ChatGPT: An ingenious predicament

A chatbot is a program by the computer that simulates human-to-human conversation through the Internet. As an artificial intelligence (AI) language model, identity is given to a chatbot as a Chat Generative Pre-trained Transformer (ChatGPT) owned and developed by an AI research and deployment company, OpenAI. Headquartered in San Francisco and was designed by some prominent players – including Elon Musk, Sam Altman, Peter Thiel, OpenAI chief scientist Ilya Sutskever, Jessica Livingston, and LinkedIn cofounder Reid Hoffman. ChatGPT is a large language model (LLM). It is basically a man-made neural network that is pretrained using self- and semisupervised learning. ChatGPT is OpenAI’s LLM. Other such sister LLM is Google’s PaLM, which is used in Bard, Meta’s LLaMa, and Anthropic’s Claude 2. ChatGPT chatbot has raised the bar for AI and demonstrated that machines can actually “learn” the nuances of human communication and engagement. On November 30, 2022, OpenAI made a preliminary demo of ChatGPT available online. As people posted samples of what the chatbot was capable of, the chatbot soon gained popularity online. Everything from composing fables to coding computer programs to organizing trips was covered in the stories and examples. The chatbot gained over a million users in just 5 days. Generative Pre-trained Transformer’s Major Milestones ChatGPT is the brainchild of British-Canadian computer scientist, Geoffrey Hinton, who is known as the “Godfather of AI.” ChatGPT’s evolution has been distinguished by incremental improvements, with each version building on prior tools. The first iteration of the GPT, i.e., GPT-1, was introduced in June 2018. The release of GPT-2 in February 2019 indicated a significant increase in text production capabilities, producing coherent, multi-paragraph text. However, due to its potential misuse, GPT-2 was not initially released to the public. GPT-3 was a huge leap forward in June 2020. The trend of exponential improvement was continued and has also been maintained by the latest iteration GPT-4. ChatGPT-4, the most recent natural language processing technology in use, has evolved with GPT-4 vision, which allows the AI model to capture images, analyze, and revert to text-based queries related to the image. This enables users to extract details from visual content to transcribe text and solve visual calculative problems. ChatGPT can quickly produce a business plan, fix coding issues, and compose essays for students, blogs, and even Shakespearean sonnets. The chatbot can be used in the field of medicine for tasks such as disease surveillance, clinical decision support, maintaining electronic medical records, and evaluating and interpreting medical literature. The World Economic Forum predicts that roles such as bookkeeping, insurance underwriting, and credit analysis are jobs that will be jeopardized the most by automation. The novelty to the chatbot is due to the use of Reinforcement Learning from Human Feedback which makes ChatGPT a unique one.[1] This AI chatbot has been vested with amazing potential; it has fanned the same anxieties about the future of jobs as automation has, reinforcing an assumption that robots will outnumber humans in the future. Another important application of ChatGPT in medicine is the development of virtual assistants to aid patients in managing their health. However, the use of ChatGPT and other AI techniques in medical diagnosis and patient counseling has raised ethical and legal concerns despite its potential advantages. A matter of concern with the use of Chat GPT is the fabrication of data and references. A study by Bhattacharyya et al. revealed that out of 115 references generated by ChatGPT, 47% were fabricated, 46% were authentic but inaccurate, and only 7% were authentic and accurate emphasizing the need for execution of caution while seeking any medical information on ChatGPT.[2] Can Chat Generative Pre-trained Transformer Replace Human Brains? There have been rising concerns regarding AI chatbots eclipsing or degrading human intelligence. For example, the chatbot may possibly replace the requirement for a human writer by writing an article on any subject accurately and effectively, making it easier for students. This in due course leads to a decrease in ability of quality writing, cheating, loss of interest for creative penning, and avoiding to adapt skills to learn required for proper article writing. This has led to some school districts blocking access to it.[1] Are Chatbots Surrogates for Diagnosis by Doctors? AI lacks domain-specific training, but still, the algorithms designed by AI show accuracy in knowledge-based tests indicating that AI has the potential to revolutionize health care and make it more efficient in terms of diagnostics, detecting medical errors, and decreasing paperwork burdens. However, it is unlikely that AI will ever be a substitute or replace the physicians. In the United States Medical Licensing Examination, ChatGPT achieved 66% and 72% on Basic Life Support and Advanced Cardiovascular Life Support tests, respectively, and performed at or near the passing threshold. However, they are notoriously weak at context and nuance, both of which are critical for safe and effective patient care, which requires the application of medical information, concepts, and principles in real-world circumstances that would never be justified by any machine. According to Frey and Osborne’s examination of the future of employment, while the probability of administrative health-care positions being automated is rather high (e.g., 91% for health information technicians), the possibility of physicians and surgeons’ professions being automated is 0.42%. While some evidence suggests that fully autonomous robotic systems are “just around the corner,” the duty of a surgeon extends much beyond executing surgical procedures. The physician’s task is complicated by the ability to provide fully integrated care by delivering both treatment and compassion. We were taught as medical students to always take care of patients rather than their medical records – a clinical ability that computer algorithms are yet unable to comprehend. As a result, the huge promise of AI in health care lies not in the ability to replace physicians but rather in the ability to boost physicians’ efficacy through workload rebalancing and performance optimization.[3] With ChatGPT boasting a plethora of applications, it is widely being used in writing scientific literature. The literary works produced by using ChatGPT were found to be eloquent, pleasant to read, and could be also used as a search engine too. ChatGPT generates articles which escapes plagiarism checks. It has strengthened students and scientists to deceive others or pass off the AI-generated texts as their own. ChatGPT can also be used for monitoring and follow-up of patients and screening activities too in case of exacerbation of risk factors. However, with every good comes something bad like ethical concerns and limitations which cannot be overlooked. Medicolegal complications, infringement of copyright laws, biased inaccurate results, irrelevant references, and scientific misconduct are some of the issues with ChatGPT which need to be addressed.[4] Another major challenge is AI hallucination wherein the chatbot perceives patterns or objects that are imperceptible to human observers creating outputs that are inaccurate altogether. Hence, total dependence on AI-enabled chatbots may prove a double-edged sword in the long run.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.