Understanding ChatGPT

In this article we provide insights on the trending AI tool of the moment, ChatGPT. Learn more about how the AI-powered chatbot came about, how it’s being used, and key limitations which should considered.

How did AI come about?

The fundamental vision of artificial intelligence (AI) was born in 1950 when Alan Turing published a paper titled “Computing Machinery and Intelligence”, asking the question ‘Can machines think? Today, AI is an interdisciplinary science that uses machine and deep learning and involves the theory and development of computer systems able to perform tasks emulating human intelligence. Examples of processes and technologies that use AI include recommendation algorithms, financial forecasting, production processes, autonomous vehicles, language processing, smart speakers and chatbots.

The proliferation of AI in recent years has attracted both excitement and serious concern from many directions and is becoming an increasingly contentious issue. AI raises many ethical issues such as the replacement of human labour, privacy, deep fakes, and the potential dangers of technology making potentially ‘life or death’ decisions (i.e., self-driving vehicles).

The legal industry is also faced with the question of whether the law and legal practice are able to keep up with rapid advancements in AI technology and the way in which we are using it. ChatGPT, released in November 2022, is the most recent AI platform that has taken the world by storm.

What is ChatGPT?

Chat Generative Pre-Trained Transformer (ChatGPT) is an AI-powered chatbot developed by US-based company OpenAI using the Generative Pretrained Transformer language model. In its own description ChatGPT uses deep learning techniques to generate human-like responses to text inputs in a conversational manner. Generative AI tools allow users to enter written prompts and receive new human-like text, images, or videos. ChatGPT has made waves as it’s powered by a large language model that can understand human language and generate responses based on a large pool of data. The bot was trained using a large text sample taken from the internet and has built it knowledge by sorting through 300 billion words and 570 gigabytes of data.

The model currently available to the public for free is called GPT-3.5, this is one of the largest and most powerful language processing AI models. However, there is also now GPT-4 which is being made available at a cost, and currently has a wait list to try.

What is ChatGPT-4?

ChatGPT-4, released in March 2023, is a new model of the AI system, available to ChatGPT Plus customers (the paid for version).

Differences between ChatGPT-3.5 and ChatGPT-4

Advancements in the technology are impressive to say the least. When we consider the improvement from GPT-3.5 to GPT-4 we can’t help but wonder, what’s next? The potential of this AI technology is exciting and a little frightening.

How is ChatGPT being used today?

ChatGPT can be used for a wide range of tasks including; text generation, text completion, text summarisation, text classification, dialogue generation and language translation. Previously most chatbots were private or otherwise available for a fee, whilst ChatGPT is currently completely open to the public. Naturally, the open accessibility lends itself to a whole new range of uses, such as university students using the platform to write essays.

What are the limitations to the information?

ChatGPT gets its information from public internet sources and can therefore be unreliable or incorrect. Other limitations include the fact that the bot’s knowledge base only extends up until September 2021, that it can’t accurately replicate human speech and writing, and that it can get confused by prompts that are too complicated.

When asked about its limitations, ChatGPT-3.5 responded with the following:

• Lack of common sense: ChatGPT is an AI language model that relies solely on the data it has been trained on, and it does not have a real-world understanding of concepts and situations that may seem obvious to humans. As a result, it may provide answers that are technically correct but not applicable or relevant in real-world scenarios.
• Biased responses: Like any other AI system, ChatGPT is only as good as the data it has been trained on. If the data used to train ChatGPT is biased or incomplete, it may provide biased or incomplete responses.
• Inability to recognise emotions: ChatGPT is not capable of recognizing the emotional state of a user or understanding the context of a conversation. This means that it may provide inappropriate or insensitive responses to emotional or sensitive topics.
• Limited scope of knowledge: ChatGPT’s knowledge is limited to the data it has been trained on. As a result, it may not have answers to questions that fall outside of its scope of knowledge.
• Language limitations: While ChatGPT is proficient in a variety of languages, its understanding of languages other than English may be limited, and it may not be able to provide accurate responses in those languages.

It’s important to keep in mind that ChatGPT is a machine learning model and not a human expert, so its responses should be taken as general information only, and not as professional advice.

How new is the technology?

Chatbots themselves have existed for decades. ‘ELIZA’ was one of the first chatbots, developed in 1966 by Joseph Weizenbaum. ELIZA mimicked human interaction through pattern recognition and built-in scripts but could not respond to questions in their complete context. Chatbot technology has been developing ever since.

ChatGTP is set apart by its level of sophistication. Previous chatbots were able to have conversations and answer questions, but ChatGPT has a unique ability to learn and self-correct. The technology is distinctive in its ability to learn while predicting what the next word should be, improving its understanding of a prompt as it goes.

What are some areas of concern?

• Accuracy and the spread of misinformation
• Student use and academic misconduct in universities
• Copyright – infringement of existing materials and difficulties in claiming rights over AI generated material.

Looking ahead…

Chatbots and other new AI technology will increasingly impact our lives and workplaces, requiring constant adaptation. In the face of wide-spread use of chatbot technology, fact-checking and combating the spread of misinformation will become more important than ever. Privacy will also become increasing relevant, with businesses and professional industries needing to implement rules monitoring the submission of information to chatbots, and copyright protection of works submitted to or created by chatbot technology will become increasingly complex.

Follow the FAL Lawyers’ AI series to learn more about developments in AI, limitations, legal considerations, and more. Through this series we aim to drive discussion around the future of AI and gather insights on what these developments will mean for our work.

The contents of this article do not constitute legal advice and should not be relied upon as such. If this article pertains to any matters you or your organisation may have, it is essential that you seek legal and relevant professional advice.