Industry Insights AI: 28/05/24
This industry insight focuses on updates to Scopus AI and the recent buzz in the news about researchers possibly misusing large language models due to a surge in certain keywords.
New Scopus AI updates for May 2024
Scopus AI keeps evolving thanks to feedback from its users. Since its launch in 2024, it has introduced several new features, including concept map relationships, a new SML reranker, a reflection layer, and the ability to export references to SciVal. The updates to the concept map relationships allow users to gain further information on how the node relates to the topic. The new SML reranker is a new model that significantly enhances the precision of the Scopus search capabilities and intellectually prioritises the best Scopus articles relevant to the queries. The benefit of this is that it allows for higher quality and relevance when receiving an answer. Another helpful addition is the reflection layer. This provides transparency and context regarding the confidence level of the response. This confidence level can range from high certainty to instances where Scopus AI might not be able to answer a question. The export function to SciVal allows references to undergo further analysis. It is great to see Scopus AI continuing to make updates based on user feedback.
If you’re still feeling unsure about using Scopus AI, we’ve got you covered. Scopus AI has launched three on-demand webinars for you to explore. These cover practical ways to leverage Gen AI-informed results, how Scopus AI navigates ethical implications, quality control during the development of Scopus AI, and how Scopus data can inform a generative AI solution to support researchers. Short on time? No worries! Scopus has updated the FAQ section to address the questions submitted during these webinars. You’ll find information on avoiding bias and hallucinations and accessing Scopus AI.
Does the use of “delve” indicate the use of ChatGPT and other large language models?
Numerous news articles have claimed that researchers are misusing large language models (LLMs) due to a surge in certain keywords since their emergence in the public domain. ChatGPT has even acknowledged this controversy with a tweet: “Don’t mind me just doing some delving today.” Now, a preprint of a study is available, examining words and phrases that have been used more frequently since the rise of LLMs. Interestingly, the word ‘delve’ was used frequently in 2024 but also gained traction in 2021. The authors hypothesise that ‘delve’ might have featured prominently in the training data, leading to its overuse in LLMs. They also point out that language naturally evolves over time, with some phrases gaining popularity while others fade away.
At Elion Medical Communications, we keep up-to-date with the latest trends in AI. Check out our other Industry Insights or find out more about Elion at https://elion.nz/
