A memory of heat

We are witnessing a truly remarkable moment: the birth of an artificial intelligence. Large language models, machine learning and deep fakes are all running on distributed computing networks that use a tremendous amount of energy. For example, the training of chat-GPT used up approximately 1 GWh of energy. The projections are that at the current rate of development AI would require an energy equivalent to our current global energy consumption by 2030.

This is clearly unsustainable and the search for alternative computing technology is speeding up. With an average energy consumption of only 12 W, the brain is a great place to look for inspiration and researchers started to develop materials that can mimic components such as neurons. 

In this project, you will work on the latest development in imaging ‘spiking’ of artificial neurons on the nanoscale. We recently demonstrated that we can track the memory effect in a functioning memristor on the nanoscale [1]. Your role will be to pinpoint the origin of this effect by direct measurement of heat dissipation in these devices.

[1]: Sergio Salvia Fernandez et al. Near-field imaging of domain switching in in-operando VO2 devices