Artificial Intelligence (AI) is making tremendous progress leaps, as has been shown by the recently popular ChatGPT. However, we’re still at the level of narrow AI, otherwise called weak AI, that can only process simple tasks and still lacks the emotional depth and understanding of a human being and thus cannot perform on an equal level.
During Memex Poland 2022, Artificial Intelligence (AI) was one of the hot topics. One of the pressing matters involved the disintegrated knowledge between AI and human experience. Mateusz Bonecki (PhD), DAC.digital’s Head of Innovation, talked about, among others, the inability of AI to adapt and take different approaches as humans can. At the end of this summary, you can watch the video recording of the presentation.
Artificial intelligence definition
Artificial Intelligence (AI) is a computer system designed to perform tasks that usually require human work. We distinguish three main types of AI: narrow (ANI), general (AGI) and super (ASI).
Narrow AI can perform calculations and tasks of various complexity using computing power. However, it lacks the emotional depth and understanding to make it anything more than a tool. It’s the only type of AI we keep developing that is currently achievable.
General AI is still out of the current technology’s reach. However, if we could develop general AI, it would be equal to human beings in all aspects. It could build emotions and help us understand everything as we do.
Super AI remains the work of science fiction fantasy. This type of AI would surpass humans in every aspect. The Super AI would be capable of feeling emotions and forming relationships, forming a self-aware AI. Although considerably feared, Super AI will most likely remain out of reach for decades.
Capabilities and limitations
In his presentation, Mateusz highlights the issue of the narrow AI being tailored to performing one specific task and addressing one particular problem, no matter how vast. Be it machine learning, computer vision, natural language processing, or human speech, Artificial Intelligence (AI) can’t do more than focus on this one task.
The problem of knowledge discrepancy
Using the example of “Les Miserables”, Mateusz explained how even though the AI bot could carry out a meaningful conversation about the plot, characters and other nuances, it couldn’t react to some unexpected events related to it. For example, if a user calling themselves Javert entered the chat, the AI wouldn’t react differently, even though it knew the evil character from the story.
These discrepancies can create a gap that doesn’t allow the machine intelligence to look for a different path or find an entirely different solution. If a human being realised that a particular approach doesn’t bring the desired outcome, they would look for another way to solve it. Machine intelligence and computer science with limited memory don’t allow that.
Human intelligence, experience and AI
Although Artificial Intelligence and machine learning envelop an increasing range of possibilities and improvements, limited memory AI and artificial neurons can’t yet equal or surpass the narrow AI due to the knowledge discrepancy and current technology limitations. Our road to achieving general AI can still be a long one.
Watch the video recording of the presentation
Watch and listen to the entire presentation of Mateusz sharing his expertise and insights on AI’s current state and future.
Junior Content Specialist at DAC.digital