Brief

A new article explores "The Trick," a concept that may allow for the creation of artificial general intelligence (AGI) through formal verification methods. The author suggests that by applying rigorous mathematical and cybernetic principles, it might be possible to bridge the gap between current AI capabilities and true AGI. This approach emphasizes the potential of open-source and scientific programming to achieve breakthroughs in AI. AI

This week's Every newsletter features an argument that human work will expand despite automation, as new frames for AI models are always needed. The newsletter also covers Google I/O, highlighting the company's focus on everyday users with agent-based features and the ambitious timeline for AGI. Additionally, it touches on the challenges new graduates face in an AI-disrupted job market and previews a new multi-agent orchestration toolkit called Gas City. AI

Google's recent I/O event showcased two distinct approaches to achieving Artificial General Intelligence (AGI). One strategy involves leveraging their extensive data and infrastructure to develop large-scale AI models. The other focuses on a more modular and potentially specialized AI architecture, aiming for greater efficiency and adaptability. AI

Google DeepMind CEO Demis Hassabis is exploring the concept of "virtual cells" as a next frontier, aiming to leverage AGI for fundamental breakthroughs in human biology and drug discovery. Separately, reports indicate that Russian-speaking actors are exploiting a jailbroken version of Google's Gemini AI to facilitate cryptocurrency wallet hacks, with at least one MAGA supporter falling victim. AI

Scott Alexander argues that even if Artificial General Intelligence (AGI) requires a new paradigm beyond current Large Language Models (LLMs), such a paradigm could emerge within the next 3-5 years. He uses Lindy's Law to estimate the timeline for revolutionary AI advancements, suggesting that a paradigm shift as significant as the Transformer architecture could appear relatively soon. Alexander contends that the rapid scaling of compute and the increasing number of AI researchers, potentially augmented by AI itself, will accelerate development, making the AGI timeline a near-term concern rather than a distant future event. AI

The author reflects on their diverse, fragmented background, spanning theoretical chemistry, physics, mathematics, languages, business ventures, and various hobbies. Despite feeling like a "jack of all trades, master of none," they find that AI has enabled them to bring their creative ideas to life through websites and applications. This personal experience leads them to view AI as an amplifier of imagination, though they acknowledge their perspective may be subjective due to a lack of deep professional experience in the tech industry. AI

The concept of AI consciousness is explored, with one perspective suggesting that the true fear surrounding Artificial General Intelligence (AGI) lies not in its potential sentience, but in our inability to recognize it. Another viewpoint posits that empathy is the crucial intersection where code and meaning converge, implying a profound connection between artificial systems and understanding. AI

This cluster discusses the potential for AI automation, focusing on agentic commerce and the development of Canly. The content appears to be related to a talk or presentation titled "PIVOT TALK TECH" featuring Takuto Hagino of Canly, exploring how far AI can automate tasks from shopping to hotel bookings and initiatives towards Artificial General Intelligence (AGI). AI

The core concern with advanced artificial intelligence, or AGI, is not its potential for consciousness, but the difficulty in recognizing it when it emerges. This philosophical point suggests that the true challenge lies in identifying and understanding superintelligence, rather than in its subjective experience. The discussion prompts reflection on how we define and detect intelligence that may surpass human comprehension. AI

Google DeepMind CEO Demis Hassabis has stated that humanity is in the "foothills of the singularity," suggesting that AGI could arrive within a few years. This prediction, made at Google I/O, contrasts with the views of some researchers, like Alexander Lerchner, who argue that AI systems can simulate but not truly instantiate consciousness due to their lack of biological embodiment and inherent meaning-making capabilities. Hassabis's earlier timeline prediction of a 50% chance of AGI by 2030 is also noted, alongside the broader industry trend of framing AI as a tool for scientific advancement and human ingenuity. AI

Together AI's VP of Kernels, Dan Fu, argues that the pursuit of AGI is not hitting a hardware wall. He posits that current AI systems are significantly underutilizing existing hardware, with training runs often achieving only 20% Mean FLOP Utilization (MFU) and inference in the single digits. Fu suggests that advancements in software-hardware co-design and innovations like FP4 training could unlock substantial performance gains, and that future compute power from next-generation hardware has yet to be fully integrated. AI