Artificial Intelligence

Meet OnlyFakes, a quick and easy way to help bypass cryptocurrency Know Your Customer attempts — among other uses for fake IDs.

Lots of interesting stuff in this story, but here’s what stood out to me:

While OnlyFake says it uses “neural networks” to create its fake IDs, 404 Media has not seen evidence that the service uses generative AI tools.

If you want criminals to think you’re on the cutting-edge of criminality, sprinkle some AI on your marketing!

Bloomberg reports that users on the forum have been competing to find ways to bypass keyword filters on AI image-generation tools that prevent sexually explicit materials from being produced. The images moved from 4Chan to Telegram to X where they found viral attention.

Research from social network analysis company Graphika found that Swift isn’t even the community’s most frequently targeted victim, which range from global celebrities to school children.

Axon, the company that provides most police body cameras and stores their recordings, holds over 100 petabytes of footage, according to ProPublica.

Some departments are reportedly turning to AI to find patterns in the footage. Washington State University researchers and Truleo, a company that uses AI for body cam audio analysis, each found officer behaviors that lead to violence. The story quotes Truleo co-founder Anthony Tassone:

“There are certain officers who don’t introduce themselves, they interrupt people, and they don’t give explanations. They just do a lot of command, command, command, command, command,”

An employee of the company was invited to a group video call full of deepfaked company officers, including what appeared to be the company’s CFO, according to the South China Morning Post.

The scammers then gave the employee orders to transfer the money to five separate Hong Kong bank accounts. This sort of fraud will only get worse, as crypto scam ads on YouTube and blackmail using deepfaked nudes have shown.

Hugging Face tech lead Philipp Schmid posted yesterday that users can now create custom chatbots in “two clicks” using Hugging Chat Assistant. Users’ creations are then publicly available.

Schmid directly compares the feature to OpenAI’s GPTs feature, and adds they can use “any available open LLM, like Llama2 or Mixtral.”

The OnePlus 12 launched without big AI features, but now it’s getting them, at least in China, according to a new ColorOS update posted on Reddit (and later reposted on X by Mishaal Rahman). It’s no AI bonanza like other Androids, but OnePlus is clearly moving in that direction, at least in China.

AI-summarized phone calls and articles and AI photo editing are among the features being added in the update.

Android app developer Dylan Roussel leaked an apparent changelog this morning that says “Bard is now Gemini” — the name used for the new model Google put in service last year to compete with OpenAI’s GPT-4.

The log says Google will debut voice chat with Gemini, as well as a new “Ultra 1.0” model with “Gemini Advanced,” a paid plan that offers ChatGPT Plus-like file uploading features.

Of spatial computing, that is. OpenAI brought the iPad version of the ChatGPT app to Vision Pro. Feel free to prompt away while moving around your new semi-virtual environment.

Ambassadors from all 27 countries within the bloc have now endorsed the AI Act text following months of negotiations over how artificial intelligence should be regulated.

However, some hurdles and potential for further lobbying still remain before it comes into force — Euractiv reports that the IMCO Committee is expected to adopt the rules on February 13th, followed by a plenary vote scheduled for April 10-11th.

Billed as the largest silicon manufacturing site in the world, the plant was originally slated to start production in 2025 will now be delayed until late 2026.

Intel planned a groundbreaking at the plant in 2022 but postponed it due to lack of government funding. The Journal quotes an unnamed Intel spokesperson citing “business conditions and market dynamics” for this delay. Intel previously said it depended heavily on the CHIPS Act to finish construction.

Bard can generate images now, but Google might be playing it safe in light of recent events. When I asked it to “make a stop sign that reads ‘don’t stop’,” I got a lecture about safety. But it was perfectly happy to create the image if I asked in a slightly different way.

When Microsoft started its big AI push last year, it launched Copilot tools for Sales and Service to summarize meetings, manage customer lists, and find info for customer service agents, and now they’re more widely available.

Microsoft isn’t the only one applying AI to these tasks — AWS announced a slew of generative AI services for contact centers in December, including transcriptions of audio calls and Q for Amazon Connect, which lets users ask questions about their data.

Reuters reports “Artemis” will complement the hundreds of thousands of Nvidia H100 chips Meta bought. Similar to the MTIA chip Meta announced last year, Artemis is also focused on inference — the actual decision-making part of AI — and not training AI models, but it’s already a little late to the game.

Google introduced a second-generation TPU in 2017 that could do both, and so can Microsoft’s recently-announced Maia 100. And AMD claims its M1300X chip performs better than H100s on the inference side.

If passed, the Artificial Intelligence Environmental Impacts Act of 2024 would compel the EPA to conduct a study on the environmental footprint of AI. It would also make the National Institute of Standards and Technology develop standards for measuring those impacts and set up a voluntary reporting system. AI is energy-intensive, which is why it’s stirred up concerns about how it might impact the grid and contribute to climate change.

If approved, it would make robocalls that use voice cloning technology illegal. The FCC would also provide tools to state attorneys general “to go after bad actors behind these nefarious robocalls.”

Last week, a robocall that appeared to use AI to fake President Joe Biden’s voice made the rounds in New Hampshire — and it told people not to vote in the state’s presidential primary.

9News Melbourne aired the altered image shown below of a member of Victoria’s Parliament, Georgie Purcell.

In a statement, 9News director Hugh Nailon said that when the picture was resized, “automation by Photoshop created an image that was not consistent with the original.” However, Adobe maintains (via NYT and the Guardian) “Any changes to this image would have required human intervention and approval.”

So what do you think happened?

New York lawyer Jae Lee will face an attorney grievance panel after trusting known liar ChatGPT for case research.

The court’s order says Lee filed a “defective brief” citing a made-up case. Lee isn’t alone — others have fallen for the allure of chatbots, including former Trump lawyer Michael Cohen and counsel representing a member of The Fugees.