Daily AI Newsletter: Amazon has a secret way to scrape Microsoft’s GitHub

In today’s email:

🔒 CryptGPT: Privacy-Preserving Language Models Using Vigenere Cipher
❤️ Why Chinese women are looking to ChatGPT for love
📚 Tutorial: Developing an LLM: Building, Training, Finetuning
🧰 10 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

Amazon has a secret way to scrape Microsoft’s GitHub and feed its AI model

Amazon is in need of vast amounts of high-quality data to develop powerful AI models, and it has identified GitHub as a valuable source of coding metadata. An internal memo from Amazon’s Artificial General Intelligence (AGI) Group, obtained by Business Insider, reveals the company's strategy to collect this data despite GitHub’s scraping limits of 5,000 requests per hour per account. With over 150 million public repositories on GitHub, traditional data collection methods would have taken years. To expedite this process, Amazon proposed a workaround: having its employees create multiple GitHub accounts and share their credentials, effectively allowing the company to collect the necessary data within weeks instead of years.

The memo includes detailed instructions on creating and managing these accounts, using Amazon work emails, specific GitHub tokens, and setting appropriate permissions. Although Amazon claims this method has been approved by its legal and security teams, the strategy raises ethical concerns regarding data privacy, consent, and the proper use of platform resources. By encouraging employees to share their GitHub credentials, Amazon may be accessing data without explicit permission from GitHub or repository owners, potentially violating ethical standards even if it remains within legal boundaries.

Amazon’s need for GitHub data is driven by its goal to advance AI capabilities, crucial for understanding human language and making predictions. GitHub's diverse and extensive repository of open-source projects offers valuable insights into project evolution, developer collaboration, and contribution patterns. This metadata is essential for training AI models to improve accuracy and problem-solving abilities. While access to such comprehensive datasets can give Amazon a competitive edge, it also highlights the ethical dilemmas tech companies face in responsibly using and protecting digital information.

Two Forbes 30U30 Founders Transforming Mental Wellness

Forbes 30 Under 30 winners founded Aura to solve the $100B problem - mental wellbeing.

Aura has quickly grown to 8 million users & 100k+ paying subscribers, and attracted investments from legendary Silicon Valley VCs & executives from Spotify, Facebook, and Apple.

The best part? You can invest now before the opportunity ends in a few days.

Black founders are creating tailored ChatGPTs for a more personalized experience

John Pasmore, an experienced AI founder, initially welcomed ChatGPT but soon noticed its lack of cultural nuance, particularly for Black communities. This gap is reflective of a broader issue where most AI models, trained primarily on Western data, fail to capture the diverse cultural experiences of people of color. To address this, Pasmore created Latimer.AI, a language model designed to reflect Black and brown experiences more accurately. Similarly, Erin Reddick's ChatBlackGPT and Tamar Huggins' Spark Plug are emerging as personalized AI tools that prioritize Black cultural contexts, enhancing educational and conversational experiences for Black and brown users.

In Africa, the challenge is even more pronounced due to the vast linguistic diversity with over 2,000 languages. AI models like CDIAL.AI, developed by Yinka Iyinolakan, are being created to bridge this gap. CDIAL aims to support nearly all African languages and dialects, focusing on speech patterns. Efforts like these highlight the limitations of mainstream AI models in representing non-Western cultures and emphasize the need for more inclusive AI development.

Additionally, initiatives like pocstock by Steve Jones and DeSean Brown aim to diversify stock images, addressing the lack of representation in visual AI outputs. These efforts reflect a broader movement towards personalized AI that can better serve diverse communities. As AI continues to evolve, there is a growing call for more cultural-specific models and for people of color to take leading roles in AI innovation, ensuring that future AI developments are inclusive and representative of all cultures.

Tutorial: Developing an LLM: Building, Training, Finetuning

Other stuff

Apple's new AI is made in Google data centers
Microsoft’s star AI chief peers into OpenAI’s code, highlighting an unusual rivalry
Apple joins the race to find an AI icon that makes sense
McDonald’s will stop testing AI to take drive-thru orders, for now
AI Agents: Hype vs. Reality
AI is quickly becoming a regular part of children’s lives. What happens next?
Apple’s Slow Rollout of Intelligence Features Will Stretch Into 2025
A practical guide to making your local AI chatbot smarter
CryptGPT: Privacy-Preserving Language Models Using Vigenere Cipher
Why Chinese women are looking to ChatGPT for love