- AI KATANA
- Posts
- OpenAI is reportedly working on more advanced AI models capable of reasoning and ‘deep research’
OpenAI is reportedly working on more advanced AI models capable of reasoning and ‘deep research’
Also: Will K-pop's AI experiment pay off?
Morning!
In today's newsletter, we delve into OpenAI's ambitious new project "Strawberry," aimed at enhancing AI's reasoning and research capabilities, which could revolutionize problem-solving across various domains. We also explore the K-pop industry's embrace of AI for creative processes and the mixed reactions from fans. Universities are being called to innovate and integrate more closely with industry to stay relevant in the AI era. A feature on AI agents highlights their evolving roles and challenges in achieving full autonomy. Additionally, we assess ChatGPT's mixed performance in coding tasks, discuss Figma's AI tool controversy, and introduce a new app that uses AI to boost recycling efforts. Lastly, we examine the surge in venture capital funding driven by the AI boom, amid cautionary notes reminiscent of past tech bubbles.
Sliced:
🧪 OpenAI is reportedly working on more advanced AI models capable of reasoning and ‘deep research’
🇰🇷 Will K-pop's AI experiment pay off?
🎓 Role of universities in the AI era
🤖 What exactly is an AI agent?
🧑💻 How Good Is ChatGPT at Coding, Really?
OpenAI is reportedly advancing its AI models with a secret project named "Strawberry," which aims to enhance reasoning and deep research capabilities. This project involves developing AI that can autonomously browse the internet to plan and execute complex tasks, marking a significant leap from current capabilities. Earlier iterations, known as Q*, demonstrated proficiency in answering challenging science and math questions. The project aspires to achieve human-like reasoning, positioning OpenAI's models to perform deep, autonomous research, potentially transforming how AI tackles intricate problems and contributes to various fields.
The K-pop industry is embracing AI, with major bands like Seventeen using the technology for music videos and songwriting. Member Woozi of Seventeen explained their approach as a way to evolve with technological advancements, aiming to leverage AI's benefits. However, this has sparked a divide among fans. Some, like podcaster Ashley Peralta, worry that AI-generated lyrics could erode the personal connection between artists and fans, while others see it as a useful tool for overcoming creative blocks. Music producer Chris Nairn notes that while AI can assist in lyric creation, it lacks the innovative spark of human creativity, predicting that fan demand for authentic, personal music will grow. This technological trend in K-pop highlights broader concerns about the regulation and ethical use of AI in the music industry.
Universities are being urged to transform their traditional roles of teaching and research to meet the demands of the AI era, by integrating closely with industry and society. This shift is seen as essential for fostering innovation and addressing real-world challenges. Examples like Stanford University's Hasso Plattner Institute of Design illustrate successful industry-academia collaboration, which combines academic research with practical applications, creating an ecosystem that drives both industrial development and social progress. The establishment of independent colleges focused on industry-academia collaboration is proposed as a model for universities, especially in non-technical fields, to remain relevant and competitive. Korean universities, facing demographic challenges, must adopt proactive and innovative approaches to sustain their relevance and contribute meaningfully to societal and industrial advancements.
The definition of AI agents varies widely, with no consensus among experts. At its core, an AI agent is software designed to perform tasks autonomously, often those previously handled by humans, such as customer service or IT support. AI agents leverage technologies like natural language processing, machine learning, and computer vision to understand their environment, make decisions, and take actions to achieve specific objectives. Companies like Google and Asana see AI agents as task-based assistants or additional virtual employees. Despite the optimistic outlook for AI agents, challenges remain, especially in integrating multiple systems and achieving fully autonomous operation. As AI continues to evolve, the development of a robust infrastructure to support AI agents is crucial.
A study published in IEEE Transactions on Software Engineering assessed the effectiveness of ChatGPT in coding, revealing a wide range of success rates based on the complexity of tasks, programming languages, and other factors. ChatGPT's ability to generate functional code varied significantly, achieving success rates as high as 89% for some tasks and as low as 0.66% for others. The AI excelled in solving pre-2021 coding problems due to their prevalence in its training data but struggled with newer problems. While ChatGPT often produces efficient code, it lacks the critical understanding of algorithm problems, leading to errors and security vulnerabilities. Researchers recommend providing additional context and guidance to enhance its performance and mitigate risks.
🛠️ AI tools updates
Figma recently pulled its new AI tool, Make Designs, after facing criticism for creating app designs that closely resembled Apple's iOS weather app. The tool, which uses generative AI to help users quickly mock up apps, raised concerns about potential legal issues and copyright infringement. Figma CEO Dylan Field acknowledged the problem, attributing it to the use of third-party models and a bespoke design system that lacked sufficient variation. Figma CTO Kris Rasmussen stated that the AI models were not trained by Figma but were powered by OpenAI’s GPT-4o and Amazon’s Titan Image Generator G1. The company plans to refine its design system and take additional precautions before re-enabling the tool, ensuring that future models only learn general design patterns to better serve professional designers.
A new AI-powered app developed by Swedish recycling platform Bower transforms smartphone cameras into recycling tools by allowing users to scan products to determine their recyclability and the CO2 savings from recycling. Funded by a €1.75 million grant from Google's charitable arm, the app uses Google Gemini-trained AI models and crowdsourced data to identify item details and provide disposal instructions. With a user base of 680,000, the app offers real-time guidance on recycling options at various locations. By partnering with over 600 brands, Bower aims to enhance recycling rates and promote the circular economy, demonstrating AI's potential in environmental sustainability.
💵 Venture Capital updates
The hype surrounding AI is significantly boosting VC funding, with AI-related deals accounting for nearly half of the total value of all VC deals in the U.S. this year. In the second quarter alone, major funding rounds for AI startups like CoreWeave and xAI comprised over 26% of total VC deal values. Despite greater scrutiny in private capital markets, AI investments continue to attract substantial interest, though some caution is advised due to potential risks of low-diligence deals. The sector's rapid advancements and proven use cases drive optimism, yet concerns about a possible AI bubble persist, echoing sentiments from the dot-com era.
🫡 Meme of the day
⭐️ Generative AI image of the day
Before you go, check out Amazon AI Shopping Assistant Rufus Answers Existential Questions Too.