Revolutionizing Algorithmic Reasoning: DeepMind’s Journey at ICML 2022

Petar Veličković, a research scientist at DeepMind, will be presenting his paper The CLRS Algorithmic Reasoning Benchmark at ICML 2022. Throughout his undergraduate courses at the University of Cambridge, Petar was intrigued by the limitations of deep learning systems. His interest in AI was sparked when DeepMind’s ground-breaking gameplay in a challenge match against a world-renowned Go player made seemingly impossible things possible.

Now, as a research scientist at DeepMind, Petar’s role involves continually learning, researching, communicating, and advising. He takes active steps to broaden his knowledge and brainstorm ideas with his team to positively impact the world. He also spends time presenting their work at conferences like ICML and co-advising students.

At ICML, Petar and his team will present their research on The CLRS algorithmic reasoning benchmark, which aims to support and enrich efforts in the emerging area of neural algorithmic reasoning. The benchmark aims to improve on the challenges of current research by providing publicly available code and dataset generators to track progress in the field.

Looking to the future, Petar’s dream is to capture classical algorithms inside high-dimensional neural executors, enabling the deployment of these executors over raw or noisy data representations. This approach has the potential to enable data-efficient reinforcement learning, making the application of strong classical algorithms possible in standard environments.

As he looks ahead, Petar is excited about the ICML Workshop on Human-Machine Collaboration and Teaming, where he will discuss the broader implications of AI in collaboration with human domain experts alongside his fellow panelists.

Finally, Petar emphasizes the importance of diversity and inclusion in the field of AI, and is committed to outreach opportunities in Eastern Europe and Africa to empower future AI leaders and strengthen the local AI communities. He recently co-authored a proto-book and taught a course on Geometric Deep Learning at the African Master’s in Machine Intelligence, where he has met several students who have gone on to join DeepMind for internship positions. He also participated in organizing the EEML Serbian Machine Learning Workshop, with the hope of strengthening the local AI community in Eastern Europe.

Source link

Stay in the Loop

Get the daily email from AI Headliner that makes reading the news actually enjoyable. Join our mailing list to stay in the loop to stay informed, for free.

Latest stories

You might also like...