LAGOS, Nigeria (VOICE OF NAIJA) – Artificial intelligence chatbots have had a big influence since they were made available to the public and have ignited a race for worldwide development.
However, they have also given rise to worries about the possible damage they pose to mankind.
Users have been enthralled by OpenAI’s ChatGPT since it was unveiled in November, and conversations regarding its implications for numerous businesses have been sparked.
It was anticipated that ChatGPT-4, which was just released in March, would be much more potent than its predecessor.
However, recent research from Stanford University and the University of California, Berkeley suggests that creating AI will be quite difficult. It was revealed that ChatGPT is now less efficient in carrying out several fundamental arithmetic operations.
This reduction is due to a process known as “drift,” when attempts to improve one part of complex AI models lead to the detriment of other components.
READ ALSO: Nigerians More Interested In Artificial Intelligence Than Ever – Google Trends Reveal
James Zou, one of the study’s authors and a professor at Stanford who is associated with the AI lab, claims that changing AI in one way can unintentionally make it perform worse in other areas.
It is difficult to continually improve AI models given this issue.
The researchers discovered that ChatGPT 3.5 performed better than ChatGPT 4 in a variety of activities, including resolving mathematical problems.
This difference draws attention to the problem of “drift” and how it affects the development of AI.
Although AI chatbots have demonstrated outstanding capabilities, this research underscores the necessity for developers to recognize and minimize any possible downsides in their quest of innovation.
This study offers insightful information on the challenges of AI development and the importance of giving improvements considerable thought.
It serves as a warning that even the most cutting-edge AI systems may run into unforeseen difficulties and failures.
It’s critical to achieve a balance between development and the preservation of fundamental capabilities as the area of artificial intelligence continues to advance.