Andrew Dai

I am a Researcher at Sakana AI. My research background spans Large Language Models (LLMs), Quality Diversity (QD), and open-endedness. I also worked as an AI Researcher at Aleph Alpha.

My humble beginnings started with Mechanical Engineering in Trinity College Dublin, where I co-founded Formula Trinity Autonomous with amazing close friends and teammates.

After reading Ken and Joel's book in 2021 with close friends, I've learned that ambitious objectives can deceive us from greatness, inspiring me on a collaborative journey to reflect on the way that I (and artificial systems that I build with my team) can "achieve our highest goals by being willing to abandon them", ultimately making us more creative.

GitHub  /  Google Scholar  /  Twitter  /  LinkedIn

profile photo

Research

Authors marked with an asterisk (*) denote equal contribution first authors. Representative works are highlighted.

project image

Surveying the Effects of Quality, Diversity, and Complexity in Synthetic Data From Large Language Models


Alex Havrilla, Andrew Dai, Laura O'Mahony, Koen Oostermeijer, Vera Zisler, Alon Albalak, Fabrizio Milo, Sharath Chandra Raparthy, Kanishk Gandhi, Baber Abbasi, Duy Phung, Maia Iyer, Dakota Mahan, Chase Blagden, Srishti Gureja, Mohammed Hamdy, Wen-Ding Li, Giovanni Paolini, Pawan Sasanka Ammanamanchi, Elliot Meyerson
arXiv pre-print, 2024
arXiv

Investigates quality, diversity, and complexity (QDC) of LLM synthetic data (data evaluation/downstream model effects, and data generation methods)

project image

From Text to Life: On the Reciprocal Relationship between Artificial Life and Large Language Models


Eleni Nisioti*, Claire Glanois*, Elias Najarro, Andrew Dai, Elliot Meyerson, Joachim Winther Pedersen, Laetitia Teodorescu, Conor F. Hayes, Shyam Sudhakaran, Sebastian Risi
ALIFE, 2024
arXiv

Investigates how LLMs can serve as tools in ALife research and how ALife principles can enhance LLM development.

project image

Quality-Diversity through AI Feedback


Herbie Bradley*, Andrew Dai*, Hannah Teufel, Jenny Zhang, Koen Oostermeijer, Marco Bellagente, Jeff Clune, Kenneth Stanley, Grégory Schott, Joel Lehman
ICLR, 2024
project page / arXiv / code / blog

QDAIF utilizes LMs to both vary texts and evaluate subjective quality/diversity of creative texts, enabling search of diverse high-quality solutions in vast spaces of subjectivity.

project image

MultiFusion: Fusing Pre-Trained Models for Multi-Lingual, Multi-Modal Image Generation


Marco Bellagente*, Manuel Brack*, Hannah Teufel*, Felix Friedrich, Björn Deiseroth, Constantin Eichenberg, Andrew Dai, Robert Baldock, Souradeep Nanda, Koen Oostermeijer, Andres Felipe Cruz-Salinas, Patrick Schramowski, Kristian Kersting, Samuel Weinbach
NeurIPS, 2023
project page / arXiv

MF image-gen system enables expression of complex, nuanced concepts with arbitrarily interleaved inputs of multiple modalities and languages.

project image

Interpretable Video Transformers in Imitation Learning of Human Driving


Andrew Dai, Wenliang Qiu, Bidisha Ghosh
3rd ICML 2021 Workshop on Human in the Loop Learning, 2021

ViTs applied to behavioural cloning for AVs from offline video data, lending itself to more interpretable decision-making.

Other Projects

These include coursework, side projects and unpublished research work.

project image

Neural Radiance Fields (and Friends)


TCD CS7GV4: Augmented Reality
2021-07-21
code / video / report

Applying advancements in NeRF (e.g. NeX-MPI, PlenOctrees) to AR demos within a home (and notebook).

project image

Formula Trinity Autonomous


TCD Extracurricular
2020-09-01
tutorial

Autonomous Racing (Formula Student Driverless racing, F1TENTH) is cool - co-founded with Jakub Pyszka, Senan Stanley, Katherine Hardgrave, and curious TCD students

project image

Ground and UAV Robot Sim Control


NUS EE4308: Autonomous Mobile Robots
2020-05-11
code / report

Control, state estimation, perception, path planning on sim robots. Project done while on exchange in Year 4.


Design and source code from Jon Barron's website. Jekyll Now fork of design from Leonid Keselman