• AI Boost
  • Posts
  • 🚀AI Boost: AI helped the feds catch $1 billion of fraud

🚀AI Boost: AI helped the feds catch $1 billion of fraud

PLUS: New Mistral AI Model and More!

Good morning! It is July 29, 2024, today. We've got a fantastic lineup of AI news for you, from breakthroughs in robotics to Reddit's new policy on AI web crawlers. Let's dive in!

1. AI Powers $1 Billion Check Fraud Recovery: US Treasury's New Weapon Against Financial Crime

The US Treasury's use of machine learning AI has significantly boosted fraud detection efforts, recovering $1 billion in check fraud during fiscal 2024—nearly three times the previous year’s amount. In total, AI helped prevent and recover more than $4 billion in fraud, marking a six-fold increase from the prior year. While not using generative AI like ChatGPT, the Treasury relies on AI models that swiftly analyze vast data to uncover hidden fraud patterns. This effort aims to protect taxpayer money as fraudsters target federal payouts.

2. Mistral Unveils AI Models for Laptops and Phones

French AI startup Mistral has launched its first generative AI models, "Les Ministraux," optimized for laptops and phones. The two models, Ministral 3B and 8B, support applications like on-device translation and autonomous robotics, offering local, privacy-first AI solutions. With a 128,000-token context window, they can process data equivalent to a 50-page book. Available for research and cloud use, these models promise low-latency performance and outperform rivals like Meta's Llama on key AI benchmarks. Mistral, which raised $640 million, continues to expand its AI portfolio while navigating revenue challenges.

3. Nvidia Quietly Releases AI Model That Outperforms GPT-4

Nvidia has introduced Llama-3.1-Nemotron-70B-Instruct, an AI model outperforming GPT-4 and Anthropic’s Claude 3.5 Sonnet in key benchmarks. Released on Hugging Face, it signals Nvidia's move beyond hardware into leading AI software development. The model excels at language understanding and is tailored for business applications, offering a cost-effective alternative to high-end AI solutions. Nvidia’s expanding AI strategy challenges the dominance of software giants, positioning itself as a top player in both hardware and AI software.

4. Hong Kong Police Arrest 27 in $46M AI-Driven Romance Scam

Hong Kong police have arrested 27 individuals involved in a romance scam using AI face-swapping technology to defraud victims of $46 million via fake cryptocurrency investments. The operation involved scammers creating fake personas on dating platforms, using AI to enhance their appearance and even deepfakes for video calls. Victims were misled into thinking they were investing in legitimate crypto platforms. The arrested include six university graduates, with some linked to the Sun Yee On triad. This highlights the rising issue of AI-powered fraud.

Credit: Amazon

5. Amazon Launches New Kindle Scribe, Emphasizing Notebook Features

Amazon’s updated Kindle Scribe, priced from $399.99 and shipping on December 4th, enhances its notebook functionality with improved design and writing experience. Key features include a refined stylus, paper-like feel, and Active Canvas, which lets users write notes directly on books and PDFs. The Scribe can now convert handwriting to text, with AI-powered tools for summarizing and refining notes. While still a reading device, the Scribe is increasingly being used for work and note-taking, signaling a shift in Kindle usage.

6. Yann LeCun: AI is Far from Human-Level Thinking, But World Models Could Change That

Meta’s chief AI scientist Yann LeCun argues that current AI systems, like those from OpenAI, are far from human-level intelligence. While some suggest AI can "remember" and "reason," LeCun explains that AI lacks common sense, memory, and the ability to interact with the physical world. He advocates for the development of "world models"—AI systems that understand three-dimensional spaces and plan actions, predicting outcomes. LeCun believes achieving human-level AI could take years or even a decade.

How would you rate today's newsletter?

Vote below to help us improve the newsletter for you.

Login or Subscribe to participate in polls.

Stay tuned for more updates, and have a fantastic day!

Best,

Zephyr