On December 21, 2022, New York Times carried an article: “ChatGPT and Other Chat Bots Are a ‘Code Red’ for Google Search”. ChatGPT is being considered the software industry’s next big disrupter. In this article, we are going to examine and understand ChatGPT.
What is ChatGPT?
ChatGPT can be understood as a smart computer that can write like a person. It can understand and respond to text inputs and can write paragraphs, stories, articles, etc. It can also solve mathematical equations and can answer questions related to various fields of study. It internally uses machine learning techniques to generate human-like text.
How is it different from Google Search?
Google search is a web search engine that is designed to help users find information on the internet.On the other hand, ChatGPT provides information in clear and simple sentences instead of internet links. It explains concepts in easy-to-understand ways and can perform a wide range of language-based tasks.
What does GPT denote?
GPT stands for “Generative Pre-trained Transformer”. It is a type of neural network model that is trained to generate text.
In layman’s terms, GPT is a computer programme that can understand and generate human language. It’s like a virtual writer that can write sentences and paragraphs that sound like they were written by a person. The programme is trained on a big dataset of text and can understand patterns and structure of human language, so it can write text that is similar to what a person might write.
ChatGPT is a specific version of GPT that is fine-tuned for conversational use cases, like in a chatbot or a dialogue system.
How can we access ChatGPT?
The easiest way to access it is by browsing the URL: https://chat.openai.com/chat. User needs to register/log in before they can access it. Apart from this, there are several other ways to access it; for example, OpenAPI has exposed an Application Programming Interface (API) for developers to access it.
OpenAPI?
OpenAPI (or OpenAPI initiative) is a research and development company working on Machine Learning (ML) and Artificial Intelligence (AI). By the way, the term OpenAPI is also used for a set of open-source specifications used for building APIs.
Machine Learning, Artificial Intelligence, open source!
Machine Learning is a method of teaching computers to learn from data, identify patterns, and make decisions with minimal human intervention. The easiest (though not the best) example is Gmail spam-identifying software. Every time you mark one email as spam, Gmail is able to identify similar emails and mark them as spam. In a sense, the Gmail spam filter is continuously learning to identify emails as spam and/or legitimate.
Artificial intelligence is the simulation of human intelligence in machines that are programmed to think and learn like humans. One example is any game that is played against a computer. The computer (thanks to AI) is able to respond based on user inputs.
Open source refers to a type of software whose source code is available to the public. Normally, they are free to use, modify and distribute.
Is ChatGPT open source?
The original version of the ChatGPT model and its codebase are not open source. GPT-2 and GPT-3 (which are similar to ChatGPT) are open-source though and their codes are available on the GitHub platform.
Is Elon Musk behind it?
ChatGPT is developed by OpenAI; one of this organisation’s co-founders is Elon Musk (the others are Sam Altman, Greg Brockman, and Ilya Sutskever).
So, what can ChatGPT do?
ChatGPT has the ability to perform a wide range of language-based tasks. For example, it can
Figure 1. An essay
- Language Translation: ChatGPT can translate text from one language into another.
- Text summarisation: ChatGPT can summarise long text into shorter, more concise versions.
- Text completion: ChatGPT can complete a given text prompt; for example, it can write a story or a news article based on a given headline.
- Question answering: ChatGPT can answer questions based on given context and information.
- Dialogue generation: ChatGPT can generate natural-sounding dialogue between multiple characters.
- Sentiment analysis: ChatGPT can understand the sentiment behind a given text, whether it is positive, negative or neutral.
- Text classification: ChatGPT can classify text into different categories, such as news articles, opinion pieces, product reviews, etc.
- Text generation: ChatGPT can generate new text that is similar to a given input text.
- Language model fine-tuning: ChatGPT can be fine-tuned to perform specific tasks, like writing poetry, composing music, and more.
Figure 1 shows ChatGPT response to one such task.
Is the response trustworthy?
Figure 2. The correct answer is 7
No! ChatGPT is not foolproof, it can make mistakes and produce incorrect output. One of the main limitations of any AI model (including but not limited to ChatGPT) is that it responds based on the data it was trained on. This means ChatGPT won’t be able to provide a correct response when faced with inputs or tasks that it has not seen before or are outside of its training data.
What are other limitations?
Some of the limitations of ChatGPT are:
- Bias. ChatGPT has been trained on a dataset of text. It may reflect the biases and stereotypes present in the data, which can result in biased, discriminatory or stereotypical responses.
- Limited understanding of context. ChatGPT may not be able to fully understand or respond to complex or nuanced questions or situations.
- Lack of common sense. ChatGPT is not able to understand common sense reasoning and its understanding of the world is based on the data it was trained on.
- Lack of creativity. While ChatGPT can generate text that is similar to human-written text, it does not have the same level of creativity and imagination as humans.
- Lack of understanding of sarcasm, irony or other forms of figurative speech.
- ChatGPT was trained in 2021. It’s not been updated with the latest data. This may result in obsolete and/or stale responses.
Will ChatGPT replace Google Search?
It’s too early to tell. As such ChatGPT has not been designed to work as a search engine but there is a big overlap between what it does and what search engine like Google does. Though not completely replace, but future version of GPT can definitely have an effect on Google Search.
I asked this question to ChatGPT and it responded: “ChatGPT is a tool for generating text, while Google is a tool for finding information on the internet. ChatGPT could potentially be integrated into other systems, like search engines or chatbot, but it is not intended to replace them.”
ChatGPT is a powerful tool. It just shows a glimpse of the power of artificial intelligence and machine learning. OpenAI is already planning the next version of GPT (called GPT-4). Though ChatGPT has a long way to go, yet this is something that is interesting and has got a lot of potential in shaping the future of technology.
[The writer is Sr. SAAS Engineer associated with Fenergo. This article is written with the help of the ChatGPT application. Some portion of the article is based on the response it provided.]