After participating in the famous ChatGPT workshop held by Jordi, an importa question was raised, which has now become the main topic of this article:
Will natural language processing ierfaces (such as ChatGPT) replace graphical ierfaces?
In the past few years, chatbots have been proving their presence coinuously and sometimes iermittely, and their small arena has suddenly turned io a big bomb. However, the chatbots created at that time were not powerful enough to be recognized as a symbol of a revolution; Therefore, some of them accepted defeat and other examples coinued to work in silence.
In those years, technology was moving along its way and the user experience was far from the word “enjoyable”. But now several mohs have passed since 2023 and communicating with ChatGPT feels natural and scary to the user.
Natural ieraction
In the book “The Design of Future Things”, Donald Norman has discussed various aspects of design and human ieraction with technology. While the main focus of this book is on designing future technologies, Norman also takes a look at the concept of natural ieraction.
In his book, Norman emphasizes the importance of designing technology that is compatible with human ability and behavior. He acknowledges that technology should be designed to support natural ieraction. In addition, users should easily cope with how to ieract with the user ierface and corol smart devices without any complexity or training.
Norman also criticized the idea that users should adapt to graphical ierfaces and suggested that developers and technologists should create a product that adapts to users. One of Norman’s requests and the subject that has been paid atteion to in this book is the need to design a perceptual ierface that requires natural human skills and abilities to ieract with it; This topic includes the design of a device and a system that can communicate with a person through understanding, memory, and cognition and free the user from complex and diverse graphical ierfaces.
But now think about natural language processing. Thanks to this technology, there is no more ad hoc language and new semiotics, and this means achieving natural ieraction progress.
All that glitters is not gold
Although natural language processing has experienced significa advances in the past few years, we still face various challenges that remain unanswered.
- Understanding texts and common sense: Natural language processing models have always struggled to understand the texts and common sense used to accurately ierpret language. Models such as: ChatGPT 3, although it can give users’ answers in a reasonable way, but their wrong and meaningless answers are also obvious to everyone due to the limitation in understanding the vast texts of conversations and knowledge of the world.
- Dealing with ambiguity and resolving it: Language is inherely ambiguous, and natural language processing models may not deal well with disambiguation. Solving ambiguous references, word meanings or ierpreting and accurately and consistely translating phrases and idioms is a challenging task for natural language processing systems.
- Dealing with rare words and items that are not in the language model dataset: Linguistic models learn from a large set of data, but sometimes we may encouer rare words and items that are not available to linguistic models; Therefore, dealing with words beyond the reach of the linguistic model and updating it is still considered as a challenge.
- Observing bias and ethical issues: Natural language processing models are condemned to unieionally provide false information. These models may provide answers about gender, ethnic, and cultural biases that many users may not like. This situation is similar for ethical cases, and although many governmes and users are not satisfied with its treatme, it remains an unsolved challenge.
How to delete a ChatGPT accou with a few simple methods
- The basis of language in the physics of the world: Understanding the language in relation to the physics of the world was and is a challenge. While progress has been made in areas such as image description or answering visual questions, fully iegrating visual information or sensor data with language understanding remains a challenge.
- Understanding texts when the user’s conversation with the natural language model increases too much: Most of the time, natural language processing models can’t perform properly while augmeing the conversation with the user. Although it is iended for them to maiain a reasonable conversation and understand the topics, they can hardly and sometimes cannot maiain the quality of the conversation; A problem that some people also struggle with.
- Reasoning and providing explanations: While natural language models can provide somewhat reasonable answers to the user, they do not benefit from the ability to reason logically and clearly, and they remain sile when explaining their decisions.
Graphical ierfaces and features
Ideally, the desired design should provide visual and clear features and provide this possibility for the user to easily learn and predict how to use and ieract with the system and compones.
Feedback helps users to be aware of the results of their actions and provides information about the state of the system. Functional feedback helps the user to ieract with the ierface design and not get confused.
When graphical user ierfaces are designed correctly, they will have the ability to ieract with users in a way that users understand their abilities and capabilities and predict system behavior. In general, natural language ierfaces struggle to provide these evaluations and, like humans, you don’t know what to expect.
Conclusion
Given the curre limitations of the technology, it is unlikely that GUIs will be completely replaced by chat-based ierfaces in all situations. Differe types of ierfaces also serve differe purposes and each has its own advaages and disadvaages. Graphical ierfaces are useful for providing visual effects, combining complex data, and tasks that require precise input and corol. This ierface is suitable for tasks such as: graphic design, video editing or working with large sets of data.
On the other hand, chat-based ierfaces are desirable for tasks that boil down to natural language ieraction, personalized suggestions, and information retrieval. They can be used especially in customer support, personal assista and ieractive storytelling.
In many cases, a combination of these two ierfaces may be used to impleme a comprehensive user ierface. For example, a chat-based ierface can handle high-level commands and queries, while a graphical ierface can be used for finer-grained corol.
In conclusion, although chat-based ierfaces have the poteial to significaly impact ieractions with digital systems, they will not replace graphical ierfaces. The coexistence and iegration of differe ierface models will most likely shape the future of user ieraction.
