Uploaded January 2026 | Updated September 2026, 47 minutes ago
At AGI-25, Faezeh Habibi describes how objectives shape what neural networks learn and how to design a better objective for human-level intelligence. Recorded at the 18th Annual Artificial General Intelligence Conference (AGI-25) at Reykjavík University, Iceland, on August 12, 2025.
Are we scaling the wrong thing if AGI still isn’t here? Faezeh Habibi, AI researcher at SingularityNET and Ph.D. student of the Neural Adaptive Computing Laboratory (NAC Lab), argues that the missing ingredient of human-level AGI is world modeling, a core feature of biological intelligence that today’s deep learning doesn’t build in; the reason why scaling is not the solution.
Habibi contrasts this framework with contemporary AI architectures, arguing that most deep learning models discard intermediate prediction errors and rely primarily on output loss. This simplification, she suggests, limits deep learning model's ability to make world models and explains why current AI systems require massive scale to approximate capabilities that biological brains achieve with far fewer resources.
Revisit all AGI-25 keynote presentations from industry leaders like Ben Goertzel, Richard Sutton, and Joscha Bach, technical sessions, and workshop recordings: youtube.com/playlist?list=PLAJnaovHtaFQZIZqNxIaBaNrFpCLmMPWx
Explore the conference proceedings on Springer:
- Part 1: link.springer.com/book/10.1007/978-3-032-00686-8
- Part 2: link.springer.com/book/10.1007/978-3-032-00800-8
#AGI25 #AGI #AI
----
SingularityNET was founded by Dr. Ben Goertzel with the mission of creating a decentralized, democratic, inclusive, and beneficial Artificial General Intelligence (AGI). An AGI is not dependent on any central entity, is open to anyone, and is not restricted to the narrow goals of a single corporation or even a single country.
The SingularityNET team includes seasoned engineers, scientists, researchers, entrepreneurs, and marketers. Our core platform and AI teams are further complemented by specialized teams devoted to application areas such as finance, robotics, biomedical AI, media, arts, and entertainment.
Website: singularitynet.io
LinkedIn: linkedin.com/company/singularitynet
X: https://x.com/SingularityNET
Instagram: instagram.com/singularitynet.io
Discord: discord.gg/snet
Telegram: https://t.me/singularitynet
ASI Alliance on X: https://x.com/ASI_Alliance
ASI Alliance on LinkedIn: linkedin.com/company/artificial-superintelligence-alliance
At AGI-25, Faezeh Habibi describes how objectives shape what neural networks learn and how to design a better objective for human-level intelligence. Recorded at the 18th Annual Artificial General Intelligence Conference (AGI-25) at Reykjavík University, Iceland, on August 12, 2025.
Are we scaling the wrong thing if AGI still isn’t here? Faezeh Habibi, AI researcher at SingularityNET and Ph.D. student of the Neural Adaptive Computing Laboratory (NAC Lab), argues that the missing ingredient of human-level AGI is world modeling, a core feature of biological intelligence that today’s deep learning doesn’t build in; the reason why scaling is not the solution.
Habibi contrasts this framework with contemporary AI architectures, arguing that most deep learning models discard intermediate prediction errors and rely primarily on output loss. This simplification, she suggests, limits deep learning model's ability to make world models and explains why current AI systems require massive scale to approximate capabilities that biological brains achieve with far fewer resources.
Revisit all AGI-25 keynote presentations from industry leaders like Ben Goertzel, Richard Sutton, and Joscha Bach, technical sessions, and workshop recordings: youtube.com/playlist?list=PLAJnaovHtaFQZIZqNxIaBaNrFpCLmMPWx
Explore the conference proceedings on Springer:
- Part 1: link.springer.com/book/10.1007/978-3-032-00686-8
- Part 2: link.springer.com/book/10.1007/978-3-032-00800-8
#AGI25 #AGI #AI
----
SingularityNET was founded by Dr. Ben Goertzel with the mission of creating a decentralized, democratic, inclusive, and beneficial Artificial General Intelligence (AGI). An AGI is not dependent on any central entity, is open to anyone, and is not restricted to the narrow goals of a single corporation or even a single country.
The SingularityNET team includes seasoned engineers, scientists, researchers, entrepreneurs, and marketers. Our core platform and AI teams are further complemented by specialized teams devoted to application areas such as finance, robotics, biomedical AI, media, arts, and entertainment.
Website: singularitynet.io
LinkedIn: linkedin.com/company/singularitynet
X: https://x.com/SingularityNET
Instagram: instagram.com/singularitynet.io
Discord: discord.gg/snet
Telegram: https://t.me/singularitynet
ASI Alliance on X: https://x.com/ASI_Alliance
ASI Alliance on LinkedIn: linkedin.com/company/artificial-superintelligence-alliance
