OpenAI has served up its vision of the AI assistant of the future by announcing "Her"-like voice upgrades for ChatGPT. But while technology advances, the question arises whether its true capabilities are being deliberately limited.
How close are we to technology truly understanding us? With the new upgrade of ChatGPT, OpenAI promises a real interactive experience, but those of us who use it on a daily basis know that the potential is still untapped. Is OpenAI creating the tool of the future or is it just consciously holding back the key to a true AI breakthrough?
On Monday, OpenAI unveiled a new feature of its voice assistant that not only responds to human speech, but can also read human emotions and translate spoken words in real time, with a remarkable demonstration. The ability to pause the AI while speaking and continue without prompting again sounds revolutionary, adding a new dimension to human-machine interactivity. This approximation of the character played by Scarlett Johansson in the movie "She" is not a coincidence — it is a reflection of OpenAI's efforts to make artificial intelligence more human.
However, as everyday users of ChatGPT, we know that despite the apparent progress, there are still barriers preventing AI from truly developing to its full potential. From inconsistencies in responses to illogical errors that prevent automation without human oversight, many of its abilities appear to be intentionally limited. Why would OpenAI, which has the technical know-how and capacity, hold back the development of such an advanced tool?
In North America, ChatGPT users have been able to take advantage of the “Memory” feature for some time now, which remembers everything you tell it to do. This could be a breakthrough in interacting with AI, as you can teach it how to address you and respond based on your preferences. But still, ChatGPT seems to live in a joke — between promised usefulness and actual unusability. In order for it to be at least partially functional, it is necessary to use a series of external applications, which calls into question the autonomy and independence of this tool.
One of the most frustrating limitations in using ChatGPT is that, despite the availability of the “Memory” feature in the main models, customGPTs — customized models for specific purposes — still do not have this capability. This means that each customGPT operates in a vacuum, with no ability to learn or adapt based on past interactions. This limitation greatly reduces their usefulness and autonomy, as they cannot build continuous knowledge or follow long-term trends in dialogue. Furthermore, without memory, customGPT cannot tailor its responses to individual user needs and preferences, which is critical to creating a truly personalized and effective AI experience. Ironically, while OpenAI advances in voice and visual technology, the key functionality that would transform customGPTs into powerful tools remains elusive, leaving users and developers in a state of expectation and constant desire for more functionality.
The biggest challenge facing the development of ChatGPT is not only technological, but also of a practical nature: multimodal usability, or, in other words, connectivity with other applications. In an ideal world, ChatGPT would be able to seamlessly access your calendar, automatically reschedule or add appointments, even make phone calls on your behalf to report being late. At present, however, this level of integration seems to be a long way off. Despite the fact that ChatGPT can now "see", "hear" and even sense our emotions, the question remains whether these capabilities are really a step forward or just impressive and basically useless gimmicks. While OpenAI is admirable for its innovation, it often feels like developers are just teasing us with the potential of AI, while key functionalities that would truly increase ChatGPT's usefulness remain locked behind security and ethical locks that slow down real progress in human-machine interaction .
In the future, it seems that ChatGPT will become a permanent fixture on our devices, in part due to its integration with various systems via applications, which is already indicated by the recently launched desktop application. This application is a promising step towards fulfilling the long-term expectations we have of artificial intelligence: performing certain tasks automatically or semi-automatically. For example, the app could automatically judge which emails are important or alert you to traffic jams on your route. This ability of ChatGPT to act as a proactive assistant that anticipates and reacts to the user's needs opens the door to many practical applications in everyday life, where AI will be able to serve as a key mediator between the complexity of the digital world and our daily needs.
Perhaps OpenAI really wants ChatGPT to be a versatile and interesting interlocutor, but this seems to serve more than to mask the real limitations facing AI development. While the latest upgrade may work impressively, the real test will be whether OpenAI will ever open the door to ChatGPT's full potential, or whether we will continue to witness a technological “game of thrones” where real power is hidden behind a veil of security and ethical concerns.
