The AI-ronman 🚀

Hello, fellow tech enthusiasts! Did you hear about the AI that got into stand-up comedy? It bombed—turns out, its timing was just a bit… robotic! 🤖

Quick Takes ⚡

1. Authors accuse Anthropic of "large-scale theft" in Claude AI lawsuit

2. OpenAI's gift to devs: Free GPT-4o fine-tuning for a limited time

3. Microsoft releases powerful Phi 3.5 open-source AI models

4. OpenAI collaborates with Condé Nast to integrate premium content into ChatGPT

5. Nvidia demos new AI model in Mecha BREAK at Gamescom

6. AI Error: Microsoft Copilot falsely labels journalist as criminal

7. South Korean AI chip firms Rebellions and Sapeon finalize merger agreement

Deep Dive 🔍

• Authors accuse Anthropic of "large-scale theft" in Claude AI lawsuit💼
  
  A group of authors has filed a lawsuit against Anthropic, claiming the company engaged in "large-scale theft" by training its Claude chatbot on pirated copies of their works. This is the first lawsuit against Anthropic, following similar legal actions against OpenAI and ChatGPT. The authors allege that Anthropic used a dataset called The Pile, which contains numerous pirated books. While Anthropic argues that its practices fall under fair use, the lawsuit reignites the ongoing debate about AI training data and copyright infringement.

• OpenAI's gift to devs: Free GPT-4o fine-tuning for a limited time 🎁
  
  OpenAI has launched free fine-tuning for GPT-4o, allowing developers to customize the model for higher performance and accuracy. With 1 million free tokens per day through September 23, devs can fine-tune GPT-4o to improve structure, tone, and domain-specific instructions. Fine-tuning is available on all paid tiers, costing $25/million tokens after the free period ends. Early access users have already achieved state-of-the-art results on benchmarks using fine-tuned GPT-4o.
  
  

  

• Microsoft releases powerful Phi 3.5 open-source AI models 💻
  
  Microsoft has released three new open-source AI models in its Phi 3.5 series: mini-instruct, MoE-instruct, and vision-instruct. The mini-instruct model, with 3.8 billion parameters, is optimized for resource-constrained environments. The MoE-instruct model has 60.8 billion parameters but only 6.6 billion active, allowing it to outperform larger models. The vision-instruct model processes both text and images at GPT-4o level. Microsoft recommends combining these models with search systems to compensate for limitations.
  

  

• OpenAI collaborates with Condé Nast to integrate premium content into ChatGPT 📲✨
  
  OpenAI has announced a collaboration with Condé Nast to incorporate content from renowned media brands like Vogue and The New Yorker into ChatGPT and the SearchGPT prototype. This integration allows SearchGPT to provide direct links to source articles. OpenAI compensates publishers for their content through individual agreements, which may disadvantage smaller publishers not selected as "preferred." This partnership raises concerns about the impact on the online publishing model, as publishers face a dilemma: sign exclusive deals with OpenAI or risk being sidelined in an AI-driven landscape.

• Nvidia demos new AI model in Mecha BREAK at Gamescom 🎮
  
  Nvidia has showcased its new AI language model, Nemotron-4 4B Instruct, at Gamescom, specifically designed to enhance character conversations in video games. This model is part of Nvidia's ACE technology and operates on GeForce RTX graphics cards. It is currently demonstrated in the multiplayer mech game "Mecha BREAK," where it works alongside other AI technologies like Audio2Face-3D and Whisper. Optimized for low memory usage and quick responses, Nemotron-4 is expected to benefit over 100 million GeForce RTX devices.
  
  

  

• AI Error: Microsoft Copilot falsely labels journalist as criminal 🚫
  
  Microsoft's Copilot has mistakenly accused court reporter Martin Bernklau of serious crimes, including child abuse and involvement in a prison escape. Bernklau believes these false claims stem from his extensive coverage of trials in Tübingen, which the AI may have misinterpreted. Although Microsoft temporarily removed the erroneous entries, they reappeared shortly after, and prosecutors opted not to pursue charges. This incident highlights the unreliability of AI systems like Copilot, which can generate misleading information without understanding its truthfulness.
  
  

  

• South Korean AI chip firms Rebellions and Sapeon finalize merger agreement 🤝 
  
  South Korean AI chip manufacturers Rebellions Inc and Sapeon Korea Inc have finalized a merger agreement aimed at competing with global leaders like Nvidia. Sapeon's shareholders include major players such as SK Telecom and SK Hynix, providing strong backing for the merger. Rebellions recently secured a $15 million investment, bringing their total funding to over $225 million. This merger is expected to enhance their capabilities in the AI semiconductor market, which is increasingly competitive.

🔚 Thanks for tuning in! If you enjoyed this newsletter, don’t forget to share it! If you didn’t, just blame it on a glitch in the matrix! 😜💻

Ciao for now!

Author: Poonam 👧 

Karan 😎 🚀 
CEO, GlazeGPT.com