In conversation with the new AI language model, ChatGPT

With the launch of its latest product – the ChatGPT chatbot – OpenAI has been making headlines across the world. 

The company has been taking the internet by storm with the launch of what is one of the most intelligent, sophisticated AI chatbots that has ever been created, leading to a burgeoning slew of creatives attempting to devise an app that can identify ChatCPT’s work. 

In fact, OpenAI (the research lab that created ChatGPT) has since been valued at around a staggering US$29bn. This makes OpenAI, a company that originally started out as a nonprofit, one of the most valuable startups in the entirety of America. 

A monumental success story, it can be largely attributed to the interest garnered by ChatGPT – the responsive capabilities of which are seemingly endless. Across the world, users are asking it about everything from coding advice to generating formal email responses. 

And yes, it’s certainly amusing to ask it to ‘describe the value of AI technology in the style of The Beatles’, and get back the response of: 

"AI technology, it's a revolution

It's changing the way we live and the way we solve problems, too

It's faster than a rocket and smarter than we've ever known

It's helping us in industries from healthcare to finance, and it's just getting started”

But, beyond the entertainment value that has fuelled its viral status, this technology represents a major breakthrough in AI models. If its early success is any indication of what’s to come, this tech could kickstart a transformation across a wealth of industries, spanning everything from global corporations to individual users. 

So, we decided to put ChatGPT to the test. To explore the capabilities of ChatGPT and precisely how ‘human-like’ its organic responses are, we sat down for a pretty singular interview with the AI software. 

ChatGPT’s training & development

It’s not very often that you can ask the software itself about how and why it was developed. 

“ChatGPT is a variant of GPT-3 (Generative Pre-trained Transformer 3), which is a natural language processing (NLP) model developed by OpenAI,” it explains. 

“GPT-3 is a neural network-based model that was trained on a large dataset of text to learn how to generate human-like text.”

In an ongoing phase of ingenious AI development, how exactly is ChatGPT different from other AI language models? 

“One key difference is that ChatGPT is specifically designed for use in chatbots and other conversational applications. This means that it has been trained on a dataset of conversational text, such as chat logs or transcriptions of conversations, and is able to generate text that is more suited for use in a chatbot or other conversational system.”

“Another difference is the size and capacity of the model. ChatGPT is based on the GPT-3 model, which is one of the largest and most powerful language models currently available. This allows ChatGPT to generate high-quality text that’s more coherent and human-like than many other language models.”

To achieve this level of diversity and the high quality of its responses, extensive training was required, utilising a vast, varied dataset. 

“I am a large language model trained by OpenAI. I was trained using a technique called transformers, which allows me to process and understand natural language input. I was trained on a dataset of billions of words from the internet, which means that I have a broad range of knowledge on a wide variety of topics.” 

ChatGPT does not have access to the internet, meaning its knowledge is limited to the point at which its training ended (in 2021). But, despite its current limitations, the ChatGPT tool has the potential to be used in a huge variety of different industry applications across almost all of the world’s sectors. 

According to ChatGPT itself, NLP is needed in areas spanning everything from chatbots, language translation, and text summarisation to text generation and language modelling.

ChatGPT on addressing the key issues of AI technologies

When assessing the potential for the global deployment of technologies like ChatGPT, it is critical to consider how the solution plans to navigate existing issues that we know are present in many AI technologies.

For instance, many of ChatGPT’s predecessors have fallen short in the realm of data bias, with their answers proving to be flawed – particularly in relation to racial and gender parity – as a result of the data that they were trained on.

Interestingly, ChatGPT itself actually had suggestions to offer to help ensure AI models successfully avoid bias: 

“To avoid bias, it is important to ensure that the data used to train AI models is diverse and representative of the population it is intended to serve. This may require actively seeking out and including underrepresented groups in the data collection process. It is also important to carefully examine the data for any potential biases and to correct for them if necessary.”

“In addition, it may be useful to use fairness algorithms or other techniques to try to mitigate bias in AI models.”

“Finally, it is important to regularly evaluate and test AI models for bias, making adjustments as needed to ensure that the models are making fair and unbiased decisions.”

AI’s ability to advance telecommunications

AI technologies are set to make serious waves in the telecommunications sphere – and that’s a point both industry experts and ChatGPT firmly agree on. 

“AI has the potential to be a valuable tool for telcos, and there are several ways in which it could be implemented.”

According to ChatGPT, the potential applications of AI in telecommunications include network optimisation, with AI being used to analyse data from telecom networks to identify bottlenecks and other problems, thereby helping companies optimise their networks for better performance.

Alongside this, ChatGPT recommends that telcos use AI to improve their customer service with AI-powered chatbots, improve fraud detection strategies, improve network security, and perform predictive maintenance – with AI analysing data from telecom networks to predict when equipment is likely to fail, enabling companies to schedule predictive maintenance to a high degree of success. 

“Overall, the use of AI in telecommunications has the potential to improve efficiency, reduce costs, and enhance the customer experience, making it a valuable tool for companies in this industry,” ChatGPT explains. 

Assessing AI’s value through a microlens, we asked ChatGPT how AI is being used to improve mobile phones. Its response described how, among other points, AI is proving instrumental in advancing predictive text capabilities. 

Firstly, “predictive text on mobile phones often uses machine learning algorithms to analyse a user's typing patterns and predict what word or phrase the user is likely to use next. As the phone gathers more data on the user's typing habits, the predictive text function can become more accurate over time”.

Alongside this, there’s the value provided by NLP.

 “Predictive text functions on mobile phones often use natural language processing (NLP) to understand the context of a conversation and make more accurate predictions. For example, if a user is typing a message about a specific event, the predictive text function might suggest words and phrases related to that event.

“Overall, AI is being used to advance predictive text capabilities on mobile phones by enabling them to analyse and understand language and typing patterns more effectively, resulting in more accurate and helpful suggestions,” says ChatGPT. 

Through just this one example, we can see how much AI language tools stand to offer the world’s industries. And, let’s not forget – if there were any doubt as to the power and sophistication of these tools, these benefits can now be described by AI language models themselves.