Sockeye Researcher Profile: Muhammad Abdul-Mageed
“Some of the work we’ve been able to do wouldn't even be possible without access to Sockeye. I hope UBC continues to invest in more digital research infrastructure.”
- Muhammad Abdul-Mageed, Assistant Professor, School of Information and Department of Linguistics
For Muhammad Abdul-Mageed, Assistant Professor, School of Information and Department of Linguistics, Sockeye has been a dream come true. As a member of the original UBC Digital Research and Compute Infrastructure committee, he has championed investment into digital research infrastructure, resulting in a high-performance computing platform with nearly 16,000 CPU cores and 200 GPUs, available to UBC researchers across all disciplines.
THE RESEARCH & ITS IMPACT
Muhammad’s research is in deep learning, the branch of artificial intelligence that has broken records on speech and language processing, machine translation, and image recognition. His work focuses on developing large-scale language models that can improve a wide range of technologies. This research requires specialized hardware such as graphical processing units (GPUs). Sockeye has been used in a number of Muhammad’s research projects, including the analysis of a billion-scale dataset of Tweets in 104 languages to understand the impact of COVID-19 on how humans communicate. He’s expanding this research to detect machine generated text. “We can build generative language models that are able to write language that is very realistic," he explains, noting that this research could help with security concerns and the reduction of political and health misinformation online.
Muhammad’s goal is to build `social’ machines for improved human health, safer social networking, and reduced information overload. He is also a member of the Royal Society of Canada’s Working Group on Language and Literacy, working with academic and industry to develop educational applications of his research at home and abroad. He mentions his recent work on Arabic and other African language translation. “People don’t have access to the same level of material that other countries do, because there’s no translation available for medication, education, news and other valuable information.”
HOW SOCKEYE COMES INTO PLAY
Muhammad’s research involves large volumes of simulations to train language models in deep learning and natural language socio-pragmatics. “Very big models like ours take a lot of time to train. Sockeye helped us model faster and get results in shorter periods of time - weeks rather than months,” Muhammad explains.
Muhammad is thankful for the access that Sockeye has provided. “Some of the work we’ve been able to do wouldn't even be possible, without access to Sockeye. I hope UBC continues to invest in more digital research infrastructure.”
For more information on UBC ARC Sockeye and other services provided by Advanced Research Computing, please visit the UBC ARC Services page here.
Muhammad Abdul-Mageed is an Assistant Professor with the UBC School of Information and Department of Linguistics.