AI Today and Everyday — March 20, 2024

What’s new in the world of AI.

Sam Altman discusses GPT-5, Sora, Elon Musk, Ilya, AGI, power, Board Saga & Aliens.

In the recent discussion between Lex Fridman and Sam Altman, Sam openly discussed a wide range of topics that are of great interest to many. These included OpenAI, the development of GPT-5, the project Sora, updates on the board, Elon Musk’s legal challenges, insights from co-founder and chief scientist Ilya, issues around rights, and the pursuit of AGI, among others.

Altman‘s Perspectives:

• GPT-5: Sam Altman discusses the evolution and future of AI models, particularly looking forward to GPT-5. He expresses excitement about AI becoming smarter across the board, rather than in isolated capabilities. Altman hints at ongoing work and suggests OpenAI will release several significant updates before officially moving on to what may be termed GPT-5, emphasizing a continuous improvement trajectory rather than singular leaps.
• Sora: Altman shares insights on Sora, OpenAI’s project capable of generating simulated visual worlds. He reflects on the potential of Sora to transform content creation and hints at the broader implications for AI’s ability to create increasingly complex simulations. Despite the technological breakthroughs, Altman remains cautiously optimistic, recognizing the challenges and ethical considerations involved in advancing such technology.
• Legal Dispute with Elon Musk: The conversation touches on the complex relationship between Sam Altman and Elon Musk, particularly in light of Musk’s lawsuit against OpenAI and his criticisms. Altman expresses respect for Musk’s contributions to technology and hopes for an amicable relationship moving forward, despite the current tensions and differences in their approaches to AI development and safety.
• Ilya Sutskever’s Future: Altman discusses Ilya Sutskever’s role in OpenAI and the nuances of their professional relationship, especially in the context of OpenAI’s board saga and Ilya’s less public presence. He highlights Ilya’s ongoing importance to AI safety discussions and his hopes for continued collaboration.
• 7 Trillion Chip Project: Altman clarified that he hadn’t tweeted about wanting to raise $7 trillion for a chip project. However, he acknowledged that there is a significant need for more computing power. He pointed out that meeting this demand is a big challenge due to the requirements for energy, data center expansion, supply chain development, and chip production.
• Predictions about AGI: Altman predicts that by 2030, or potentially sooner, we’ll see the development of extremely advanced AI systems. He believes that the arrival of AGI isn’t the end but more like a beginning.
• Power: Altman contemplates the implications of achieving AGI, emphasizing the need for responsible governance and the dangers of concentrating too much power in any single entity’s hands. He reflects on the balance of power within OpenAI and the broader AI community’s responsibility to guide AGI development ethically.
• No Google Clone: Altman isn’t interested in creating a copy of Google; he thinks that’s an unnecessary use of AI. Instead, he aims to incorporate LLMs with search features to develop his unique business model. Additionally, he has a strong dislike for advertising.
• Board Saga: The dialogue includes Altman’s recounting of the tumultuous period surrounding OpenAI’s board restructuring, detailing the personal and professional challenges he faced. He shares insights into the lessons learned from the experience and the importance of resilience and governance in navigating the future of AI.
• Aliens: In a lighter moment, Altman and the host discuss the possibility of alien civilizations and how advancements in AI might help humanity explore such profound questions. This speculative conversation underscores the wide-ranging curiosity and ambition that drive Altman’s vision for AI’s potential.

Why It Matters:

According to the interview, Sam Altman seems to view computational power as a vital resource, almost like a currency, that will be essential in the future. He appears to recognize that as we move toward the goal of AGI, the demand for increased computational capacity will be immense. With the exponential growth of AI capabilities, Altman is likely hinting at the inevitable race to secure this ‘currency’ to power the next generation of AI systems and initiatives, reflecting on his discussions around the challenges and the financial intensity, such as the $7 trillion figure mentioned for a chip project.

Furthermore, Altman’s belief that the journey to AGI represents a massive, all-out struggle underscores the technological, ethical, and governance hurdles that must be navigated. The ‘struggle’ not only pertains to the technological advancements needed to achieve AGI, such as creating smarter AI models like GPT-5 or innovative projects like Sora, but also extends to ensuring responsible and safe development. This includes addressing legal disputes, such as those involving Elon Musk, managing power dynamics, as he contemplates in the scope of AGI governance, and pushing for collaborations in the face of board reshuffles. He sees the push towards AGI as a comprehensive challenge that affects us all, both inside and outside of OpenAI, and implies the need for a concerted, interdisciplinary effort.

In summary, Sam Altman sees computational power as the key to unlocking the future of technology and AI. He regards the path to AGI not merely as a technological endeavor but as a complex battle that will test our collective resolve in multiple domains. The advancements at OpenAI, escalating AI startups’ valuations, and global chip-making strategies all interplay in this vast, intricate contest to reach AGI. His vision is a testament to AI’s profound potential and its entwined challenges, necessitating a broad perspective that spans far beyond code and algorithms.

Microsoft hires Suleyman to lead AI push

Image source: Bloomberg

The Rundown: Microsoft just announced the hiring of Inflection AI’s co-founder Mustafa Suleyman, who will head up a new division called Microsoft AI focused on advancing Copilot and the company’s consumer AI efforts.

The details:

• Suleyman will report directly to CEO Satya Nadella, overseeing projects like integrating AI assistants into Windows and Bing search.
• Inflection co-founder Karen Simonyan also joins Microsoft AI as chief scientist, along with several other members of the Inflection team.
• Inflection tapped Sean White as its new CEO, with plans to focus on the company’s enterprise API platform instead of its consumer chatbot.

Why it matters: Microsoft is doubling down on talent as it races other tech giants to put AI tools in consumer’s hands. The move brings a team experienced in building AI systems with a focus on human-centric interaction — meaning Copilot may get a major personality makeover in the near future.

OpenAI CEO Teases ‘Amazing’ New AI Model for 2024: OpenAI CEO Sam Altman has teased an “amazing” new AI model set for release in 2024. While details are scarce, Altman hinted at the possibility of smaller releases before the much-anticipated GPT-5. The Lex Fridman podcast offered further insights into the company’s direction, including the potential of a video model named Sora and the continued focus on AI logic with the Q* project. Despite the excitement, Altman expressed a degree of dissatisfaction with recent GPT-4 updates, suggesting the next iteration will have a major leap in performance. Read more.

Google Researchers Develop ‘MELON’ for Easier 3D Object Creation: Google Research has unveiled a new method, dubbed ‘MELON’, that makes it easier to create 3D models of objects from images. The technique addresses the common difficulty of needing precise camera information to reconstruct 3D shapes. MELON uses a streamlined approach to overcome this hurdle, potentially simplifying 3D modeling for various applications. Demonstrated at the 3DV 2024 conference, MELON shows promise in fields like e-commerce and self-driving car technology. Read more.

Image-to-video model builds 3D videos

• Open-source, AI start-up — Stability AI — has released Stable Video 3D (SV3D) that can transform 2D images into 3D videos, ideal for e-commerce, retail, and gaming.
• SV3D is built on the foundation of Stability AI’s existing image-to-video model, which can take a still image and generate a photorealistic, moving video clip.
• It’s already available for commercial use — with a Stability AI Membership — and developers can access its open-source code via Hugging Face (for noncommercial use).

Google DeepMind introduced TacticAI, an AI system that can advise coaches on football tactics by using predictive and generative models to analyze previous play data, created in a collaboration with Liverpool FC.

Nvidia announced Earth-2, a cloud platform that uses AI and digital twin technology to forecast climate change and extreme weather events with high resolution and speed.

Saudi Arabia reportedly plans to create a $40B fund to invest in artificial intelligence, potentially becoming the world’s largest investor in the sector.

Pika Labs’ new Sound Effects feature is now available to all users, allowing for automatic or custom SFX generations to pair with video outputs.

Samsung is creating a research lab to design a new type of semiconductor specifically tailored for AGI, aiming to meet the growing demand for AI chips.

Buildbox released an alpha version of Buildbox 4, an AI-first game engine allowing users to create games and generate assets from text prompts.

Fitbit is integrating AI features in collaboration with Google Research, developing a Personal Health LLM to provide users with personalized coaching and insights on biometric data.