Projects & Publications

The goal of my work is to use all data available to the modern neuroscience lab to uncover the mysteries of complex thought and translate our findings into more powerful artificial agents. My empirical work is currently mostly based on neuroimaging data of the developing brain, large scale data of human decision making and single cell recordings in non-human primates. My modelling work currently focuses on artificial neural network models, using a combination of supervised training, reinforcement learning, and biologically inspired wiring rules.

Many of my current ideas on this topic are reviewed in the paper: Building artificial neural circuits for domain-general cognition: a primer on brain-inspired systems-level architecture https://arxiv.org/abs/2303.13651 

Ongoing

• We recently introduced the new spatially-embedded Recurrent Neural Network model which can recapitulate numerous neuroscientific findings, from large scale connectome structure down to functional codes used by single neurons. We are in the process of expanding this model with further biophysical constrains and a theoretical analysis of dynamics for optimising network communication. 

  • See details of project here

• Uncovering the neuronal algorithms underlying structural inference and reasoning in a large population of PFC neurons. 

  • No preprint expected until 2024. Get in touch if you want to know more.

• General generative AI: Together with the Cognitive AI Lab at Intel Labs I am working on methods to improve generative AI agents to develop their domain-general problem solving skills. Some of our work on generative image algorithms is described in this this preprint and our related open-source project.

• Upcoming workshop on representational alignment of brain and AI: I am part of the group of researchers developing a proposal for a NeurIPS workshop on aligning representations in brains and AI. If this topic is of interest to you, check out our workshop webpage (https://representational-alignment.github.io/).

Selected publications

• Artificial neural networks: Achterberg, J., Akarca, D., Strouse D., Duncan, J. & Astle, D (2022, preprint). Spatially-embedded recurrent neural networks reveal widespread links between structural and functional neuroscience findings. bioRxiv. doi: 10.1101/2022.11.17.516914

  • I gave a seminar on this preprint at SNUFA: https://www.youtube.com/watch?v=DaG92jBCu68&t 

• Highly functional biological networks: Achterberg, J., Kadohisa, M., Watanabe, K., Kusunoki, M., Buckley, M., & Duncan, J. (2021). A One-Shot Shift from Explore to Exploit in Monkey Prefrontal Cortex. The Journal Of Neuroscience, 42(2), 276-287. doi: 10.1523/JNEUROSCI.1338-21.2021 

• Human decision making: Ruggeri, K., Panin, A., Vdovic, M., Veckalov, B., Abdul-Salaam, N, Achterberg, J., [...], & Garcia-Garzon, E. (2022). The globalizability of temporal discounting. Nature Human Behaviour doi: 10.1038/s41562-022-01392-w 

All academic publications

A complete list of my academic publications is available on my Google Scholar profile.

Other publications

• "Domain-general cognition in brains and artificial neural networks". Jascha Achterberg. Presentation and interview as part of the Researcher App Series on AI in Neuroscience 2023. Recording available on: https://www.researcher-app.com/paper/14604093

• “Intelligence in brain and machines”. Jascha Achterberg. Short video for the Cambridge Science Festival 2021. Hosted on YouTube.

• “Want to Build Intelligent Machines? Mind the Brain!”. Jascha Achterberg. Article in the 2020 version of Gates Cambridge’s magazine “The Scholar”