Why a Computer Scientist Questions Whether Artificial Superintelligence Is Possible
- Published
- Source
- IEEESpectrumAI
The article examines whether artificial superintelligence (ASI) is technically possible, focusing on recursive self-improvement (RSI)—the proposed cycle in which an AI improves itself at an accelerating rate. The author, drawing on a computer science doctoral seminar at Georgetown University, argues that RSI remains speculative and identifies engineering obstacles. Anthropic has reported that AI is accelerating AI development, while OpenAI has stated a goal of developing an automated researcher by March 2028; the article questions whether these efforts lead to uncontrollable self-improvement.
A central constraint is the availability of training data. After model scaling produced diminishing returns following GPT-4, developers increasingly used post-training methods, including reinforcement learning, which require many examples of relevant inputs paired with correct outputs. Such data is easier to generate in structured fields such as programming and mathematics, where AI progress has been strong, than in less structured tasks. The article cites French AI researcher François Chollet’s hypothesis that human-generated data may bottleneck progress beyond math and code.
OpenAI’s stated March 2028 goal of developing an automated researcher is a specific milestone, but the article does not establish that it would produce RSI or ASI. It argues that automated coding agents and LLM-powered researchers will not overcome the underlying data limitation. The author also warns that pursuing superintelligence could cause real harm even if an intelligence explosion never occurs.
Original report
Why a Computer Scientist Thinks Superintelligence Won’t Happen
IEEESpectrumAI · Cal NewportThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
Related Suppliers
Anthropic
AI safety and research company and public benefit corporation that builds the Claude family of large language models.
OpenAI
US AI research lab and product company founded 2015; develops GPT, DALL-E, and robotics/embodied AI foundation models targeting physical-world AGI.
Versatile Robotics
Wessling, Germany DLR spinoff founded 2024; builds patented MPTC controller software for safe, high-speed human-robot collaboration.
EZ-Robot
Canadian educational robotics company founded 2011 in Calgary, offering JD Humanoid and Six Hexapod kits plus the ARC AI platform for STEM learners.
Hyodol
Korean AI companion robot for elderly, providing 24/7 health monitoring, medication reminders, and emergency alerts via a huggable doll form factor.
Aether AI
Seed-stage San Diego startup building causal world models that predict action outcomes, with robotics as its first target market.
Tau Robotics
Palo Alto AI robotics startup building general-purpose robot learning systems that enable robotic arms to acquire skills autonomously in the real world.
Assured Robot Intelligence
San Diego robot-AI start-up founded by Lerrel Pinto and Xiaolong Wang, building humanoid foundation models; acquired by Meta in May 2026.
Twolabs Inc.
Y Combinator Spring 2026 startup in San Francisco building Tobi, a modular humanoid for caregiving tasks in nursing homes and senior living.
GrayMatter Robotics
US AI robotics startup delivering physics-informed robotic cells for surface finishing — sanding, grinding, polishing — in high-mix manufacturing.
Learning Robots
Learning Robots is a Paris-Saclay CNRS spin-off making AlphAI, an educational robot and software for teaching artificial intelligence.
BridgeDP Robotics 桥介数物
Shenzhen developer of universal robot 'cerebellum' motion-control algorithms for humanoid, quadruped and wheeled-legged robots using RL and motion capture.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.