ChatGPT is just the beginning: text assistants are going to get very personal
By: Enrique Dans
Grammarly, an interesting company with Ukranian roots that has been helping people (including many of my students at IE University who come from all corners of the world and speak, in many cases, English as a second language) to write correctly in English for almost 14 years, has just introduced a generative text assistant similar to ChatGPT to create texts written in any context based on a few simple instructions.
Grammarly is a grammar and spelling correction tool that uses natural language processing technology to identify errors in texts and provide suggestions for improving grammar, style and clarity. It can be used in a variety of contexts, such as academic, professional or personal writing, and is available as a browser extension, a desktop application or a mobile app. In short, it’s is a useful tool to improve writing and communication in English.
What makes its text assistant, called GrammarlyGO, interesting is that Grammarly’s functionality is available, in the form of a contextual menu, in practically any environment, and routinely inspects everything the user writes in order to propose corrections, which means that the pre-trained Large Language Model database will be partially based on the user’s own texts, and thus able to generate texts adapted to the user’s personal style.
See more: Shopping app usage hits new record
It will be like having functionalities very similar to those of a ChatGPT but customized and adapted to users’ writing style, and can be asked to answer an email, write a document, post something on a social network or many other things, ensuring on the one hand that the use of English is correct and that there are no typographical, spelling or grammatical errors; and on the other, that the style used, the tone, the choice of words, etc. is reasonably similar to the user’s own.
As the entry barriers to create these types of algorithms begin to lower, we can expect to see a lot more of them: assistants focused on a particular feature, but partially trained with data generated by the user, providing very interesting contextualizations adapted to individual needs. I have not yet had the opportunity to try GrammarlyGO, which will not be available for a few months, but the reasoning behind incorporating a series of user-generated materials into the training of the algorithm seems to me to make perfect sense.
Generative assistants like ChatGPT will evolve to an ever-increasing level of personalization. As more information is gathered about users and the capability of natural language processing models improves, more personalized and accurate assistants can be created. In fact, current generative assistants using conversational interaction models already have some level of personalization, as they can learn from interactions with users and adapt to the way they speak and express themselves. In the future, more advanced machine learning and deep learning techniques will be used to further enhance the personalization of the user experience.
That said, we should also expect our personal data, preferences, search history and behavior to be used to provide more accurate and personalized responses. In this sense, personalization can be a valuable tool to enhance the user experience and make interaction with generative assistants more natural and effective.
