Advanced language proficiency for AI such as translation and summarization is reaching human ability. Modeling human cogitation, AI will gain new capabilities through varied and diverse research efforts. Such ability includes flexibility by applying knowledge and experience once learned and logical thinking by inference using causality.
Having gained notoriety based on its successful image recognition, third-generation AI continues to strengthen its capability in the field of language processing. This is enabling AI to genuinely assist in the intellectual activity and capacity of humans.
The end of 2018 saw the emergence of an AI that takes in huge amounts of discourse directly from the internet, learns language rules and performs various processes that involve language, sometimes at a level that exceeds human capabilities. Stimulated by this progress, more learning data, robust computing power and efficient algorithms are being generated daily. Thus the global AI development competition continues.
While AI's ability is getting closer to humans, many issues remain. AI models lack versatility, and a different one is needed for each application. The one-off nature of this makes it inefficient. Additionally, AI performance depends on the amount and quality of learning data. This means that AI is of little value in fields with very little historical data.
For this reason, efforts are continuing to create AI that overcomes these restrictions. Just like human children learning about the world, this future AI will accumulate and apply the knowledge it has learned in one area to others, hypothesizing and validating to establish new rules. The continued development of such technologies will likely improve the versatility and capacity of AI learning, making it far closer to that of the human brain in order to apply knowledge in diverse ways.