TURK 2025

Recent advances in deep learning have catalyzed the widespread adoption of artificial intelligence (AI) across diverse fields. However, as traditional silicon CMOS technologies approach their physical scaling limits, further improvements in AI performance are becoming increasingly constrained. In response, alternative hardware paradigms are being explored, particularly those that physically embody artificial neural networks (ANNs) through neuromorphic and brain-inspired architectures. These emerging "AI systems" integrate dynamic functionalities—such as pulse generation and noise—by leveraging the inherent nonlinearity of nanomaterials[1,2]. Such dynamics are essential for achieving low-power, highly integrated hardware platforms. A key challenge in realizing efficient ANNs is the continuous adjustment and retention of synaptic weights during learning. At our research center, we are addressing this by developing novel materials compatible with CMOS, including molecular systems and nanocarbon-based devices[3-6]. The material not only support energy-efficient information processing but also hold promise for autonomous AI robotic platforms. This paper reviews our recent progress in nanomaterial-based AI devices, focusing on network construction methods, performance metrics, and practical implementations. Selected experimental results are also highlighted.[7-18] Additionally, It will be also introduced how to use the technique not only to apply to robots but to detect human activities.

Acknowledgments: HT would like to thank to Prof. J. Gimzewski of UCLA for fruitful discussion on reservoir device measurement.
References:
[1] H. Tanaka et al., Adv. Mater. 18, 1411 (2006).
[2] H. Tanaka et al., Nat. Commun. 9, 2693 (2018). The article was selected as the most read 50 articles published in Nat. Commun. in 2018 (Physics).
[3] D. Banerjee, H. Tanaka et al., Adv. Intell. Syst. 4, 2100145 (2022).
[4] S. Azhari, H. Tanaka et al., IEEE sensors J. 21, 27810, (2021).
[5] K. Kimizuka, H. Tanaka et al., JP patent application 2022-071679, and PCT application (2022).
[6] K. Kimizuka, H. Tanaka et al., Adv. Intell. Syst. 7, 2400640 (2025).
[7] T. Kotooka, H. Tanaka et al., Adv. Elect. Mater. 2400443 (2024).
[8] Hadiyawarman, H. Tanaka et al., Jpn. J. Appl. Phys. 60, SCCF02 (2021).
[9] O. Srikimkaew, H. Tanaka et al., ACS Appl. Elec. Mater. 6, 688 (2024).
[10] T. T. Dang, H. Tanaka et al., Appl. Phys. Lett. 124, 091903 (2024).
[11] O. Srikimkaew, H. Tanaka et al., Adv. Elect. Mater. 10, 2470033 (2024).
[12] O. Srikimkaew, H. Tanaka et al., Adv. Elect. Mater. 10, 2470039 (2024).
[13] Kyutech team won the RoboCup World Series of Domestic Standard Platform League by TOYOTA HSR in 2017, 2018 and 2024, and World Robot Summit in 2018, 2021 and 2024. The same robot was used in this work.
[14] Y. Usami, H. Tanaka et al., Adv. Mater. 33, 2102688 (2021).
[15] H. Tanaka et al., Neuromorph. Comput. Eng. 2, 022002 (2022).
[16] T. Kotooka, H. Tanaka et al., Appl. Phys. Express, 16, 014002 (2023).
[17] Y. Tanaka, H. Tanaka et al., IEEE Int’l Symposium on Circuits and Systems (ISCAS2023), 2056, (2023).
[18] M. Desu, H. Tanaka et al., Jpn. J. Appl. Phys. 64, 04SP12 (2025).