Weekly Tech News of May Last Week 2024

What Novelties Can Be Found in the Outcomes of Reddit and OpenAI’s Collaboration, Let’ Reveal

OpenAI has partnered with Reddit to access real-time content via Reddit’s data API. As an outcome, you are guys free to enhance the role of ChatGPT and other AI models by incorporating Reddit conversations with them.

In this deal OpenAI enable to improve its AI’s precision and context-awareness, similar to a previous $60 million agreement with Google. The collaboration also enables Reddit to develop new AI-powered tools for users and moderators, potentially enhancing content moderation and user engagement.

However, it raises privacy and ethical concerns, with community acceptance dependent on OpenAI’s adherence to privacy norms. This strategic move positions OpenAI against giants like Google and Microsoft, while Reddit could gain a competitive edge in social media. Let’s have an experience with this new launch and leave your thoughts as well.

Official Source: https://openai.com/index/openai-and-reddit-partnership/

Elon Musk’s xAI Secures $6Billion By Announcing the Funding Round Called Series B

xAI, Elon Musk’s AI startup that was established last summer, has announced a $6 billion investment round to launch products, develop infrastructure (such as a data center powered by Nvidia H100 chips, which should become obsolete by 2025), speed up research, development, and much more.

Musk, the former OpenAI co-founder, returned to the field of artificial intelligence after splitting from the company in 2018 due to its adoption of proprietary models. Grok-1 from xAI beat models such as Llama-2-70B and GPT-3.5,
but it is still lagging behind GPT-4 and Claude3.

Still, it is running behind in the race of the AI industry’s leaders, such as GPT-4. The notable investors in this deal include Saudi Prince Al Waleed bin Talal, Andreessen Horowitz, and Sequoia Capital. xAI starts with the goal of creating a supercomputer powered by Nvidia chips by 2025, whereas Musk is more ambitious about the goals for AI’s new achievements.

Official Source: https://x.ai/blog/series-b

AI Innovation is Contributing to Finding the Mate for the World’s Most Lonely Plant

A University of Southampton-led project now using AI and drones to find out the mate for the world’s loneliest plant, the Encephalartos woodii (E. woodii). This endangered plan seris especially male plant, dating back before dinosaurs, has no female partners and can’t reproduce naturally.

Dr. Laura Cinti leads the AI-driven search in South Africa’s Ngoye Forest, aiming to recognize plants by shape using their image recognition algorithms. The project, inspired by a tale of unrequited love, hopes to save this ancient species through natural reproduction, a groundbreaking endeavor with less than 2% of the vast forest explored so far.

As a consequence, Artificial Intelligence is now contributing to enhancing the Exploration of Untouched areas of conservation and Natural Reproduction especia;y for the endangered Plant Species in real time.

Official Source: https://www.southampton.ac.uk/news/2024/05/the-high-tech-hunt-for-a-lonely-plants-partner.page

CodecLM: A Collaboration of Language Model with Synthetic Data

CodecLM, a framework is now tailoring with high-quality synthetic data to align large language models (LLMs) for specific tasks without having any human annotation. It emphasizes the importance of instruction tuning in LLM (Larege Language Model) alignment to enhance the ability to follow instructions effectively across various formats.

Now you are free to leverage CodecLM a strong LLM as a codec to encode and decode seed instructions into tailored synthetic data, employing strategies like Self-Rubrics and Contrastive Filtering to improve the quality and understanding concept of the data.

Utilizing this approach you can outperform baselines on open-domain instruction-following benchmarks. Plus, are free to demonstrate its effectiveness in aligning LLMs for different instruction distributions.

Official Source: https://research.google/blog/codeclm-aligning-language-models-with-tailored-synthetic-data/