- Superintelligence
- Posts
- 2024: Grok 2 &3
2024: Grok 2 &3
In today's edition of Superintelligence, explore Alibaba's Qwen performance, the impact of Fintech AI in India, significant AI investments from South Korea, and other key global developments in the AI world.
The AI World Today
Grok 2 Coming August
Alibaba's Qwen Tops Leaderboard
TIFIN Launches MyFi India
SK Group's AI Expansion
CIR Sues OpenAI, Microsoft
+
Heads Up
Save the Date
Good Reads
Grok 3 Set for End-Year Launch

Screenshot: Elon Musk/X
Elon Musk announced on X that Grok 2, an advanced AI language model, will be released in August. Grok aims to remove content from other LLMs found on the internet. Grok 3, which will utilize much larger compute resources, is scheduled for release by the end of the year. Musk emphasized that Grok 2 will significantly improve over previous versions, exceeding current AI metrics. In March, Musk noted that Grok 2 was in training and would surpass existing AI. Grok, initially launched earlier this year, is Musk's response to advancing AI technology, and Grok 3 is expected to train on 100,000 H100 GPUs, promising substantial advancements.
Alibaba’s Qwen Dominates LLM Leaderboard Rankings

Screenshot: Hugging Faces
latest Open LLM Leaderboard update reveals that Alibaba's Qwen-2-72B-Instruct model is currently the top-ranked open-source model on Hugging Face, outperforming competitors like Meta's Llama3-70B and Mistral AI's Mixtral 8x22B. Qwen-2-72B-Instruct demonstrated remarkable proficiency across various tasks, including knowledge assessment, comprehension of lengthy and complex inputs, and advanced mathematics. The model achieved an impressive average score of 43.02, significantly surpassing Meta's Llama-3-70B-Instruct, which scored an average of 36.67. This performance showcases Qwen's advanced capabilities and positions it as a leader in the current AI landscape, with its high scores attributed to its robust instruction fine-tuning and versatile task performance.
MyFi Offers Customized Investment Guidance in India

AI-Generated Illustration
US-based fintech platform TIFIN has launched its conversational AI assistant, MyFi, in India. MyFi offers personalized investment guidance using advanced AI models and research-driven insights, aiming to revolutionize wealth management for Indian users. The platform, compatible with both Android and iOS, simplifies investment decisions by providing tailored advice based on users' existing portfolios. TIFIN's CEO, Vinay Nair, highlighted MyFi's role in enhancing financial decision-making for Indian investors. Initially targeting mutual fund investors, the app plans to expand its inclusivity in the future.
SK Hynix Commits $74.6 Billion to AI

South Korea's SK Hynix, the world’s second-largest memory chip maker, plans to invest 103 trillion won ($74.6 billion) through 2028 to enhance its chips business with a focus on AI, according to SK Group. Additionally, SK Group aims to secure 80 trillion won by 2026 for investments in AI, semiconductors, and shareholder returns while streamlining its 175 subsidiaries. Following significant losses by SK Hynix and its EV battery arm, the conglomerate aims to improve competitiveness by focusing on AI-related technologies. SK Group projects a profit before tax of around 22 trillion won this year, with a target of 40 trillion won by 2026. South Korea’s government has also announced a 26 trillion won support package for its chip industry.
Investigative Group Files Suit Against AI Giants

Illustration: The Verge
The Center for Investigative Reporting (CIR), behind Mother Jones and Reveal, is suing OpenAI and Microsoft for copyright infringement. CIR alleges that the defendants copied and used their valuable content without permission or compensation, damaging CIR’s relationships with readers, consumers, and partners, and depriving them of revenue. The lawsuit emphasizes the fundamental purpose of copyright law, which is to protect creators' works and ensure they are properly compensated. CIR claims that OpenAI and Microsoft populated their training sets with copyrighted journalism without respecting the copyright, leading to unauthorized copying and distribution. The defendants allegedly trained ChatGPT to ignore copyright laws, further exacerbating the infringement.
Heads Up
AI Robotics : First Commercial Humanoid Robots Deployed
AI Warfare: Chinese Military Unveils Armed Robot Dog
AI Competition: Can DeepL Surpass Big Tech?
AI Education: WhatsApp AI Tortur Launches In India
AI Investment: Amazon's $100 Billion AI Investment
AI Features: Google AI Supports 110 Languages’ Translation
Save the Date
July 1-2: Berlin Hosts Artificial Intelligence in Mental Health Summit
July 3-4: India to Host Global India AI Summit 2024
Good reads
July 1: Scaling Synthetic Data Creation with 1,000,000,000 Persona
June 29: Regurgitative Training: The Value of Real Data in Training Large Language Models