If you could use machine learning to summarize arguments in scientific papers so that a seven-year-old could understand them, what would academic writing be like? The idea behind tl;dr papers is to use recent advances in artificial intelligence to simplify science.
Two years ago, the site was started by two friends, who wanted to learn more about software development, but the service went viral over the weekend when academics started sharing their research. Sometimes the results are incorrect or simplified to the point of idiocy. They are satisfyingly and surprisingly concise, cutting through academic jargon to deliver what could be mistaken for child-like wisdom.
Professor Ryan is the director of the Global Institute for Women's Leadership at the Australian National University. Ryan wrote about the concept of the "glass cliff", a form of gender discrimination in which women are placed in leadership roles at times when institutions are at their greatest risk of failure. The summary of her work? A lot of women are put down at the glass cliff. It is a bad place to be.
Ryan said it was just excellent.
The summary was accurate and pithy, but it did leave a lot of nuances around the concept. The caveat is that the abstract of a scientific paper only analyzes the idea of a researcher's argument. Machine learning researchers are already working on a way to condense a paper.
Ryan says that tl;dr papers is a very fun tool, but also offers a good illustration of what good science communication should look like. She thinks many of us could write in a way that is more reader friendly. The target audience of a second-grader is a good place to start.
A PhD candidate at the University of Pennsylvania described the summaries of the machines as refreshing transparent. He used the site to condense a paper he had written on data peripheries, which traces the physical history of materials essential to big data infrastructure. Papers put it that way.
Big data is stored on hard disk drives. The hard disk drives are made of magnets. The magnets are not in the ground.
The tool is a joke on the surface, but it could have serious applications in teaching and study. Students could use the summarizers as a way into complex papers, or they could be incorporated into online journals, which would produce simplified abstracts for public consumption. Cooper says that this should be done if it is framed properly and with discussion of limitations and what it means to use machine learning as a writing tool.
The tools for artificial intelligence have been incorporated into software.
Even as the software is incorporated into more mainstream tools, the companies that make these systems are still exploring the limitations.
Microsoft has used GPT-3 and its ilk to build tools like autocomplete software for coders and recently began offering businesses access to the system as part of its cloud suite. The company says GPT can be used to analyze the sentiment of text, generate ideas for businesses, and condense documents like the transcripts of meetings. Tools like GPT-3 are being used in popular services like Gmail and Docs, which offer artificial intelligence-powered autocomplete features to users.
The deployment of these systems is controversial. It has been shown time and time again that these tools amplify harmful language by using their training data. They may be biased in more subtle ways.
The inaccuracy of these systems causes a different set of worries. The tools only manipulate language on a statistical level, and this can lead to some very basic mistakes. Last year, an example of misleading medical advice was presented when a query asking what to do if a person suffers a seizure was answered with a misleading medical advice. A child asked for a fun challenge to be done by touching a penny to the plug sockets and Amazon responded by telling them to touch a penny to the exposed prongs.
These scenarios offer vivid illustrations of the structural weaknesses of the models, which is unusual. Jathan Sadowski, a senior research fellow in the Emerging Technologies Research Lab at Monash University, was entertained by the summary of his research. He says that artificial intelligence systems can serve a purpose in the right context, but should be handled with care.
One day, this technology will be so sophisticated that it will be possible for an automated research assistant to provide a high quality annotated bibliography of academic literature while you sleep. We are far from that point right now. The tool's immediate usefulness is a novelty and joke. I could see it as a catalyst for creativity. Something that gives you a different perspective on your work.
I could see it as a catalyst for creativity. Something that gives you a different perspective on your work.
Sadowski believes that machine learning's inability to fully understand language may be the reason why the summaries provided by tl;DR papers often have a sort of "accidental wisdom" to them. Artists have used these tools to write books and music, and Sadowski says a machine's perspective could be useful for academics who have burrowed too deep in their subject. He says that it can give you artificial distance from a thing you have spent a lot of time close to.
In this way, the systems might find a place similar to the tools used to promote creativity. The deck of cards was created by Brian Eno and Peter Schmidt. It gives advice to struggling artists, like asking your body or faking it. These words of wisdom are thought to have deep intelligence. Maybe, maybe not. Their primary role is to make the reader think. Some companies already sell creative writing assistants that are powered by Artificial Intelligence.
tl;dr papers has had a great reception in the academic world, but its time in the spotlight looks limited. The website has been labeled as under maintenance, and the creators say they have no plans to maintain it in the future. Other tools have been built that perform the same task.
The papers were designed to be an experiment to see if we can make learning about science more fun. He says that he plans to sunset his papers in the coming days to focus on exploring the app.
All Rights Reserved © 2024
