ChatGPT: will robots soon write our texts?

We have already explained the necessary AI tricks in recent months and the end is not yet in sight. This month we look at ChatGPT, the new AI tool that writes text based on certain inputs. How does it work and will it soon be robots that write our texts? We’ve already explored the world of artificial art using tools like MidJourney, and seen how you can use practically any person in the world in your own movies through deepfakes. The advancements in artificial intelligence seem to be endless.

This time it is the turn of writing text, because yes, we can now have artificial intelligence do that too. If you ever scroll through web texts or news items online, chances are you have already come across a few articles written by an AI. These are often so-called clickbait articles that are generated to attract as many people as possible to a website. The difference is that with those articles it is often clear that something is not quite right. But writing by robots is also evolving very quickly. We may be moving towards a future where an AI writes natural-looking texts. And the new hype in technology is another step in that direction. Meet ChatGPT.

What exactly is ChatGPT?

ChatGPT is a large AI language model trained to generate ‘human’ texts based on input from its users. The tool uses the GPT-3 (Generative Pretrained Transformer 3) architecture to generate text that is coherent and reads naturally. One of the most promising potential uses of ChatGPT is in chatbots and virtual assistants. Using ChatGPT allows these systems to have more natural and engaging conversations with users. This can significantly improve the user experience, making it easier for people to get the information they need or complete tasks without having to deal with frustratingly stern or unnatural conversations. Another potential application for ChatGPT is machine translation. Because it has a broad knowledge base and can generate coherent text, the tool could be used to accurately convert text from one language to another. This would be a huge improvement over current translation systems, the results of which are often strange or difficult to understand. In addition to chatbots and translations, ChatGPT could also be used for text summaries. By analyzing a large amount of text and reducing it to key points, ChatGPT can be used to quickly create summaries of long documents or articles. This can save time and make it easier for people to quickly extract the most important information from a text.

Read also: What is ChatGPT? All about the intelligent chatbot

How does GPT-3 work?

GPT-3 has been trained on an enormous amount of data to create text that is as realistic as possible. The exact data used to train GPT-3 is not public, but it is widely known to include a wide variety of text from various sources, such as books, articles, and websites. The goal of GPT-3’s training was to give it a broad knowledge base and the ability to generate highly coherent and natural-sounding text. To achieve this, the data to train the model was carefully selected and curated by the OpenAI team to represent a wide range of topics and writing styles. Using such a large and diverse dataset is one of the main factors that sets GPT-3 apart from other language models. As a result, GPT-3 has a broad understanding of the world and the tool generates text that is highly coherent and reads naturally, making the technology useful for a wide variety of language processing tasks.

But isn’t this plagiarism?

No, GPT-3 is not plagiarism. Plagiarism is using someone else’s work without giving due credit or treating someone else’s work as your own. This is not the case with GPT-3, as it is a machine learning model that is trained on text data, but does not copy or steal that text. Instead, GPT-3 uses its training data to learn the patterns and structures of human language, and uses that knowledge to generate its own text. This text may resemble text it has seen before, but it is not copied verbatim from any particular source. GPT-3 is therefore not plagiarism.

Not really?

Well, you can discuss it. The last three headings of this article are in part by ChatGPT. Of course we made some minor adjustments here and there, but 90 percent of the text was created by artificial intelligence. So also the piece above about plagiarism, and that remains a difficult subject. In ChatGPT you can just ask simple questions like ‘What is ChatGPT?’ and then the AI ​​gives you a response. The questions we used to create these texts were: ‘Explain ChatGPT’, ‘Why is ChatGPT useful?’, ‘Is GPT-3 potentially dangerous?’, ‘What data is used for GPT-3?’, ‘ Is GPT-3 plagiarism?” and more. You can also ask the tool to write an article about ChatGPT, but unfortunately that was not long enough to publish. That was very easy. From our experience, the texts are mostly Wikipedia-like texts that are very helpful, but don’t contain an ounce of ‘unique’ content. Still, it is sometimes scary to see what is possible, but this does not mean that the tool has no flaws. If you ask ChatGPT for a ‘literary’ description of a Christmas tree, you get a nice but extremely cliche description of the Christmas tree with its twinkling lights, the typical pine scent and the beautiful Christmas decorations. If you then ask to write a story about a couple spending Christmas together, he will write that story, but often the ‘originality’ is missing. This is a very abstract concept, but you notice it immediately in some texts. In addition, ChatGPT cannot provide content that does not exist. For example, if you ask what one of the partners is thinking at that moment, you will get the answer that ChatGPT cannot know what someone is thinking, because it is an AI tool. But who knows, maybe GPT-4 will come up with a solution in a few months or years.

What is the future of GPT-3?

The future of GPT-3 is difficult to predict with certainty, but there are a number of possible developments. One possibility is that GPT-3 or similar large language models continue to be developed and refined, leading to even more sophisticated and powerful AI systems. Another possible development is the use of these language models in more applications. Currently, GPT-3 is used for tasks such as translations and text summarization, but in the future it could be applied to even more complex problems. For example, it could be used to generate more realistic and engaging texts for games or other interactive applications. In addition, the development of GPT-3 and other large language models can also lead to further research into the ethical and social implications of AI. As these technologies become more powerful and more widespread, it is important to consider the potential risks and challenges they may pose and develop strategies to address them. In general, the future of GPT-3 will likely be characterized by continued innovation and advancements in AI. It’s an exciting time for language processing technology, and we can’t wait to see what the future holds for GPT-3 and other major language models. Oh, by the way, this paragraph was also largely written by ChatGPT.