There has been a lot of buzz about it recently among the AI development community. What is ChatGPT really? In this article, we will try to answer this question as well as consider what the future holds for this tool.
Table of contents:
General information about ChatGPT
In a nutshell, ChatGPT is used to generate elaborate responses to user input. Although a more illustrative definition for this tool is a natural language response system.
ChatGPT was developed by San Francisco-based artificial intelligence company OpenAI. The company's main investor and partner is Microsoft, which has invested $1bn in the company's development. They previously developed the Azure AI platform.
What makes ChatGPT different from classic search engines?
Everyone has spent a lot of time searching for the answer to a specific question among the vast amount of data available on the Internet. It often becomes time-consuming and sometimes impossible to find the answer. It is certainly easier for those who are more familiar with how popular search engines such as Google, Bing, etc., work. Even if you know what keywords you need to type into a search engine, it is still not an easy task to filter the results returned from the web. The tool mentioned above already comes to your aid.
According to the Statista portal, in 2022, the amount of data/information created, captured, copied, and consumed worldwide equates to 97 zettabytes of data. In 2025, it is expected to be as much as 181 zettabytes.
The Internet contains both structured data, such as databases and spreadsheets, but also unstructured data, such as all kinds of documents, blog articles, and replies to specialist forums. Tools that answer questions in natural language are already coming to the rescue.
So what is this system, and how does it differ from a classic answer search? In contrast to classic search engines, ChatGPT does not return a complete document to the user. Instead, the users ask a question in his or her natural language and gets a specific answer in return. In other words, the user asking the question receives direct information without having to search through an often extensive document.
How does ChatGPT work?
The current tool is based on the GPT-3.5 model. An additional layer of learning is RLHF, which uses human feedback to make the answers more satisfying for humans. GPT-3 has 175 billion parameters and was trained on 570 gigabytes of text. In comparison, its predecessor, GPT-2, was more than 100 times smaller at 1.5 billion parameters.
This increase in scale has resulted in the current version being able to perform tasks such as writing essays and dissertations compared to the older version. It handles this quite well, and for proof of this fact, it can be mentioned that the ChatGPT tool is banned on all devices and networks in some public schools in the US due to "concerns about negative effects on learning".
ChatGPT drawbacks and limitations
But ChatGPT also has drawbacks, as the developers are consciously making it clear. Many machine learning applications optimize for simple metrics that are only rough proxies of what the designer intended. This can lead to problems such as YouTube links promoting clickbait.
Another limitation of the tool is that it is designed to avoid providing toxic and harmful responses. Therefore, ChatGPT currently avoids answering such questions. It also sometimes provides wrong answers that seem correct at first glance. One of the first to notice this were the employees of the coding platform Stack Overflow. At the peak of ChatGPT’s popularity, the platform – well-known among programmers – was flooded with answers, many of which turned out to be wrong. This prompted Stack Overflow administrators to ban all user posts based on answers generated from ChatGPT.
The success of ChatGPT
Despite these obvious drawbacks, ChatGPT has been an undoubted marketing success. It gained one million users in five days (by comparison, it took Netflix 41 months, Facebook 10 months, and Instagram 2.5 months).
Currently, ChatGPT is a free tool, and it seems that it will remain so at least until the next version of the software is released. OpenAI is currently focused on improving its tool and is encouraging its users to test the platform. The product owner not long ago ran a competition with a prize of $500 in ChatGPT credits to encourage the public to rate answers.
What will the future hold?
The most important question about this tool at the moment is: what is its future? There have been claims among experts that ChatGPT could replace Google in the future. Microsoft itself reportedly has plans to integrate the tool into its Bing search engine, which is currently not as popular as Google.
At the moment, Google engineers can rest assured that in its current form, ChatGPT is unable to threaten their search engine, but it is not impossible that this may change in the future. Google itself is currently developing a similar tool called LaMDA, which has not currently been released. It is currently used by a small group of developers developing the tool.
Market rumors claim that LaMDA was never released because the chatbot AI conversational platform did not work with Google's current advertising strategy, which generated 81% of its revenue in 2021. As chatbot AI is known to mimic human speech from the Internet and spread certain types of hate speech, fake news, racist and sexist statements, and a high degree of informational inaccuracy, there may also be reputational risks.
One thing is certain, the current success of ChatGPT will result in increased investment in the development of the google tool, which will most likely bring us all a breakthrough in searching for answers on the Internet in the coming years.