By Chelsea Chung, President at the King’s Commercial Awareness Society
ChatGPT is the latest chatbot developed by OpenAI, an artificial intelligence (AI) research company founded by Elon Musk and supported by Microsoft. Since its launch in November, the programme has taken over the internet, stunning users and onlookers with what it can achieve. This article seeks to examine how the AI model works as well as its implications for the technology industry and wider society.
What is ChatGPT?
ChatGPT is the latest text-generating AI of the GPT family, which furthers the capabilities of its powerful ancestors including GPT-3, the writer of an extensive opinion piece for the Guardian in 2020. The programme allows the user to type in a question using natural language, to which an elaborate – sometimes essay-like answer is then given. The application accommodates countless types of questions ranging from encyclopaedia questions such as ‘explaining Newton’s laws of motion’, to creative questions such as ‘write a poem’. A CNET writer even instructed the bot to write an ‘exciting’ poem, which it did, using emotionally charged words such as ‘battlefield’ and ‘thunder’. There are therefore seemingly very few limits to the programme’s knowledge base, prompting some to proclaim that ‘Google is done’ after witnessing the quality of its responses.
How Does ChatGPT Work?
It is not hard to imagine that the model was developed using a vast amount of text databases from the internet. According to Science Focus, this included 570GB of data acquired from books, webtexts, Wikipedia, articles and other written information. 300 billion words in total were fed into the system to provide the versatile answers to its users. Probability forms the working basis of the language model, which means that it has to be able to correctly and accurately guess the next word in a sentence. It therefore went through an interactive testing stage with its human trainers, where they fed questions and put correct answers into the system, hence continuously refining the bot’s understanding of prompts and ultimately making it a know-it-all.
Implications For The Technology Industry
The establishment of ChatGPT has brought the popularity of chat-based models to a whole other level. This could lead to serious disruptions in web search technology, a domain long claimed by Google. According to Krishna Gade, CEO of AI model monitoring startup Fiddler, Google will possibly have to adjust its core search technology to chat-based to maintain its top spot in the market. This is due to the increasing attraction towards acquiring information through a direct question-answer model, which is more time efficient than the traditional search query. This is not to water down the fact that chatbots are not 100% accurate, meaning that ultimately Google is still a favoured fact-check mechanism and so possesses a major advantage over chatbots. However, the possibility of web search technology becoming more chat-based is not to be neglected as the accuracy of chatbot answers become increasingly refined.
Implications For Society
The emergence of sophisticated chatbots like ChatGPT heavily affects the education sector, as it provides an untraceable and seamless gateway for students to cheat during the completion of essays and online exams, which have been increasingly popular post-Covid. The chatbot’s writing ability has even prompted some to proclaim ‘the death of the college essay in realtime’, as it could be an easy shortcut to achieving solid A grade work in seconds. It is therefore possible that universities will have to come up with new modes of assessment to maintain academic integrity. Anna Chechitelli, Chief Product Officer at Turnitin, said that the anti-plagiarism software company is developing a tool to further identify traces of AI-assisted writing. However, it is unclear whether universities will move towards more interactive assessments as it could be less cost-effective and efficient, according to a higher education consultant.