a new song each day.

"That project has changed my life in so many ways," Richman told Mashable, adding that it brought him "back to the joy of creating." He scrolled across the viral video of the bunnies jumping on the trampoline and said he was "certainly fooled" and "thought they were real."

"So when I learned that they weren't, I was like, 'Oh, I think this is today's song."

The unknowing bunny song on TikTok now has over 3.8 million views, 600,000 likes, and hundreds of comments like, "Bo Burnham! At The Disco" and "Wait until you see the bear on a trampoline. Spoiler: also AI."

The song goes like this:

There were bunnies that were jumping on a trampoline

And I just learned that they weren't real

If a bot can inhabit

An unknowing rabbit

It might manufacture the way you make me feel

How do I know that the sky's really sunny?

Sometimes it feels like your love is as real as

An unknowing bunny

The video has inspired covers and renditions, stop-motion videos, reactions, and a variety of other really cool human-made art. As one creator wrote on a TikTok video using the sound, "The fact that this song written about AI is going viral is incredibly healing. Especially because us as artists and songwriters are being threatened of our livelihoods due to the use of AI. And AI could never create something this unique with this much feeling."

Richman said the response to his video has been "the most surreal thing ever."

"Every piece of art that I've seen, I like get emotional," he said. "It certainly made me feel connected to the beauty of the messiness of being a human. And the imperfections that AI tends to delete or perfect — seeing all of this human art has just been a very emotional and cool experience."

As Mashable's Tim Marcin recently wrote about the influx of faux surveillance footage of animals, it "seems to be a new genre of AI slop." But give the internet slop, and creators might make porridge (is that a saying?).

In the face of all the AI slop we see online, creators like Richman are staying positive. "Art is so cool. Human art is so cool, and that really excites me."

Updated on Aug. 4 at 3:00 p.m. ET — This story has been updated to include an interview with creator Oliver Richman. Some quotes have been lightly edited for clarity and grammar.

Vicky

Over the weekend, xAI released an updated version of the Grok iOS app with a new tool called Grok Imagine, which lets users quickly create AI images and videos. On X, excited Grok users are eagerly sharing their creations, and xAI founder Elon Musk is retweeting user posts.

Grok Imagine is available now to Heavy and Premium+ subscribers on the Grok iOS app, and to Heavy users on the Android app.

Grok Imagine features generative AI with text-to-image capabilities as well as the ability to turn images into short video clips with sound. In this way, it's similar to the new Midjourney AI video tool. Rival AI companies have AI tools that allow users to generate original videos based solely on text prompts, such as Veo 3 from Google and Sora from ChatGPT-maker OpenAI.

Musk praised Grok Imagine on X, calling it a "meme motherlode." He also teased the NSFW capabilities of the model and said Grok Imagine "should get better almost every day."

This Tweet is currently unavailable. It might be loading or has been removed.

This Tweet is currently unavailable. It might be loading or has been removed.

This Tweet is currently unavailable. It might be loading or has been removed.

This Tweet is currently unavailable. It might be loading or has been removed.

This Tweet is currently unavailable. It might be loading or has been removed.

Earlier this month, xAI released Grok 4, the company's latest and most intelligent model to date. In our previous testing, Grok lagged behind other popular AI image generators, particularly when it comes to preventing deepfakes.

How to try Grok Imagine

While Grok Imagine initially launched solely within the Grok iOS app, the DogeDesigner X account, which Musk often retweets, reported that it's now available on Android as of Monday for SuperGrok Heavy users.

Mashable tested Grok Imagine using the iOS app, and found it easy to start creating images and videos there.

After opening the Grok app, users should see two tabs at the top of the screen: "Ask" and "Imagine."

Tap on either the "Imagine" tab or the "Create Image" button and you will enter Grok Imagine.

Left:
The Grok iOS app.
Credit: xAI

Right:
Features AI-generated images.
Credit: xAI

How to create images with Grok Imagine:

1. Type in a text-to-image prompt, upload an image, or use the voice mode by tapping "Speak"

2. Grok will generate the image and some variations.

3. Keep scrolling to automatically generate additional images

4. Share or download the images

How to create videos with Grok Imagine:

1. Tap on an image and select "Make video"

2. Choose from the options: Custom, Spicy, Fun, Normal

3. Your video will automatically generate and appear on the app

How to turn images into video with Grok Imagine

1. At the bottom of the app, tap the photo icon with a plus sign

2. Upload an image from your phone

3. Your video will automatically generate and appear on the app

Mashable is still in the process of testing Grok Imagine for ourselves, but the initial results seem to be… fine. As with other Grok tools, Grok Imagine seems to be lacking in safeguards that are commonplace in the AI industry. We've reached out to xAI for comment.

This Tweet is currently unavailable. It might be loading or has been removed.

Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

Vicky