Topic: How Far We Are from Real Artificial Intelligence

Speaker: Prof. Matti Pietikäinen, University of Oulu, Finland

Time: Thursday, August 15, 2024, 10:00 a.m.

Venue: G849, New Main Building, Xueyuan Road Campus

Abstract:

Eleven years ago, a breakthrough occurred in AI and ML with deep learning neural networks due to increased training data, advanced graphics processors, and better algorithms, enabling diverse AI applications. Deep neural networks with hundreds of layers demand substantial energy for training. Energy usage for teaching these networks reportedly surged tenfold yearly in the 2010s. The natural language Al model of Google requires about a thousand trillion calculations.

Human brains differ significantly with 86 billion neurons for parallel processing and cognitive functions. The neural network of Digital Reasoning contains up to 160 billion neurons. Animals exhibit remarkable functions surpassing human capabilities. Ants, with 250,000 neurons, build communities. Flies, with 150,000 neurons, learn from trial and error. Dragonflies navigate with minimal latency using four-layer networks. Human brains consume 25 watts, akin to a small light bulb.

Current AI falls short of human-like intelligence, lacking multitasking abilities and integrating data-driven and symbolic reasoning. Machines should learn continuously akin to children, necessitating new neural network approaches combining symbolic and data-driven AI. Dense wireless networks aid environment sensing, with energy-efficient, distributed AI promising smarter AI solutions.

About the Speaker:

Matti Pietikäinen is currently Professor (emer.) in the Center for Machine Vision and Signal Analysis, University of Oulu, Finland. He was Director of the Center for Machine Vision Research, and Scientific Director of Infotech Oulu. He has authored over 350 refereed scientific publications. He has made pioneering contributions to local binary patterns (LBP) methodology, texture-based image and video analysis, and facial image analysis. He has been Associate Editor of IEEE TPAMI, PR, IEEE TIFS, etc. He is a Fellow of lEEE, IAPR, and AAIA. In 2018, he received the lAPR's King-Sun Fu Prize. He was named a Highly Cited Researcher by Clarivate Analytics in 2018. Since February 2023 he will be listed by Webometrics among the Highly Cited Researchers whose h-index is at least 100.

School of Computer Science and Engineering