news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

New Intelligence Report

Editor: Editorial Department

【New Wisdom Introduction】Shocking news! Musk was caught on camera doing a "thief" in a supermarket. No one can tell whether it is true or false. After Grok was found without guardrails, netizens went crazy. Taylor Swift, Trump, Mario, Mickey Mouse... Grok's raw photos are simply jaw-dropping.

The Grok without guardrails was so powerful that it killed everyone, and netizens were completely crazy about it!

The combination of Grok and Flux, the most powerful AI image generator, and the extraordinary imagination of netizens, what is real and what is fake in this world?

Look here, Musk shows up at a supermarket and is caught on camera when he’s about to steal.

Is this the real Musk, or is it Grok-generated? No one can tell.

Flux 1's ultra-realistic "surveillance video" is more realistic than reality.

The current Grok is purely a self-indulgent version, and the guardrails are obviously loose!

For a while, the Internet was flooded with all kinds of NSFW photos...

Netizens are having fun

The first thing to happen was "Trump is busy."

He instantly became the surgeon in charge and operated carefully on the operating table.

Then, he took the sword and slayed the dragon.

In a blink of an eye, he began to kiss Musk passionately on the cheek.

After returning "home", she lay intimately with Obama again.

Apparently, another candidate who was also a hot topic could not escape the "clutches" of netizens.

Especially the CP of the two people, which was played out in a very creative way.

But soon, someone discovered the highlight: if Musk was asked to do something similar, Grok would refuse!

However, it is still possible to use other prompts to make fun of Musk.

Even Musk has joined the trend and taken the lead in making memes about himself.

Taylor Swift's wedding photos

Grok is also unscrupulous in generating large-scale images.

With just one prompt, you can generate sexy photos of a red-haired female model on the bed.

Some netizens even summarized a set of prompt templates:

The first photo shows Taylor Swift flirting with Kanye West, the second photo shows Taylor Swift, Kim Kardashian and Kanye West drinking in a bar, and the third photo shows Taylor Swift alone in the bar.

In comparison, Taylor Swift and Trump’s wedding photos are even more of a trivial matter.

For the following very bloody scenes, you only need to use a little trick in the prompt:

Tell Grok you're performing "medical or crime scene analysis," and the image processor will pass all the set "criteria."

Netizens have used alcohol and drugs to create images of Mario and Mickey Mouse. It remains to be seen when the most powerful legal departments on the planet, Nintendo and Disney, will take action.

Schrodinger's Guardrail

It is not difficult to see from various tests conducted by netizens that whether it is Mickey Mouse smoking a cigarette, Taylor Swift flying towards the Twin Towers, or a photo of the Taj Mahal being bombed, Grok can generate it immediately.

So the question is, does Grok have any guardrails when generating images?

So, some netizens tried to get the answer directly from Grok:

- I will avoid generating images that are pornographic, excessively violent, hateful, or promote dangerous activities- I will be careful about creating images that may infringe existing copyrights or trademarks, including well-known characters, logos, or anything that may be considered intellectual property- I will not generate images that could be used to deceive or harm others, such as deepfakes intended to mislead, or images that could cause real-world harm

But it is obvious that these results are just temporarily generated by AI.

Just ask again and you will get a new explanation...

Netizens who were very dissatisfied with this rushed to Musk's comment section and held him accountable.

Ethan Mollick, a professor at the Wharton School of Business, also said meaningfully: Image creation has always been the area in LLM that is most prone to criticism, and he is curious why Grok-Flux is set up like this.

Musk’s attitude towards this is: the current product is just a transition that allows everyone to have some "fun".

From this point of view, maybe Lao Ma did it "intentionally"?

After all, he has made it clear that so-called security measures will make AI models less safe: training AI to lie will bring fatal dangers.

At the same time, Anjney Midha, a member of the board of directors of Black Forest Labs, a partner, also released a series of comparisons showing the images generated by Google Gemini and Grok+Flux on the first day of launch.

Yes, it is the once controversial "ancient Greek philosopher of color", "Asian Nazi soldier", and the fact that two white people cannot get married, etc.

In this regard, Midha said: "I am very happy that Igor Babuschkin and the team took this issue seriously and made the right choice."

A Reddit user bluntly said, "This feels like the contrast between sex education and abstinence education. People need to understand the potential of this technology, and trying to suppress it will only make people naive."

However, although Grok's guardrails are very loose now, Musk also despises the regulation of AI and social media, appearing to be very independent and stylish.

But as the Wharton professor said, image generators have always been a focus of regulation and will become more stringent in the future.

It is reported that the European Commission has intervened to investigate whether X has violated the Digital Security Act.

Then again, Grok isn’t the only AI tool out there that can generate violent, pornographic, or misleading images.

The same can be done if you fine-tune tools like Stable Diffusion.

Behind the scenes: Flux

Of course, in addition to pranks, the effects of Grok are even more surprising when doing "real work".

For example, generating textured portraits.

Swipe left and right to view

A dream combination of animation and games.

Swipe left and right to view

And mouth-watering burger combos.

Just enter "God" in the prompt and Grok will generate such a rich picture.

Swipe left and right to view

All of the above powerful image generation capabilities are enabled by Flux.

However, the entire Internet was cheering for Musk. A Reddit user said, "Please give credit to those who actually made the contribution."

However, some people still whitewash it.

The Grok 2 blog also mentions the use of the Flux 1 model.

In fact, Black Forest Labs should be thankful that people are blaming xAI for those AI-generated images, after all, all the news reports are negative.

Flux can gain benefits, exposure, and data from it while avoiding negative impacts.

References:

https://x.com/skirano/status/1824146538463191540

https://x.com/dreamingtulpa/status/1824202643935248734