Will Large Language Models Like ChatGPT Change the Way We Interact With AI?

Even if you're not interested in artificial intelligence, by now, you've heard the buzz about ChatGPT. Everyone is talking about this new chat tool that allows you to interact with an artificial intelligence system as if you're talking to a human. It will answer your questions and write text if you ask it to. And it's way better than everything you've seen so far.

Now you're probably wondering what's so special about it.

After all, we've had "intelligent" chat systems for a while. Any business website you visit sports a chatbot offering to answer your questions and is usually unable to do so beyond the simplest request.

Then there are sites like Jasper.ai that promise to write content for your blogs or social media posts. I experimented with such a site for a couple of weeks and soon realized that the output was often incoherent, repetitive and, worst of all, highly inaccurate. It made up facts and quoted non-existent research sources. In short, it was pretty useless beyond writing 4-5 sentences as an intro to a blog post and required heavy editing.

So when I decided to give ChatGPT a try, my expectations were low, but frankly, I was blown away.

Talking to this AI assistant feels like talking to a human being. No matter my question, it answered coherently, and the most exciting thing: it understood my follow-up questions. I was able to have a real conversation with it. A skill my Alexa hasn't mastered after years and years of development.

What is the ChatGPT?

According to the website, they've: "trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests."

So what does this mean?

The OpenAI Assistant told me it is a type of artificial intelligence designed to generate human-like text based on the input it receives. It uses machine learning algorithms to process the input and generate coherent and appropriate responses in natural language.

The training data consists of a vast amount of text from a variety of sources, including books, articles, and online forums. This allows the Assistant to understand the context and structure of the language used in different situations and to generate responses that are appropriate for the given input.

To generate text, the Assistant uses a combination of techniques, including a recurrent neural network (RNN) and a transformer model. The RNN processes the input the assistant receives and uses to generate a sequence of words or tokens. The transformer model then takes the output from the RNN and uses it to generate the final text. This combination of techniques allows the Assistant to generate coherent and appropriate responses in real time.

In addition to the assistant's ability to generate text, it also has a built-in attention mechanism that allows it to focus on specific parts of the input and use that information to generate more accurate and relevant responses. This allows the Assistant to have more natural and engaging conversations with users.

Overall, the Assistant's primary function is to assist users by generating human-like text based on the input it receives. It uses advanced machine learning algorithms and techniques to process the input and generate appropriate responses in real-time, allowing for natural and engaging user interactions. You can use any language you like to communicate with it, and it will try to generate appropriate and coherent responses in the same language as the input it receives.

How will ChatGPT be useful?

There are many potential use cases for a large language model like the ChatGPT Assistant. For example, it could be used to assist with customer service inquiries, providing quick and accurate answers to a wide range of questions. This could help to improve the customer experience and reduce the workload for your customer service team.

Content creation: The output the assistant delivers is coherent and sounds natural. Unlike other tools, it could actually be used to generate content for websites or social media channels. This could help save time and resources while still providing engaging and informative articles or posts.

In addition, it could be used in education, providing students with access to a wealth of information and the ability to ask questions and receive accurate and helpful answers.

It could also be used to provide students with information about a particular subject or to assist teachers with answering common questions from students. It could assist with creating educational materials, such as lesson plans or study guides. Or help with the organization and analysis of educational data, such as by summarizing the findings of research studies or identifying trends in student performance.

The Assistant could also be used in healthcare, assisting doctors and other medical professionals with diagnosing and treating patients. For example, it could be used to assist with the creation of medical documents, such as patient records or treatment plans. Or it could be used to help with the organization and analysis of medical data, such as by summarizing the findings of clinical studies or identifying trends in patient records.

Additionally, its ability to engage in natural language conversations could potentially be useful in healthcare settings. For example, the assistant could be used to provide patients with information about their health or treatment options, or to assist with answering common patient questions.

Is this a brave new world?

Overall, the possibilities for a large language model like this are vast, and there are many use cases we probably haven't even thought of yet.

But of course, there are several risks associated with using a system like ChatGPT.

It will become increasingly difficult to know whether you're talking to a human or a machine. Teachers will have issues knowing if an essay submitted was written by the student or by AI. And even the best-trained model can sometimes get it wrong and deliver inaccurate information. There's even a disclaimer about this on the website: "ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers." How will we know when this happens?

As language processing models evolve and improve, we as a society are tasked to discuss how we're willing to interact with and use such tools.

Do we need disclaimers to let people know when they're talking to a machine or if a text is created by a model? Will we limit the use of AI in the educational sector to ensure students write their own text and learn to source information? Or will we find ways to use AI to enhance education?

These are some of the discussions we have to have. Right Now!

And we should all be involved in this discussion to ensure that we don't wake up in a world where models are doing all the thinking and we're simply consuming the output of these models.

Disclaimer: Part of this article was written by the Assistant. I'd love for you to try and find out which part that is!