Every year, a whole lot of industry, academia, government and numerous other organisations contribute to R&D in AI via their research papers, articles, journals, etc., on topics of machine learning, computer vision, natural language processing, and more. The number of AI publications in the world has increased from 162,444 in 2010 to 334,497 in 2021 - nearly doubled as per the AI Index Report 2022. 

As a well-known fact, research publication has a complex and long lifecycle. Now, the thing to ponder upon is - Can AI come to the rescue for researchers, making the entire process simpler, and helping them focus on core work that can ultimately lead to an uptick in their productivity and throughput. 

Meet Nishchay Shah, Chief Technology Officer at CACTUS. We got in a conversation with Nishchay to understand how CACTUS Labs, where he is the Head, works on AI solutions that automate and augment various processes of the research cycle, how AI is impacting the content domain, and much more.

Integrating AI into the researchers' ecosystem

"Fortunately, research is one domain where we are really good at since we've been in research communication for 20 years now. Research is a long cycle, and we are here to assist researchers in making this journey hassle-free with our multiple AI capabilities at CACTUS labs," said Nishchay.

CACTUS Labs is the Machine Learning unit of CACTUS, and its expertise lies in developing wide-ranging AI tools for researchers, which include:

• Literature Surveillance (to find what to write) R Discovery: Labs works closely with R Discovery, a CACTUS product that assists in generating research paper recommendations based on a proprietary concept extraction algorithm. Labs' big data and semantic intelligence team help in curating millions of data points for recommender logic and creates (and constantly evolves) the concept extraction algorithm. 
• Writing enhancement software (grammar checks, suggestions, etc.) - Paperpal Edit: Real-time grammar checks, sentence formation suggestions, and potential reference checks from a corpus of millions of research artefacts help one with focused writing. CACTUS helps with real-time NLP pre-processing, grammar error detection + correction, and reference search for one of CACTUS' products – Paperpal Edit. 
• Academic technical checks (format, guidelines, etc.) - Paperpal Pre-Flight: Once a manuscript is written, many things need to be checked manually or by a reviewer. Language quality, layouts and formats, relevant references, ethical use of images, table citations, and journal guidelines are but a non-exhaustive list of items to be checked. CACTUS Labs builds predictive models to automatically and instantly check these aspects with zero human intervention. 
• Which journals to publish in? (predictability of whether the paper will get published): Their Global Journal Database product works primarily based on historical data and data from the manuscript content and helps us figure out the possibility of the manuscript being accepted in a journal. "With supervised learning and custom feature engineering on top of a decade worth of data from our editing service, we can generate this insight for the author/researcher," said Nishchay.
• Create visualisations to share research work with the masses: CACTUS recently acquired Mindthegraph, a DIY product that can help researchers create seamless infographics and posters without any graphic designing or coding skills. The product works on a patented content element generation using AI + tagging method with a visual interface for the researcher to quickly create posters, abstracts, and infographics. 

However, many ML machine models used in real-world applications are often referred to as "black boxes." While the models and algorithms are complex to comprehend in the initial iterations, explainability helps in providing a clear picture of certain patterns and methods that the algorithm is focused on to derive a result.  

So, what's your take

Mentioning his commitment, Nishchay says, "We at Labs are building hybrid models, where we can keep an element of explainability in place. While it is sometimes not possible for a few of the cutting-edge models, we have succeeded in creating sanity checks and cases which help us chart a course towards understanding how our models are performing over time and identifying certain traits."

• The team at CACTUS Labs rigorously looks at model explainability and spends quite a lot of time understanding the data- given that it comes from 1000s of subject areas. 
• The company also have a lot of customised tests to understand more about the impact the data has on the models, which helps us create a nice feedback loop to gain overall, system-wide explainability. 

The team presented one example from one of their products powered by the CACTUS Labs' NLP solution– "here, we have gone one step beyond and have added an explanation as to why using another word might help improve the readability of the sentence," explained Nishchay. 

Is AI going to replace humans?

"It's mostly bad PR which AI has gathered over decades of its existence – It will replace humans, take away jobs, make us redundant"

Say, for example, in the publishing industry, it's nearly impossible to replace an editor simply because the language has so many nuances, and people make mistakes that are not just random but very erratic. This leaves massive room for errors that cannot be picked up by ML and AI, leading to very unpleasant results and a bad user experience.  

However, smart enterprises and businesses know that if they don't use AI to scale and augment their solutions and products, eventually, they will get disrupted by cheaper, faster, and better competitors in the market. So, the collaboration between humans and AI is the key. To be precise, AI still can't do very basic tasks at which humans excel; it can help with repetitive and shallow tasks, helping us focus more on expanding businesses and services and creating new products. 

For Nishchay, it is definitely going to be an AI-augmented workforce in the future. Now, AI is out of labs and has real-world applications. It is here to stay– if not as a total replacement, then at least as a trusted workforce ally. "At CACTUS Labs, we have constantly been pushing for assistive AI in multiple services and business processes, and the results have helped us improve our efficiencies," he said.  

Future plans

While talking about his plans for the future, Nishchay said: "The Global AI race has begun, and I am not done yet. We are committed to continuing to build solutions and products that help accelerate global research." Further, he intends to take things a notch higher and work to,

•  Provide AI as a Service business model to not just businesses in the CACTUS ecosystem but beyond. 
• More products in the Computer Vision or Extended Reality space
• Build and sustain global research teams to fuel more research into AI for this niche and double our team size. 
• Open source some of the team's work to increase collaboration with AI communities.