How does Digital Science use AI? We ask ChatGPT
The year 2023 will forever be remembered as the year that AI finally broke into the mainstream. But in the research ecosystem, what do we mean by AI and how is it already being used? Simon Linacre does what everyone has been doing and asks AI itself for some answers
Anyone would be forgiven for thinking that artificial intelligence (AI) had only just been invented given the huge interest that has suddenly been generated in it in 2023. However, AI actually has a long history, being first conceived of in the 1950s and gradually becoming a reality in the decades after. Nevertheless, a real step forward has been made with Generative AI, which can create seemingly flawless text or images in just seconds. But is it really that new?
Writing these words on a Google Doc, I have already accepted one or two suggestions from the predictive text function, itself a form of AI. When you realize that the words you have chosen have not 100% been your own, you can see how lines have started to blur with the more advanced Generative AI’s capabilities. However, caution must still be used as AI is more than capable of getting things wrong – just read Digital Science’s post on ‘The Lone Banana Problem’. Moreover, a lack of context or key detail can lead to misrepresentation or redundancy in the content being created.
Digital Science and AI
Digital Science has been using AI and related technologies such as Machine Learning (ML) and Natural Language Processing (NLP) for several years now – indeed, Dimensions has had https://www.dimensions.ai as its URL since its launch in 2018 – and it has become a trusted partner for hundreds of organizations that depend on its AI expertise. These are the ways Digital Science and its portfolio companies uses AI-related technologies in 2023:
- Document classification; Researcher Disambiguation; Research Integrity Trust Markers
- Image classification to decide whether an image shows a chemical compound; Entity recognition to decide whether a term describes a chemical compound
- Sentiment Analysis for Mentions; Article Recommendation Engine
- Knowledge Graph and graph-based reasoning
- Core writing assistance offering is AI-based; AI detection
The information above is correct because, well, I work at Digital Science and I know it to be true. It also breaks the information down by constituent parts in terms of each Digital Science company and how it uses the relevant technology. So what happens if we ask a Generative AI model how Digital Science uses AI? Well, we get the text below courtesy of OpenAI’s new ChatGPT iPhone app:
“Digital Science, as a company that provides scientific software and research solutions, utilizes AI in various ways. Here are a few examples:
- – Text and data mining: AI algorithms are employed to analyze large volumes of scientific literature, extracting valuable insights, identifying trends, and accelerating the discovery process.
- – Natural Language Processing (NLP): Digital Science uses NLP techniques to enhance the understanding and processing of scientific texts, enabling advanced search capabilities, automatic categorization, and semantic analysis.
- – Predictive analytics: AI models are trained on scientific data to predict outcomes, such as the success or impact of research projects, funding opportunities, or potential collaborations, helping researchers make informed decisions.
- – Recommendation systems: By employing machine learning algorithms, Digital Science can provide personalized recommendations to researchers, suggesting relevant articles, journals, or research resources based on their interests and previous activities.
- – Data visualization and exploration: AI-powered tools enable the creation of interactive visualizations and data exploration platforms, allowing scientists to gain deeper insights from complex datasets and facilitate data-driven decision-making.
These are just a few examples of how Digital Science incorporates AI into its offerings. The field of AI is rapidly evolving, and new applications are continually emerging to support scientific research and discovery.”
While all of the information presented by the AI is correct, it is broken down by technical functionality, only presenting ‘examples’ of how technology is used and not what it is used on. As such, it is technically very accomplished, but of little interest to most audiences who would miss the context of how the technology is actually being applied.
AI vs Humans
In 2023 while all the hoo-haa around Generative AI was at its peak, Digital Science launched a campaign called ‘Mind the Trust Gap’ which sought to underline the importance of trust in research among those who both carry out research itself and the rest of us who depend on its integrity and quality. The developments in AI technology are often compared to humans like it is some sort of competition or even battle, but it is perhaps more sensible to view it as a relationship entering a new phase, and trust is key to that relationship. As long as we can use AI with transparency and openness like the examples above, we can build a better understanding of the world around us.