Today, my internship, which lasted for three and a half months, came to an end, and it is necessary to give it a summary…

Is Current Artificial Intelligence Intelligent?

I still remember the first time I became aware of the concept of artificial intelligence (AI) was from a very famous movie, Iron Man. Jarvis, the AI assistant in the movie, became my imagination of the ultimate development of AI. It could not only have independent thinking and control various electronic devices but also interact freely with humans.

However, looking at the current development of AI, it is still far from the level of intelligence portrayed by Jarvis. In fact, it can even be said that current AI is not truly intelligent. After studying neural networks, I believe that present-day AI is more like an answer generator. It can generate corresponding answers for a given question using pre-trained parameters. This actually differs little from earlier question-answering systems, where parameters were often manually designed based on knowledge databases. AI, on the other hand, learns through a large amount of training data. However, in my view, both approaches are fundamentally similar, as they use parameters to control answer generation. One approach is forward-thinking, determining the parameters, while the other is a reverse process, approximating the accurate parameters based on the training datasets. This is why I believe that current AI is not truly intelligent.

Furthermore, I have noticed that many people have limited understanding of AI. For example, many people believe that AI does possess intelligence, but it is just not advanced enough to solve complex problems. The popularity of ChatGPT has also made many people aware of the field of large language models, leading them to believe that AI has evolved to the point of understanding human speech. While it is true that the cognitive abilities of large language models have improved due to increased parameters and training data, this evolution is not in the biological sense. Therefore, it is problematic to simply assume that AI is evolving, which indirectly leads to the belief that AI is highly uncontrollable and opposes its development.

Lastly, I have some superficial thoughts. In my view, the remarkable aspect of the human brain lies in the fact that it often generates numerous sparks of thought, helping us solve problems. For example, coming up with a solution to a problem while sitting on the toilet and watching phones. This characteristic is precisely what present-day AI lacks or has not been added. Therefore, perhaps such “mental distractions” can contribute to the development of AI (laughter).