AGI Future and Impact

The Argument for an AGI Horizons Team In early 2023, shortly after the release of ChatGPT, a major software company reached out to me to try and understand what was happening with Gen AI and how it might impact their product roadmap and business strategy. We talked through how ChatGPT worked, compared notes on what it could mean to their product development plans, and I shared insights into what else the AI labs were working on that might affect their company and its customers in the coming years. One of my key recommendations for them was to form what I termed an “AGI Horizons Team” tasked with monitoring advancements toward artificial general intelligence (AGI) and assessing potential threats and opportunities. While there continues to be a lack of agreement on how to define AGI, I consider it an AI system that is generally capable of outperforming the average human at most cognitive tasks (e.g. ChatGPT being able to perform 90% or more of marketing tasks better than an average marketer). For more than 70 years, researchers have pursued this idea of human-like general intelligence. They were driven by a belief that we could give machines the ability to think, reason, understand, create, and take actions in the digital and physical worlds. But, progress was often slow, and the impact on our professional lives was minimal. Then, everything changed—and accelerated—with the release of ChatGPT in November 2022, and the rise of Gen AI. By early 2023, the tone and positioning on AGI from the leading AI labs had changed. They no longer talked about AGI as something that might be possible in a decade or more. They were conveying increasing confidence that there was a clear path to achieving AGI within 3 - 5 years. My point to the software company was that while everyone was racing to understand the impact of the current forms of AI on their business, far smarter and more generally capable models were on the horizon that might force them to reimagine and reinvent their products and business model. That if AGI was unlocked by OpenAI, Google Deepmind or another AI lab, it would change everything. And while the probability of that occurring in the 3 - 5 year window was relatively unknown, it certainly wasn’t zero. In other words, there was a potential transformative event (maybe even an extinction-level event for their company) possible within half of a decade. I felt that it was worth putting a team of their best people together to assess, along with outside advisors who can be more objective about the path forward. I now believe there is a greater than 50% chance of an AI lab claiming they have achieved AGI with 1 - 2 years. What that means to your business and industry is unknown. What that means to society and the economy is also unclear. But, I think it’s significant enough that we should all be doing more to consider the possibilities. Maybe it’s time for an AGI Horizons Team in your organization.

The Problem with AI Isn't Just Data, It's the Nature of Intelligence! I am often asked about my opinion on AGI and the true capabilities of AI. I am not pessimistic about AI and neither optimistic about our current scaling of AI. Here are my views shaped by my humble understanding of AI and my passion and interest to have better AI systems that are safe and trustworthy, and that augment our capabilities but not replaces us! One of the most pressing challenges in AI today isn’t just its ability to process information, it’s how it learns. Most of today's AI models are trained on massive datasets and optimized to infer patterns within that training data. But here's the real test: What happens when AI encounters something it hasn’t seen before? Will it truly understand, or will it guess? Can it reason beyond the boundaries of its training data? It does not seem so! This is the limit of current AI - it is not intelligence in the human sense. It’s powerful pattern recognition, not reasoning. It lacks intuition, context awareness, and a true sense of meaning.   The challenges with AI including: - Overfitting to data rather than learning abstract concepts. - Hallucinations and false inferences when faced with unfamiliar scenarios. - The illusion of intelligence without true understanding or common sense. The current trends in AI reasoning aim to move from shallow pattern recognition to deeper, structured thinking - more akin to how humans solve problems. Techniques like chain-of-thought prompting, neuro-symbolic reasoning, and agent-based architectures are early attempts to replicate human-like deductive steps. While promising, these methods often mimic how we think without actually understanding why. Human reasoning is built on lived experience, emotion, and adaptability - dimensions AI still struggles to grasp. The dream of Artificial General Intelligence (AGI) includes reasoning, deduction, adaptability, and understanding context across domains - hallmarks of human intelligence. But will AGI ever truly embody those traits? That remains uncertain. Even as models grow more capable, their cognition lacks self-awareness, intentionality, and moral grounding. AGI may someday match or exceed humans in narrow tasks, but replicating the richness of human reasoning - including empathy, ethics, and meaning - may require breakthroughs not just in engineering and neuroscience, but in our philosophical understanding of intelligence itself. As we build the next generation of AI, we must ask not just how well it performs in known tasks, but how gracefully it fails when facing the unknown. True progress lies in generalization and common sense, not just memorization. Let’s keep pushing the boundaries—but responsibly, and with a clear-eyed understanding of what current AI is and what it isn’t! #ai #artificialIntelligence #agi #responsibleai #aifutures #machinelearning #aiethics #reasoning #humanintelligence #topvoice #

On a personal note... I finally got the chance to meet the brilliant Professor Daniela Rus, Director of MIT CSAIL, and a contributing author to The Digitalist Papers, Vol. 2. For over a decade, I've followed her work in robotics and AI. See two slides from my AI presentation I gave in 2017 (AI TechWorld) below. It was a genuinely starstruck moment to shake her hand. Her recent essay, "Private Physical AI for the Edge," isn't just a technical discussion; it's a critical roadmap for the next phase of AI development and, potentially, the necessary step toward achieving true Artificial General Intelligence (AGI). Read the essay here: https://bit.ly/4qbfjWd The current era of AI, dominated by Large Language Models (LLMs), has achieved incredible fluency and creativity. However, as Professor Rus argues, this is only part of the intelligence puzzle. These models are "hungrier" for electricity than ever and lack a fundamental, built-in understanding of the real, physical world. Their intelligence is not grounded in physics. The work led by Professor Rus, as detailed in her essay, represents the vital shift to Private Physical Edge AI (often simply referred to as Physical AI). This paradigm moves intelligence away from sprawling, distant, energy-intensive data centers and places it directly on the devices that sense, decide, and act in the world... the "Edge." ⭐ Intelligence per Watt: It’s about building systems that are small, fast, and radically energy-efficient. This is crucial for sustainability and widespread deployment. ⭐ Grounded in Physics: Physical AI, as exemplified by breakthroughs like Liquid Neural Networks (LNNs) , is designed to be causal and physics-aware. They are inspired by the compact, adaptable brains of small species (like the C. elegans worm) and learn the task rather than just the context. ⭐ The AGI Connection: This causal, adaptable, and energy-efficient intelligence is what makes the work so critical for AGI. True general intelligence must be able to: - Understand the Physical World: It must reason about materials, motion, and uncertainty... the "messiness of the real world." - Generalize Zero-Shot: As Rus points out, LNNs can seamlessly transfer skills learned in one environment (e.g., a drone hiking in summer woods) to an entirely different one (the same task in winter or an urban setting) without retraining. This level of generalization is a hallmark of true intelligence, which current large models struggle with. By making AI compact, efficient, and inherently grounded in the world's physics, Professor Rus is helping to democratize intelligence and weave it seamlessly into the physical and social fabric of everyday life... from personal AI glasses for the visually impaired to decentralized energy grid management. This is a future where intelligence is measured not by trillions of parameters, but by intelligence per watt, and where it belongs to everyone. Thank you, Daniela Rus, for your groundbreaking vision!

𝑻𝒉𝒆 𝑷𝒆𝒂𝒌 𝑫𝒂𝒕𝒂 𝑪𝒉𝒂𝒍𝒍𝒆𝒏𝒈𝒆: 𝑾𝒉𝒂𝒕 𝑰𝒇 𝑾𝒆 𝑹𝒖𝒏 𝑶𝒖𝒕 𝒐𝒇 𝑻𝒓𝒂𝒊𝒏𝒊𝒏𝒈 𝑫𝒂𝒕𝒂 𝑩𝒆𝒇𝒐𝒓𝒆 𝑨𝑮𝑰? 🤔 Fascinating AI question to consider: What if we exhaust high-quality training data before achieving Artificial General Intelligence? Ilya Sutskever, OpenAI co-founder, has compared AI training data to fossil fuels - a finite resource that's rapidly being consumed. With models like GPT-4 already trained on substantial portions of internet text (estimated at ~100 petabytes total), this scenario deserves serious consideration. If we hit "peak data" before AGI: - Model improvements through traditional scaling approaches would face diminishing returns - Development timelines could extend significantly - Research priorities would shift toward efficiency rather than raw scale - Companies with proprietary data access might gain competitive advantages The industry is already exploring solutions: - Synthetic data generation (though this risks creating AI "echo chambers") - Transfer learning to maximize utility from limited datasets - Hybrid approaches combining neural networks with symbolic reasoning - Learning from human-AI interactions as a renewable data source This challenge could ultimately push AI development in more sustainable directions, prioritizing systems that learn efficiently from limited information - much like humans do. What do you think? Could data limitations become AI's biggest barrier, or will they inspire breakthrough innovations in how machines learn? #ArtificialIntelligence #AGI #MachineLearning #FutureOfTech #DataScience

Multimodality of AI for Education: Towards Artificial General Intelligence published at arxiv.org from Cornell University This paper presents a comprehensive examination of how multimodal artificial intelligence (AI) approaches are paving the way towards the realization of Artificial General Intelligence (AGI) in educational contexts. It scrutinizes the evolution and integration of AI in educational systems, emphasizing the crucial role of multimodality, which encompasses auditory, visual, kinesthetic, and linguistic modes of learning. This research delves deeply into the key facets of AGI, including cognitive frameworks, advanced knowledge representation, adaptive learning mechanisms, strategic planning, sophisticated language processing, and the integration of diverse multimodal data sources. It critically assesses AGI's transformative potential in reshaping educational paradigms, focusing on enhancing teaching and learning effectiveness, filling gaps in existing methodologies, and addressing ethical considerations and responsible usage of AGI in educational settings. The paper also discusses the implications of multimodal AI's role in education, offering insights into future directions and challenges in AGI development. This exploration aims to provide a nuanced understanding of the intersection between AI, multimodality, and education, setting a foundation for future research and development in AGI.   5️⃣ key takeaways from the study: #Multimodal AI in Education: The paper discusses the integration of multimodal artificial intelligence (AI) in educational contexts, highlighting its potential to achieve Artificial General Intelligence (AGI). #CognitiveFrameworks: It emphasizes the importance of cognitive frameworks, knowledge representation, and adaptive learning mechanisms in developing AGI for education. #StrategicPlanning: The study explores strategic planning and sophisticated language processing as crucial elements of AGI that can enhance teaching and learning effectiveness. #Ethical Considerations: Ethical, explainable, and responsible usage of AGI in educational settings is critically assessed, addressing the transformative potential and challenges. #Future Directions: The paper offers insights into future directions for AGI development, including the implications of multimodal AI’s role in education and the challenges ahead.

New on AI Snake Oil: Arvind Narayanan and I argue that AGI will not lead to rapid economic effects, the race to AGI is not relevant for great power competition, we won't know AGI when we have built it, and AGI does not imply impending superintelligence. In other words, AGI is not a milestone: https://lnkd.in/exDQbafU 1) Even if general-purpose AI systems reach some agreed-upon capability threshold, we will need many complementary innovations that allow AI to diffuse across industries to realize its productive impact. Diffusion occurs at human (and societal) timescales, not at the speed of tech development. 2) Worries about AGI and catastrophic risk often conflate capabilities with power. Once we distinguish between the two, we can reject the idea of a critical point in AI development at which it becomes infeasible for humanity to remain in control. 3) The proliferation of AGI definitions is a symptom, not the disease. AGI is significant because of its presumed impacts but must be defined based on properties of the AI system itself. But the link between system properties and impacts is tenuous, and greatly depends on how we design the environment in which AI systems operate. Thus, whether or not a given AI system will go on to have transformative impacts is yet to be determined at the moment the system is released. So a determination that an AI system constitutes AGI can only meaningfully be made retrospectively. 4) Businesses and policy makers should take a long-term view. Businesses should not rush to adopt half-baked AI products. Rapid progress in AI methods and capabilities does not automatically translate to better products. Building products on top of inherently stochastic models is challenging, and businesses should adopt AI products cautiously, conducting careful experiments to determine the impact of using AI to automate key business processes. A “Manhattan Project for AGI” is misguided on many levels. Since AGI is not a milestone, there is no way to know when the goal has been reached or how much more needs to be invested. And accelerating AI capabilities does nothing to address the real bottlenecks to realizing its economic benefits. We plan to keep writing on this topic, and have a series of essay planned on the theme of AI as Normal Technology. Follow the AI Snake Oil substack for more.

🧭 From AGI to ASI A new position report from a Google DeepMind team maps what AI progress could look like past human-level AGI: the possible routes to superintelligence, the things that could slow it down, and the questions nobody can answer yet. Most public forecasting is either one dramatic timeline or pure vibes. This is an attempt to structure the uncertainty instead. → AGI might not be a single "switch flips" moment. Even if base models stop getting smarter, you can still run far more copies, run them faster, and let them think for longer. The authors note effective compute has grown roughly 10x per year (better chips, more spending, and smarter algorithms stacking up). Pure quantity can start to look like new capability. → There are four plausible routes from AGI to superintelligence, and they can run in parallel: keep scaling, find a new algorithmic paradigm, let AI speed up its own research (the recursive loop everyone worries about), or have it emerge from massive groups of agents working together. No single bet required. → The "smartest possible" end has real theory behind it. They anchor it on Universal AI, and argue today's pretrain-then-finetune recipe is a rough, compute-limited approximation of it. → Superintelligence still won't be a god. It's stuck with physics (speed of light, energy per computation), real time (you cannot fast-forward a weather system or a slow biology experiment), and plain computational hardness. And knowing these limits exist tells us almost nothing about which specific dreams (curing aging, simulating a brain) are within reach. → The part I liked most: they catalog the brakes, not just the accelerators. Data, energy, real-world feedback, and an "abstraction barrier", the open question of whether high-bandwidth digital minds even build the deep abstractions humans lean on. I like this framing. Instead of one vivid timeline, you get a landscape of routes, brakes, and explicit open problems, which is a far better tool for reasoning about what might happen. There are still a lot of unknowns, but this is a useful artifact to think about what might happen next.

📝 Announcing our paper that proposes a unified cognitive and computational framework for Artificial General Intelligence (AGI) -- going beyond token-level predictions -- one that emphasizes modular reasoning, memory, agentic behavior, and ethical alignment 🔹 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬: 𝐅𝐫𝐨𝐦 𝐁𝐫𝐚𝐢𝐧‑𝐈𝐧𝐬𝐩𝐢𝐫𝐞𝐝 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐭𝐨 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐬 𝐟𝐨𝐫 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐆𝐞𝐧𝐞𝐫𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐚𝐧𝐝 𝐢𝐭𝐬 𝐒𝐨𝐜𝐢𝐞𝐭𝐚𝐥 𝐈𝐦𝐩𝐚𝐜𝐭 🔹 In collaboration with University of Central Florida, Cornell University, UT MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences, Toronto Metropolitan University, University of Oxford, Torrens University Australia, Obuda University, Amazon others. 🔹 Paper: https://lnkd.in/gqKUV4Mr ✍🏼 Authors: Rizwan Qureshi, Ranjan Sapkota, Abbas Shah, Amgad Muneer, Anas Zafar, Ashmal Vayani, Maged Shoman, PhD, Abdelrahman Eldaly, Kai Zhang, Ferhat Sadak, Shaina Raza, PhD, Xinqi Fan, Ravid Shwartz Ziv, Hong Yang, Vinija Jain, Aman Chadha, Manoj Karkee, @Jia Wu, Philip Torr, FREng, FRS, Seyedali Mirjalili ➡️ 𝐊𝐞𝐲 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬 𝐨𝐟 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬' 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞‑𝐂𝐨𝐦𝐩𝐮𝐭𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐆𝐈 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: 🧠 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: Integrates cognitive neuroscience, psychology, and AI to define AGI via modular reasoning, persistent memory, agentic behavior, vision-language grounding, and embodied interaction. 🔗 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧‑𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧: Critiques token-level models like GPT-4.5 and Claude 3.5, advocating for test-time adaptation, dynamic planning, and training-free grounding through retrieval-augmented agentic systems. 🚀 𝐑𝐨𝐚𝐝𝐦𝐚𝐩 𝐚𝐧𝐝 𝐂𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬: Proposes a roadmap for AGI through neuro-symbolic learning, value alignment, multimodal cognition, and cognitive scaffolding for transparent, socially integrated systems.