Probabilistic computer prototype; photonic AI chiplet; observing changes in 2D materials.
Researchers at Tohoku University and the University of California Santa Barbara created a prototype of a heterogeneous probabilistic computer that combines a CMOS circuit with a limited number of stochastic nanomagnets.
It aims to improve the execution of probabilistic algorithms used to solve problems where uncertainty is inherent or where an exact solution is computationally infeasible, such as in AI and machine learning.
“Our constructed prototype demonstrated that excellent computational performance can be achieved by driving pseudo random number generators in a deterministic CMOS circuit with physical random numbers generated by a limited number of stochastic nanomagnets,” said Shunsuke Fukami, a professor at Tohoku University, in a statement. “Specifically speaking, a limited number of probabilistic bits (p-bits) with a stochastic magnetic tunnel junction (s-MTJ), should be manufacturable with a near-future integration technology.” [1]
Engineers from Tsinghua University and the Beijing National Research Center for Information Science and Technology built a large-scale photonic AI chiplet and distributed optical computing architecture that provides up to 160 TOPS/W.
Alongside the high parallelism and high connectivity of wave optics, the chiplet explores a general and iterative encoding-embedding-decoding photonic computing to effectively increase the scale of the optical neural network to the billion neuron level for thousand-category-level classification and generative AI. [2]
Researchers at the University of Illinois Urbana-Champaign have developed a method to visualize the thermally-induced rearrangement of 2D materials, atom-by-atom, from twisted to aligned structures using transmission electron microscopy (TEM). They investigated bilayer transition metal dichalcogenides, which could have applications in future electronics.
Even a few degrees of twist between the two layers can change the behavior. Understanding how rearrangement happens can help tune the interfacial alignment at the nanoscale, noted Pinshane Huang, a materials science and engineering professor at the University of Illinois Urbana-Champaign, in a release. “It is impossible to underscore how excited people are about that tuneability. The macroscopic twist between the two layers is a really important parameter because as you rotate one on the other, you can actually change the properties of the entire system. For example, if you rotate the 2D material graphene to a specific angle, it becomes superconducting. For some materials, if you rotate them, you change the bandgap which changes the color of light it absorbs and what energy of light it emits. All of those things you change by altering the orientation of atoms between layers.”
To observe the changes, the team encapsulated the twisted bilayer in graphene. It was then put on a chip that could be heated and cooled quickly. To capture the fast atomic dynamics, the sample underwent half second heat pulses between 100-1000°C. After each pulse, the team would look at where the atoms were using TEM and then repeated the process.
“You can actually watch the system as it changes, as the atoms settle in from whatever configuration they were put in initially, to the configuration that is energetically favorable, that they want to be in,” Huang said in the release. “That can help us understand both the initial structure as it is fabricated and how it evolves with heat.” [3]
[1] Singh, N.S., Kobayashi, K., Cao, Q. et al. CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning. Nat Commun 15, 2685 (2024). https://doi.org/10.1038/s41467-024-46645-6
[2] Zhihao Xu et al., Large-scale photonic chiplet Taichi empowers 160-TOPS/W artificial general intelligence. Science 384, 202-209 (2024). https://doi.org/10.1126/science.adl1203
[3] Yichao Zhang et al., Atom-by-atom imaging of moiré transformations in 2D transition metal dichalcogenides. Sci. Adv. 10, eadk1874 (2024). https://doi.org/10.1126/sciadv.adk1874
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply