Hey Synapticians,

Great announcements from Mister Altman: fingers crossed that this is followed by quick action… and that it concerns Europe!

Then, we’ll see that the debates between Musk and Altman continue; that startups working on drug discovery are raising funds; and we’ll finish with the current limitations of LLMs in "understanding" language.

But most importantly, enjoy the Chuck Norris meme!

Top AI news

1. OpenAI Announces Major Simplification of AI Product Line
OpenAI acknowledges that its diverse range of AI models has led to user confusion. To address this, the company plans to simplify its product line by integrating various technologies into a unified model. CEO Sam Altman announced that GPT-4.5, internally called "Orion," will be OpenAI's last non-chain-of-thought model. Following this, the primary goal is to merge the o-series and GPT-series models, creating systems capable of utilizing all available tools and determining the appropriate level of reasoning required for tasks. With the release of GPT-5, OpenAI will discontinue o3 as a standalone model, aiming to provide a more streamlined and user-friendly experience.

2. Elon Musk Threatens OpenAI Takeover Withdrawal Over Mission
Elon Musk has proposed withdrawing his $97.4 billion bid for OpenAI if its board agrees to maintain its nonprofit mission and stop its transition to a for-profit entity. The offer, submitted to a U.S. court, aims to ensure OpenAI remains aligned with its original goals. OpenAI, led by Sam Altman, quickly dismissed the bid, stating the nonprofit is not for sale. Musk, a co-founder of OpenAI, has also sued the company for alleged anticompetitive behavior. OpenAI is restructuring into a for-profit corporation, a move Musk is trying to block, escalating tensions over the future of AI governance.

3. AI Start-up Raises $50 Million for Drug Discovery
Latent Labs, co-founded by former DeepMind scientist Dr. Simon Kohl, has emerged from stealth with $50 million in funding to develop AI models that accelerate drug discovery. The company aims to enable researchers to computationally create new therapeutic molecules, such as antibodies or enzymes, potentially opening new paths to personalized medicines. Their AI-driven research platform is designed to expedite drug development timelines and increase clinical success rates. The funding includes a $40 million Series A raise co-led by Radical Ventures and Sofinnova Partners, with participation from notable angel investors including Google's chief scientist Jeff Dean and Cohere's founder Aidan Gomez. Latent Labs is co-headquartered in London and San Francisco.

Bonus. AI Models Still Lag Behind Humans in Text Understanding
An international research team found that AI models, including ChatGPT-4, achieve only 83% accuracy in simple text comprehension tests, compared to 89% for humans. This highlights AI's limitations in understanding language. An international research team led by the University Rovira i Virgili evaluated seven AI models' language comprehension abilities, comparing them to humans. The study, published in Scientific Reports, revealed that while models like ChatGPT-4 achieved up to 83% accuracy in simple text comprehension tests, humans averaged 89%. This indicates that despite generating grammatically correct and coherent texts, these AI models do not understand language meaningfully as humans do; they rely on statistical patterns in their training data.

Meme of the Day

Chuck is the Best: ‘AI doesn’t exist, it’s Chuck Norris who generates all the answers.‘ 😉

Theme of the Week

Reasoning model - Paper of the week
OpenAI's recent study reveals that reinforcement learning can significantly boost large language models' performance in complex coding tasks. Their model, o3, achieved top honors in the 2024 International Olympiad in Informatics without human-crafted strategies, indicating a promising future for AI in problem-solving domains.

Stay Connected

Feel free to contact us with any feedback or suggestions—we’d love to hear from you !