What Exactly is ChatGPT and How Do I Use It?
March 31, 2023
Artificial intelligence (AI) does not get much better than ChatGPT. But what makes this AI chatbot technology so advanced?
ChatGPT was developed by OpenAI, which was founded on 10 December 2015 by Elon Musk, Amazon Web Services, YC research, and a few others. OpenAI is an AI research laboratory based in the United States, with non-profit and for-profit arms and offices in San Francisco. Their aim is to develop what they call ‘friendly AI’ that has a positive impact on humanity.
Launched in November 2022, ChatGPT has over 100 million monthly users already. It is a sibling model to its predecessor InstructGPT, which was launched in January 2022.
Let’s have a detailed look at the ChatGPT technology, its incredible user advantages, and how it compares with Google and its other main competitors. We’ll then look at how ChatGPT will change how we search for information and create new content.
Lastly, as with any new AI technology, ChatGPT still has many limitations and, by the company’s own admission, may occasionally generate incorrect, harmful, or biased content. We will explore these limitations in detail and look at how the company plans to resolve them in the future.
ChatGPT Technology
ChatGPT or “Generative Pre-training Transformer” is a ground-breaking new artificial intelligence technology that allows chatbots to understand complex commands and generate human-mimicking natural language responses with unparalleled accuracy and fluency.
To illustrate this, we have instructed the ChatGPT chatbot to do the following:
“Explain ChatGPT technology in simple terms.”
The following response ChatGPT generated is mind-blowing.
“ChatGPT is an advanced computer program that uses artificial intelligence to generate responses to questions and statements entered by users.
It works by analyzing vast amounts of text data to learn patterns and relationships between words and phrases, which it can then use to generate coherent and contextually appropriate responses.
The system uses a type of AI called “deep learning,” which involves training a complex neural network on massive amounts of data.
This allows ChatGPT to recognize and understand complex language patterns and respond to queries in a way that is both natural and helpful.
Overall, ChatGPT is a powerful tool for language processing and communication, enabling users to engage in sophisticated conversations with a computer in a way that feels surprisingly human-like.”
ChatGPT functions on the basis of pre-training a complex neural network on a vast dataset of existing text and then fine-tuning the network on specific tasks, such as, for example, answering a complicated question.
The neural network is made up of a series of interconnected layers, also known as transformer blocks. These transformer blocks process the input text from the user and then generate a prediction for the output or response.
A key differentiating function built into the ChatGPT technology is the AI’s ability to understand the conversational context and generate contextual responses.
This is done through Self-Attention Mechanisms, which allow the network to prioritize the relevance and importance of different words and phrases in the input text.
This has been a major advancement made by OpenAI in the field of Natural Language Processing (NLP).
What exactly can ChatGPT do?
As the screenshot below illustrates, we have again asked ChatGPT to explain itself, and its own response is remarkable.
1 – Answer Questions
What makes ChatGPT especially great is the fact that you can ask the AI almost any question imaginable. The chatbot then generates fairly decent responses in just a few seconds.
These responses negate the tedious task of searching for and reading multiple sources on a traditional search engine to get the same insights.
The downside is that you can’t truly assess the quality of the answers ChatGPT provides because it does not list the sources it used to compile the answers. Nevertheless, users can ask ChatGPT the following types of questions:
- General knowledge questions
- Personal questions
- Educational questions
- Fun fact questions
- Technological, political, and societal questions
- Future prediction questions
- Social questions
- Creative questions
2 – Write Content (Emails, Essays, Even Poetry, and Music)
When it comes to creating accurate content, few AI technologies are as advanced as ChatGPT. The AI’s advanced Natural Language Processing technology gives it the ability to understand and write content in a very similar manner to humans.
To demonstrate this, we asked the ChatGPT chatbot to write a poem about spring. In the space of 10 seconds, the chatbot produced the following:
“The world awakens from its sleep,
As winter’s icy grip begins to weep.
The air grows warmer, the skies turn blue,
And nature shows its shades anew.
The trees bud forth, the flowers bloom,
And every corner bursts with room.
The birdsong fills the morning light,
And sets the tone for the new delight.
The earth feels fresh, the breeze is light,
And all is perfect in the sight.
The world has come alive once more,
With promises of life to pour.
As springtime’s season ushers in,
We bask in warmth and love begin.
For in this time of freshened air,
We find renewal, hope and care.”
Most other AI writing tools offer the same generic responses to questions and commands, whereas ChatGPT generates a unique response that demonstrates contextual awareness every time.
3 – Summarize Documents
ChatGPT uses what’s called extractive summarization, which involves the analysis of a particular document to identify the most important sentences of paragraphs and then use those to generate shorter summaries of the document.
Although the summary function does work, the accuracy and quality of any summary become more questionable with technical and complex materials. The AI can summarize the following:
- Articles from newspapers, magazines, and journals
- Research papers and academic articles
- Books and e-books
- Reports and white papers
- Legal documents such as contracts, patents, and court rulings
- Business documents such as proposals, plans, and presentations
- Medical documents such as patient reports and clinical studies
- Government documents such as policies, regulations, and official statements
4 – Write and Debug Code
Unfortunately, ChatGPT is not going to write the next Windows or Mac OS operating systems on its own, but this amazing AI technology can assist professional programmers with writing and debugging sections of code.
The screenshot below is the response generated by ChatGPT when we asked the bot to write a function in Python to add two numbers:
A coding assessment conducted by England’s Centre for Finance, Technology, and Entrepreneurship found that ChatGPT performed better than 85% of the approximately four million programmers who took the test on Python. Coders can use ChatGPT for the following:
- Explaining code
- Improving existing code
- Rewriting code using idiomatic constructs
- Rewriting code using the correct style
- Writing test cases
- Simplifying code
- Exploring alternatives
- Translating code
- Writing code documentation
- Tracking bugs
- Code scaffolding
5 – Quickly Explain Complex Concepts in Simple Terms
Undoubtedly, one of the most valuable functions this AI technology offers is its ability to quickly explain very complex concepts and ideas in varying levels of everyday terms.
For instance, you can ask ChatGPT to explain quantum physics in simple terms. To demonstrate, we have done just that:
“Explain quantum physics in simple terms.”
“Quantum physics is the branch of physics that deals with the behavior of very small things, like atoms and particles. It’s a field of science that explains how things work at the smallest scale.
In the world of quantum physics, some things behave in ways that seem strange and different from how we experience the world around us. For example, particles like electrons can exist in more than one place at the same time (this is called “superposition”), and they can also be “entangled” with other particles so that they become linked in a way that we don’t fully understand.
Another strange feature of quantum physics is that observing a particle can actually change its behavior. This is known as the “observer effect.” When we measure or observe a particle, we can change its state or even cause it to behave differently.
Quantum physics also introduces the concept of “quantum mechanics,” which is a set of mathematical rules that help us understand how particles behave. These rules describe the probability of finding a particle in a particular place or state, rather than the precise location or state of the particle.
Overall, quantum physics is a fascinating field that challenges our understanding of the universe and raises many interesting questions about the nature of reality.”
As this example demonstrates, the AI is sophisticated enough to avoid technical jargon and complex vocabulary.
It even incorporates analogies and metaphors into the explanations. It focuses on the main points and breaks up the concept into smaller and simpler parts, which makes it much easier for the novice to comprehend.
Is ChatGPT free to use?
OpenAI launched ChatGPT as a free-to-all tool initially, but, after the AI chatbot went viral, OpenAI quickly introduced a monthly paid subscription called ChatGPT Plus that gives subscribers priority access during busy times.
Paid subscribers also have priority access to new features and faster response times.
A subscription will set you back $20.00 a month; however, you can still access the chatbot free of charge. If you have used ChatGPT for a while, you will be familiar with the very annoying “ChatGPT is at capacity right now” notification.
When this notification appears, non-paying subscribers have no choice but to wait until the bot has enough capacity to process their requests. If you are eager to use the chatbot with no interruptions, the paid subscription may be the best route for you.
What do the critics say?
As with any new technology, ChatGPT is under the microscope of critics who point out some valuable arguments.
One of the main arguments is that these AI language models can sustain and sometimes even amplify biases present in the data that was used to train the model.
Additionally, some argue that these models do not have a true understanding of the training data and simply generate responses based on statistical patterns in the data.
Finally, some concerns have been raised regarding the ethical use of the technology: in particular, it's potential to generate and spread disinformation and misleading content.
ChatGPT versus Google
There are several differences between ChatGPT and Google, the main one being that ChatGPT is a Language Model, whereas Google is a search engine.
Both Google and ChatGPT use artificial intelligence and Natural Language Processing to get their user’s information, but they each focus on different outcomes.
While ChatGPT generates detailed human-like responses to the input text, Google provides bits of information with links to online sources.
Google Bard
ChatGPT took the world by storm with its unprecedented AI and natural language capabilities, which resulted in Google launching a similar AI chatbot called Bard, an experimental AI conversation tool, rather prematurely in February 2023.
Google unveiled their next-generation large language model called LaDMA about two years ago and has since been developing Bard on the back of that technology.
CNBC recently reported that Google’s “code red” plan to respond to the launch of ChatGPT included having staff test the Bard AI model.
Google has since opened up its Bard chatbot to trusted testers ahead of making it more widely available.
In an attempt to get the word about Bard out to the public, Google posted a video on Twitter that demonstrated the AI’s capabilities. Unfortunately for Google’s parent company Alphabet, the demonstration went horribly wrong.
In the video, Bard was prompted with the question “What new discoveries from the James Webb Space Telescope can I tell my 9-year-old about?”
The first two Bard responses were accurate but the third missed the mark by a long shot. This resulted in a 10% drop in Alphabet’s share prices within a week. Their share prices have continued to be under further pressure since Microsoft announced an integration of ChatGPT into its search engine Bing.
At the moment, we don’t know exactly how Google’s Bard will compare with ChatGPT because it is still in the testing phase. One thing is for sure, a company like Google, with almost unlimited financial resources, can and most likely will find a way to get its tech up to speed quickly.
Will ChatGPT replace Google Search?
The short answer is no, ChatGPT won’t replace Google or any other search engine. Why? Because they have very different functions.
The aim of a search engine is to generate high-quality, accurate, and up-to-date information and sources from the internet. ChatGPT was trained on a vast dataset of information, but only up to 2021, and it does not pull information from the internet. Recent events like the launch of Google Bard do not form part of the ChatGPT’s training data, as the following screenshot demonstrates:
With most Google searches relating to breaking news, new trends, and current events, ChatGPT as it currently exists will not be able to replace Google.
This lack of up-to-date information also has implications for content writers who rely on the internet and up-to-date research to write blogs, news articles, books, and other materials.
What does ChatGPT mean for the content industry?
At the moment, ChatGPT and other language tools should only be used as a supplementary automation tool that helps you develop your content. These technologies, at the moment, can’t and will likely never replace human writers.
There is also a danger of over-reliance on these technologies. First, writing is a skill that must be practiced in order to keep fluent at it.
Second, you would still need to read and edit anything written by the AI, and, as any writer will tell you, in order to be a good editor you must have a good understanding of the writing process.
Third, when you get AI to do all the heavy lifting and write the majority of your content, you may lose perspective on wider issues that may be relevant to your topic.
Finally, human-written content is still preferred over AI-generated content. It really comes down to quality over quantity. While AI can quickly draft lengthy blog posts, articles, and other content, this artificial content often lacks accuracy, personality, creativity, context, and linguistic nuance.
Who else is developing similar technologies?
Although ChatGPT does appear to be the main contender in the world of Natural Language Processing (NLP) at the moment, there are a few companies with similar technologies worth keeping an eye on.
Even though they are off to a rocky start, Google’s Bard is definitely ChatGPT’s main competitor and will likely push the NLP boundaries even further. Other high-potential players include:
ChatGPT Limitations
OpenAI makes very clear that their AI technology is by no means perfect, and a very valuable function built into the ChatGPT language model is its ability to candidly explain its own limitations:
“As an AI language model, ChatGPT has certain limitations that are important to keep in mind:
- Bias: ChatGPT may exhibit bias in its responses based on the data it was trained on, which may not always be fully representative or objective.
- Lack of common sense: ChatGPT may not have the same level of common sense that humans possess, which can sometimes result in responses that seem illogical or inappropriate.
- Limited knowledge: While ChatGPT has access to vast amounts of information, it is still limited by the data it was trained on, which may not include all relevant or up-to-date information.
- Inability to perceive the world: ChatGPT has no direct sensory experience of the world, so it cannot perceive the world like humans do.
- Inability to empathize: ChatGPT does not have emotions or the ability to empathize with human emotions, which may make it difficult for it to provide nuanced or compassionate responses.
- Language-specific: ChatGPT is designed to process natural language in specific languages that it was trained on, which may limit its ability to understand or generate responses in other languages.
It is important to note that while these limitations exist, ChatGPT can still be a useful tool for tasks such as language translation, text summarization, and question-answering, among others.”
When prompted to explain how it plans to resolve these limitations in the future, ChatGPT generated the following response:
“Addressing the limitations of AI language models like ChatGPT will require ongoing research and development in a number of areas. Some potential strategies for addressing these limitations include:
- Improving training data: One approach to addressing bias and knowledge limitations is to improve the quality and diversity of the training data used to train AI language models. This can involve efforts to collect more data from diverse sources, or to better balance the representation of different perspectives and experiences in the training data.
- Incorporating common sense knowledge: Researchers are exploring ways to incorporate more common sense knowledge into AI language models. This might involve building models that can reason about cause and effect relationships or that can infer missing information from context.
- Integrating sensory information: Researchers are also exploring ways to integrate sensory information into AI language models, so that they can perceive the world more like humans do. This might involve incorporating data from sensors or cameras, or building models that can simulate the experience of perception.
- Empathetic AI: Work is underway to develop AI systems that can recognize and respond to human emotions, which could help address the limitations of empathy in AI language models.
- Multilingual models: To improve language-specific limitations, researchers are developing multilingual models that can handle multiple languages, as well as cross-lingual models that can translate between languages.
Overall, addressing the limitations of AI language models will require ongoing research and development in a range of areas, as well as ongoing efforts to ensure that AI systems are developed and deployed responsibly and ethically.”
Why are schools and some large corporations banning ChatGPT?
ChatGPT has been very well received by the public in general, but some major concerns relating to the use of this freely available technology by pupils have been raised.
Some pupils have been found to use ChatGPT to quickly do their homework for them, summarise literature, and even solve complex maths problems.
For this reason, schools across the U.S. have restricted access to ChatGPT on their computers and networks.
Teachers are also very concerned that by using this technology, students will deprive themselves of attaining important skills at school, such as critical thinking and problem-solving.
The multinational finance giant JP Morgan Chase has become the latest in a string of large companies to ban the use of ChatGPT among their staff, due to compliance concerns with third-party software. Staff across the globe were asked not to enter any sensitive information into the chatbot.
Ikea, Amazon, Verizon, and Accenture have all taken similar precautionary steps. For most businesses, ChatGPT poses a risk to the following:
- Data safety and privacy laws. Businesses that are found to breach these laws can face significant penalties and fines.
- Plagiarism and intellectual property. The undeniably valid question of “Who does the content generated by AI actually belong to?” is a big problem for many reputable businesses. Plagiarism is a very serious crime, and some businesses are just not willing to take the risk with AI content.
- False, disinformation, and misinformation. In a world where false information is deliberately being spread, reputable companies understand the risks involved when solely relying on large language models to generate content that is hard to fact-check.
In Summary
ChatGPT has made some very impressive advances in the field of Natural Language Processing.
This Chatbot has even got some tech giants like Google on edge, which may not entirely be a bad thing.
We will likely see some exponential growth and advancements made in this field over the next couple of years, with all the various developers competing for first place, including OpenAi’s GPT-4, which has just recently been released.
These technologies have a lot of value when they are utilized correctly and with great caution. ChatGPT won’t replace professional writers, but it can greatly enhance the writing process and automate some tedious tasks. The key is knowing where the line between AI’s contribution and the writer’s contribution lie
