AI – Help or Hallucination
Technology has been advancing at an exciting rate in many industries, especially in artificial intelligence (AI). AI has been around since the mid-1900s; however, significant advancements in AI have been accomplished by improvements to computing power and storage. In addition, the sheer volume of data available today requires algorithms to be improved to ease the search and use of those data. With all the science-fiction (sci-fi) movies depicting supercomputers taking over humanity, it seemed only a matter of time before algorithms and technology reached new heights. Due to these sci-fi depictions, there appears to be a lot of concern about the development in this realm of technology. However, many examples of AI technology being beneficial to the scientific community can be researched and found.
A study, published in ACS Central Science, showed how AI can be used to help in the development of improved water filters. The Chemists utilized inverse design algorithms with machine learning AI to simulate molecules and water traveling through nanopore filters to assess the effectiveness of the filter membranes. Through their research, the Chemists were able to identify patterns and relationships between molecules, functional groups, and filter membrane designs. This could lead to the design of a more effective and efficient prototype filter that bypassed the time-consuming trial-and-error process of the physical scientific method. The really exciting concept behind this study is that it was capable of showing how AI was able to streamline product design through simulation, and potentially provide solutions to issues that may not have been identified.
Another AI application that has been all the buzz since its release in November 2022, is ChatGPT, an AI-powered language model-based chatbot. ChatGPT is a deep learning generative tool that requires a prompt to create a humanlike output, which is essentially a prediction of what is to come next based on probabilities determined from scanning information from a database.
Now, one may wonder, how this may help the scientific community. Imagine all the time a Scientist could save by not having to scan through thousands of reference materials to find information pertinent to their work. A study published by the American Chemical Society resulted in the creation of its “ChatGPT Chemistry Assistant,” which Chemists were able to use in their research of metal-organic frameworks (MOFs). The astonishing part about this study is that the Chemists were able to develop this streamlined means of research with no coding expertise because of how the AI tool functions.
Some of the additional benefits of AI, other than streamlining the work process, are as follows:
- AI provides a helpful means to overcome the hurdle of creative paralysis through prompts that can be used to generate an idea.
- AI can help to improve and refine work by helping to create more concise sentence structures.
- AI can help individuals who are grammatically challenged through its advanced and improved grammar and spelling checks.
The Risks – False Information?!
Unfortunately, AI does have its limitations and drawbacks, and we must be wary and aware of those. Many generative AI applications, such as ChatGPT, are known to create false information and references by stringing similar reference ideas together to create what has become known as an “AI hallucination.” This is why many experts suggest fact-checking using credible sources to confirm that all information is accurate. In addition, if one were to ask an AI-powered chatbot the answer to an equation showing all the work, both the steps and the answer should be confirmed.
Several other examples are as follows:
- AI databases are usually limited and the information available may not be up to date.
- AI chatbots may cause plagiarism – so be extra vigilant about verification and citations.
- AI applications may deter and/or limit the critical thinking of users by providing seemingly correct solutions or outputs.
- AI applications and software may cause unfair advantages as accessibility is still very limited.
A lot of uncertainty still revolves around AI technology and whether a sci-fi human takeover is inevitably in the making; however, it appears some good can come from embracing AI technology. With the right limitations built into the algorithms, some resource checking, acknowledgement of its pitfalls, and working to understand and improve its applications more, AI can be a beneficial tool for the scientific community and the advancement of research and development. However, at the very least, using AI applications only as a means to an end product, rather than to produce a final product, would probably be wise.