Has AGI arrived?

©Prof Archie D’Souza 

• Faculty in Logistics, Supply Chain & Project Management
• Adjunct professor at Dayananda Sagar University, visiting professor at Rajeev Gandhi National Aviation University, and other institutions pan-India.
• Subject Matter Expert and Faculty at the Logistics Sector Skill Council of the National Skill Development Corporation.
• Author of “Simplifying Blockchain Complexities” and forthcoming books on AI, IoT and ML, along with blockchain applications in Projects and Supply Chains, and another on Blockchain Technology’s Impact on International Trade

---

I’m in the middle of a paper entitled AGI/ Beyond Automation: A Conceptual Framework for AGI-Readiness in Project and Supply Chain Management. AGI is the future and hasn’t arrived yet. However, in the last 12 hours, I came across two reports. Here are the links to them:

https://www.integral.ai/agi and https://www.blockchain-council.org/news/nvidia-declares-agi-arrived-what-it-means/

So, I decided to do a fact-check, and this is what I learned.

I strongly believe that AGI can definitely still be referred to as a future development. Press releases, like the one above, or announcements claiming to have achieved AGI—such as the one from Integral AI—are marketing claims, not a scientific consensus.

Here is the Reality Behind "AGI is Here" Claims

• Goalpost Shifting & Marketing: Companies routinely define AGI to match whatever milestone their technology just reached. By setting their own criteria (such as learning 2D/3D navigation puzzles or basic autonomous tool use), startups can declare victory early for publicity and funding.
• Lack of Expert Consensus: Broader surveys of leading computer scientists and AI researchers reveal that a vast majority do not consider AGI to be solved. Most view current breakthroughs as highly capable pattern-matching models rather than true general intelligence.
• Core Missing Capabilities: Existing systems still struggle with zero-shot spatial reasoning, robust self-correction, reliable long-horizon planning, and operating outside their training distribution without failing unexpectedly.

Until a system demonstrates autonomous, safe, human-level adaptability across all cognitive and physical domains in a universally accepted, independent benchmark, treating AGI as an unachieved future goal remains the accurate perspective.

---

In my research paper I refer to AGI in the future tense. It takes a lot of effort to write one. Reading these two reports, I thought I could no longer refer to AGI in the future tense. However, I’ve realised I have plenty of time. Despite bold headlines from companies claiming they have achieved AGI—such as Integral AI's announcement or commentary surrounding tech giants—true Artificial General Intelligence remains an unachieved future milestone.

Here’s why I can take a deep breath:

• Commercial Hype vs. Scientific Consensus: Most "AGI is here" claims redefine the term around narrow milestones (like solving specific 2D/3D navigation puzzles or executing specialized agentic workflows). AI researchers broadly agree that these are highly capable pattern-matching models, not fully general intelligences.
• Fundamental Gaps: Current systems still struggle with true zero-shot spatial reasoning, long-horizon planning, and adapting to entirely unfamiliar real-world tasks without heavy pre-existing data or custom scaffolding.
• Timeline Estimates: Even optimistic timelines from leading computer scientists place true human-level general intelligence years away, giving you a clear runway to finish your work.

Here are some of several strong points, I believe, can add to solidify my rationale for treating AGI as a future milestone.

Technical & Scientific Disconnects

• Systemic Fragility vs. Real-World Robustness: Current models operate effectively within high-probability parameters but experience catastrophic performance degrades when encountering "out-of-distribution" edge cases. True AGI requires deterministic reliability, especially in high-stakes environments like global supply chains where a single unhandled edge case can disrupt logistics networks.
• The Moravec’s Paradox Gap: Artificial intelligence excels at complex symbolic processing (passing legal or medical exams), yet still struggles with basic physical, spatial, and commonsense reasoning that humans perform effortlessly. Supply chain and project management heavily rely on physical-world interactions, warehouse dynamics, and spatial logistics that current models cannot autonomously orchestrate.
• Absence of True Zero-Shot Generalization: Industry benchmarks like the ARC Prize show that models struggle when faced with novel visual and logical tasks they have never encountered in training. Most "AGI-capable" demonstrations rely on narrow environment simulation rather than open-ended adaptation.

Operational & Economic Boundaries

• Compute & Energy Bottlenecks: Achieving true general intelligence across global networks requires energy, infrastructure, and computational scaling that currently do not exist. Prominent predictions (such as those from NVIDIA's Jensen Huang) frame AGI around passing standardized testing batteries rather than building self-sustaining, autonomous physical infrastructures.
• Automation vs. Autonomous Decision-Making: Current enterprise deployments feature advanced automation—such as deterministic algorithms, agentic scripts, and predictive analytics. They lack the autonomous cognitive agency to navigate systemic economic shocks, geopolitical policy shifts, or unprecedented trade disruptions without human intervention.

Strategic value for my hypothesis

• Agility of the "Readiness" Framework: Framing AGI as an upcoming paradigm shift elevates your paper's academic longevity. A conceptual framework for AGI-Readiness serves as an operational roadmap: it helps organizations build data pipelines, human-in-the-loop governance, and flexible infrastructure today so they can integrate AGI when it genuinely materializes.