|Neural networks are now capable of creating a technology that can read scientific papers / Photo Credit: dee pakpal via Flickr|
Earlier this year, Stanford University released its report on the impact of artificial intelligence on people’s lives. The study titled “One-Hundred-Year Study on Artificial Intelligence (AI100),” focused on tracking activity and progress on AI initiatives. At the same time, the study aimed to facilitate informed conversations backed with reliable, verifiable data.
According to Forbes, a global media company focusing on business, investing, technology, entrepreneurship, leadership, and lifestyle, the study findings showed that global revenues from AI for enterprise applications are predicted to increase to $32.2 billion in 2025 from only $1.62 billion in 2018. This will attain a 52.59 percent CAGR in the forecast period.
Among the enterprise application use cases that are projected to fuel its rapid growth include the use of algorithms and machine learning, predictive maintenance, localization and mapping, patient data processing, and image recognition and tagging.
The report also revealed that 84 percent of enterprises believed that investing in AI will lead to greater competitive advantages, 75 percent were convinced that it will open up new businesses, and 63 percent believed that AI can reduce costs in companies. It’s therefore not surprising that AI has been a topic of interest in many studies.
For instance, the entire database of Scopus, Dutch information and analytics company Elsevier’s abstract and citation database, contains over 200,000 in the field of computer science that have been indexed with the key term “Artificial Intelligence.”
AI That Reads Academic Papers and Highlights Key Findings
AI not only contributes to various industries but also academic papers and studies. A study recently published in the journal Transactions of the Association for Computational Linguistics offered help in summarizing the findings of a complex and technical research paper into plain English. This could potentially help journalists in communicating or translating complex studies to the public.
|AI not only contributes to various industries but also academic papers and studies / Photo Credit: Public Domain Pictures|
According to Inside Higher Ed, a media company and online publication that provides news, opinion, resources, events, and jobs focused on college and university topics, scientists at MIT and the Qatar Computing Research Institute at Hamad Bin Khalifa University used neural network—a form of AI—to create a technology that can read scientific papers. At the same time, the tool can generate easy-to-read summaries that are just one or two sentences long.
Niki Kittur, professor at the Human-Computer Interaction Institute at Carnegie Mellon University, stated that this study can help researchers go through and understand individual papers faster. This is particularly important as scientists have to sift through millions of scientific papers published every year, which is fundamental to accelerating scientific progress.
AI in Finding New Scientific Discoveries
With hundreds to thousands of scientific papers published every year, it's easy for scientists and researchers to overlook major and minor details. But the scientists from the US Department of Energy’s Lawrence Berkeley National Laboratory have a solution for this. Recently, they used AI and machine learning to reveal new knowledge buried in old research papers. This showed that both technologies prove to be handy and can be used by all sorts of industry experts.
The researchers combed through research papers for any information scientists may have overlooked in the past using the algorithm Word2Vec. According to WouldSaySo, an online site that features the latest news and stories in science, tech, health, physics, and people, they provided the system with abstract data before letting it loose on archives of scientific data.
The researchers trained Word2Vec to assess over three million abstracts related to material science. The algorithm then gathered and compiled a vocabulary of around a half-million words to preserve and represent their syntactic and semantic relationships.
The team gathered 3.3 million abstracts from scientific papers published in over 1,000 journals between 1922 and 2018. The algorithm was able to learn scientific concepts and infer relationships between data points despite having no prior knowledge about the papers.
In a statement, Anubhav Jain, a team leader and paper co-author, said, “Without telling it anything about materials science, it learned concepts like the periodic table and the crystal structure of metals. That hinted at the potential of the technique. But probably the most interesting thing we figured out is you can use this algorithm to address gaps in materials research, things that people should study but haven’t studied so far.”
According to Vice, a Canadian digital media and broadcasting company, one of the most impressive findings done by the algorithm was predicting thermoelectric materials. Even without training in materials science, Word2Vec was able to provide candidates future thermoelectric materials. In fact, some of these materials may be better than those we currently use.
At the same time, the algorithm was able to understand concepts like the chemical structure of molecules and the periodic table. This gap in knowledge has been easy for the algorithm to spot, which can be difficult to catch with a human eye.
Also, this study published in the recent edition of Nature revealed the current problem with today’s scientific literature: most of the papers were text-based—whether it’s by conventional statistical analysis or via existing machine learning methods—that made it hard to analyze.
As these results have shown, AI could be used to find new scientific discoveries altogether. The algorithm not only proves that it could do wonders in today’s scientific literature but also opens new opportunities to create tools that can be of great help for things like medical research or drug discovery.