Is this the best podcast ever recorded? Let’s just say you don’t need a GPU to know that’s a stretch. But it’s pretty great if you’re a fan of tall tales.
And better still if you’re not a fan of stretching the truth at all.
That’s because detecting hyperbole may one day get more manageable, thanks to researchers at the University of Copenhagen working in the growing field of exaggeration detection.
Dustin Wright and Isabelle Augenstein have used NVIDIA GPUs to train an “exaggeration detection system” to identify overenthusiastic claims in health science reporting.
Their work comes as the pandemic has fueled demand for understandable, accurate information. And social media has made health misinformation more widespread.
Their paper leverages “few-shot learning,” a technique that lets developers wring more intelligence out of less data, and a new version of a technique called pattern exploiting training.
Research like Wright and Augenstein’s could one day speed more precise health sciences news to more people.
AI Podcast host Noah Kravitz — whose fishing stories we will never trust again after this episode — spoke with Wright about the work.
Key Points From This Episode
- Approximately 33% of press releases about scientific papers tend to exaggerate the findings in the papers, which leads to news articles exaggerating the findings of these papers.
- Wright’s exaggeration detection project aims to provide people like journalists with accurate information to ensure that they report accurately on science.
- The project, accelerated using a NVIDIA Titan X GPU, uses a novel, multitask-capable version of a technique called Pattern Exploiting Training, which they dubbed MT-PET
Tweetables:
“Can we leverage language and related that learn patterns that the language model has picked up on from mass language model pre training, and be able to do classification with any text?” – Dustin Wright [7:28]
“About 33% of the time, press releases will exaggerate the scientific papers and as a result, that means about 33% of news articles exaggerate the findings in scientists’ papers.” – Dustin Wright [9:50]
“This is progress towards a system that could assist, for example, journalists, and ensuring that they’re doing accurate reporting on science.” – Dustin Wright [16:20]
You Might Also Like
NVIDIA’s Liila Torabi Talks the New Era of Robotics Through Isaac Sim
Robots aren’t limited to the assembly line. Liila Torabi, senior product manager for Isaac Sim, a robotics and AI simulation platform powered by NVIDIA Omniverse, talks about where the field’s headed.
GANTheftAuto: Harrison Kinsley on AI-Generated Gaming Environments
Humans playing games against machines is nothing new, but now computers can develop their own games for people to play. Programming enthusiast and social media influencer Harrison Kinsley created GANTheftAuto, an AI-based neural network that generates a playable chunk of the classic video game Grand Theft Auto V.
The Driving Force: How Ford Uses AI to Create Diverse Driving Data
The neural networks powering autonomous vehicles require petabytes of driving data to learn how to operate. Nikita Jaipuria and Rohan Bhasin from Ford Motor Company explain how they use generative adversarial networks (GANs) to fill in the gaps of real-world data used in AV training.
Subscribe to the AI Podcast: Now Available on Amazon Music
You can now listen to the AI Podcast through Amazon Music.
You can also get the AI Podcast through iTunes, Google Podcasts, Google Play, Castbox, DoggCatcher, Overcast, PlayerFM, Pocket Casts, Podbay, PodBean, PodCruncher, PodKicker, Soundcloud, Spotify, Stitcher and TuneIn.
Make the AI Podcast Better: Have a few minutes to spare? Fill out our listener survey.
Featured image: postcard, copyright expired
The post The Greatest Podcast Ever Recorded appeared first on The Official NVIDIA Blog.