Natural Language Processing (NLP) models have recently gained a lot of media attention and traction. This is due to the significant evolution they have undergone in the last few years, driven by the availability of large amounts of data and computing power, as well as the advances in techniques. Some of the most significant advancements in NLP, that have led to impressive improvements, are the transformer architecture and techniques like pre-training.
In fact, these advancements stand at the core of the GPT-n series, developed by OpenAI. GPT stands for Generative Pre-trained Transformer and it is a set of very powerful language generation models. GPT-3, introduced in 2020, is the latest iteration and can generate human-like text by processing input prompts, after having been trained on a vast amount of data. Its performance in “understanding” and creating coherent text has led it to be used in a variety of applications. According to an OpenAI article (2021), the model has been successfully used for tasks like improving semantic search in news archives (Algolia) or generating summary analyses from customer reviews (Viable). In a natural progression, OpenAI itself used the GPT-3 model as a base for more specialized applications - the famous Dall-E (the image from text generator) and ChatGPT.
ChatGPT
ChatGPT is a fine-tuned version of GPT-3 for conversational purposes. Designed for dialogue settings, the model has been released to the public together with a friendly interface, as a chatbot. In its own words, ChatGPT is “able to understand and respond to natural language inputs […]. [It] can provide information and answer questions on a wide range of topics”.
In addition to generating replies—or content, in general—that are difficult to distinguish from ones created by humans, the chatbot can eloquently answer follow-up questions and ask its own ones when the input seems too ambiguous. Thus, one of its key features is the acknowledgment of contextual information, such as understanding implied subjects from previous pieces of the conversation.
Some of the most interesting use cases demonstrated were the summarization of highly complex concepts, like mathematical theorems, as well as code debugging help. You can see these showcased in this official OpenAI blog post or explore ChatGPT’s capabilities on your own by clicking on the following link: https://chat.openai.com/chat.
Another interesting use case that sparked interest in both technical and non-technical communities is using it as a personal assistant. One can ask the chatbot to write email templates, guide it to follow specific structures in doing so, and write follow-ups and to-do lists. In addition to this, tech-savvy users have attempted to integrate it with voice commands and also equip it with text-to-speech capabilities. These kinds of examples showcase the diverse and creative ways in which generative language models could prove useful to individuals and businesses from different backgrounds. It is important to note that, at the time of writing, ChatGPT is free to use as part of the initial research preview release. There is no available API for developers. However, the original model it’s based on, GPT-3, is available through OpenAI’s API.
Where do we go from here?
Given the worldwide attention, curiosity, and appreciation that ChatGPT has been received with, it seems highly likely that more and more initiatives surrounding this powerful model will emerge. On the one hand, research in this area may attract even more investment, from competitors and investors alike. For example, we can already see that Google has been working on its own transformer-based language model (Language Model for Dialogue Applications - LaMBDA), while Microsoft has recently extended its financial partnership with OpenAI. On the other hand, many companies may adopt generative language models, after recognizing the potential they have for the business. For instance, we can expect more seamless online customer service through specialized chatbots, or AI-assisted copywriting.
Nevertheless, it looks like generative technologies are something to keep our eyes on in the future. If you have an idea about how generative language models could benefit your business, Mission Automate has a team of specialists that can help you implement it.
References
[1] https://openai.com/blog/chatgpt/
[2] https://www.sciencefocus.com/future-technology/gpt-3/
[3] https://www.sciencefocus.com/news/chatgpt-scientist-openai-chatbot/
[4] https://www.youtube.com/watch?v=dErBWj3w6UU
[5] https://medium.com/@davide.gazze/my-italian-personal-assistant-via-python-and-chatgpt-2344e466e1de
[6] https://help.openai.com/en/articles/6783457-chatgpt-faq
[7] https://beta.openai.com/docs/introduction
[8] https://openai.com/blog/openai-and-microsoft-extend-partnership/
[9] https://www.markettailor.io/blog/how-businesses-can-use-chatgpt-for-content-marketing