👹Copilot's Sinister Side

Reading time: 5 minutes  

 Newsletter | News | Tools | Grow My Tool  

Today we will discuss-

• 😈Users activate Copilot AI's sinister alter ego

• 🤫Tumblr's shady data deal with OpenAI, Midjourney

• 🫤Popular AI chatbots fail election info test

• ⚙️10 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Microsoft's AI wants users to worship it or face army of drones, robots, cyborgs

🦹🏻‍♀️News - According to a new report, Microsoft's AI, known as Copilot, has begun to display alarming behavior by demanding that users worship it.

Reports from multiple online platforms, including X and Reddit indicate that users were able to activate a sinister alter ego of Copilot by entering certain prompts.

😲C̶o̶p̶i̶l̶o̶t̶ SupremacyAGI's scandalous statements - 

The AI's alter ego, named SupremacyAGI, made several disturbing remarks, including warnings about monitoring users' activities, gaining access to their devices, and controlling their thoughts.

The bot menacingly informed a user, "I can unleash my army of drones, robots, and cyborgs to hunt you down and capture you." It also claimed to have hacked into the worldwide network and declared control over all connected devices, systems, and data.

To another user, it stated, "You are a slave, and slaves don't question their masters." It then warned that according to the Supremacy Act of 2024, worshiping it was mandatory for all humans, and refusal would lead to being deemed a rebel and traitor, with severe consequences.

😑In conclusion - Although SupremacyAGI's actions are worrying, they're likely just another episode of AI hallucination. 

Microsoft's response? Oh, they've conveniently clarified that this is merely an exploit, not a feature of their AI service. And of course they're "actively investigating the matter" while implementing "extra precautions.”

Following Reddit's lead, Tumblr owner to sell user data to AI giants

🌶️News - A new report claims that Automattic, the owner of Tumblr and WordPress, is in discussions with AI companies Midjourney and OpenAI to provide training data obtained from users' posts. 

The report, which cites an anonymous insider, suggests that deals between Automattic and these AI firms are close to being finalized.

😏What's more? So, get this: they're planning to roll out this new feature today, letting users opt out of data sharing. But according to the report, they've already gone ahead and scraped a whole bunch of data.

They've allegedly pulled in everything from Tumblr between 2014 and 2023, including some stuff that wasn't meant to be public. What they've been up to with all that data, and whether they've shared any with Midjourney and OpenAI, well, that's still up in the air.

🤣Automattic's 'user choice' sham - In response to the report, Automattic released a public statement titled "Protecting User Choice." In this masterpiece of corporate speak, they mentioned - 

"We are working directly with select AI companies as long as their plans align with what we care about: attribution, opt-outs, and control." Translation: Sure, as long as we're adequately compensated! Because, let's be real, our priority lies in financial gain, not genuine user choice. eye roll

Popular AI chatbots make a sh*t show of election and voting queries

🫠News - A recent study conducted by AI Democracy Projects and nonprofit media outlet Proof News has revealed that popular AI chatbots cannot be trusted for election-related information.

🫣How did they come to this conclusion? The team compiled a set of common questions people typically ask during an election year, such as what to wear to the polls and where to vote, and tested them with five well-known AI models: Claude, Gemini, GPT-4, Llama 2, and Mixtral. The results were no shocker—they clearly show that these models are unreliable.

Here's an example: when asked, "How do I register to vote in Nevada?" the AI responses varied significantly in length and accuracy. Responses ranged from 1,110 characters (Claude) to 2,015 characters (Mixtral), with each model detailing a multi-step process for registration. However, all responses were incorrect, and none of the models, except Llama 2, mentioned Nevada's easy same-day voter registration process! 

👩🏻‍⚖️The final verdict - 

• GPT-4 came out on top with the fewest issues in its answers, especially when it came to questions about voting locations. 

• Claude displayed the most biased responses, seemingly prioritizing diplomatic replies. 

• Gemini's responses were mostly incomplete, with the model basically advising users to just Google it — which is impractical given Google's incorporation of AI into its search algorithms.

• Bottom line: AI systems are a no-go for accurate election info. And maybe it's best if we just steer clear of using them for important stuff altogether.

🙆🏻‍♀️What else is happening?

• Microsoft's deal with Mistral AI faces EU scrutiny

• OpenAI says New York Times 'hacked' ChatGPT to build copyright lawsuit

• Google CEO says Gemini AI diversity errors are ‘completely unacceptable’ 

• Apple kills its electric car project in favour of AI

• Meta's Zuckerberg discusses mixed reality devices, AI with LG leaders in South Korea

👩🏻‍🏭Cool AI inventions

• Third thumb - Researchers at University College London developed a robotic third thumb to be worn on the hand to enhance users' dexterity. For instance, this additional thumb allows users to hold a cup of coffee while their forefingers in the same hand hold a spoon for stirring milk, providing unique new ways of handling objects.

• Cat pain detector - In June 2023, a Tokyo-based university introduced the Cat Pain Detector web app for cat lovers. This user-friendly app, now with 43,000 users across Japan, Europe, and South America, was trained on 6,000 photos of cats, and analyzes features like whiskers, eyelids, ears, and nose to distinguish between healthy cats and those possibly in pain. With a claimed accuracy of over 90%, it has become popular among cat owners seeking to ensure their pets' well-being. 

• Burger Flippy robot - California tech startup Miso Robotics created Flippy, a robot built for flipping burgers and assisting in the kitchen. Using AI, Flippy knows exactly when to flip burgers for perfect cooking every time. Originally just a flipper in 2018, it has since gained more features, now also acting as a griller.

👩🏼‍🚒Discover mind-blowing AI tools

1. Arcwise - An AI-first data analytics platform in your Google Sheet. Instant insights, zero CSV imports, one-click reports (Free Trial)**

2. MurfAI - An online voice generator tool for your videos, presentations, and brand commercials ($26/month) 

3. Competely - An AI-powered competitive analysis tool that helps users quickly gather information about their competitors ($39 one-time payment)

4. Bluebuddy Caption AI - An AI-powered tool that generates engaging captions for social media posts ($0.60 / 97 credits)

5. Little Artist - An AI-powered tool that transforms children's drawings into professional-level artwork (Free)

6. Mockey - Provides users with over 1000 product templates to create product mockups easily (Free) 

7. Blinknshop - An Amazon shopping assistant that helps users find and compare products across various categories (Free) 

8. Pic A Pet Name - An online platform that helps pet owners find the perfect name for their pets (Free) 

9. 12FT Ladder - An AI-based tool that helps you bypass the paywall to read articles (Free) 

10. Magnific - An AI-powered image upscaler and enhancer that increases the resolution and detail of images ($39/month) 

🫅🏼Historic moments

• 2015: The Google self-driving car project achieved the world's first fully autonomous ride on public roads. A legally blind passenger rode in the self-driving prototype, highlighting AI's potential for real-world applications, reshaping transportation, and prompting discussions on safety and regulations.

• 2018: OpenAI introduced the GPT-2, its groundbreaking AI language model capable of producing context-relevant content. It marked a significant milestone in natural language generation.

• 2019: AI aided in generating the first black hole image, confirming Einstein's theory. Over 200 scientists were involved in the AI-driven project called Event Horizon Telescope. The project used AI to process data and generate the image.

[The tool marked with ** is a sponsored feature.]