If you think that lack of computing power, shortage of manpower, and lack of trust are the challenges to AI, think again. Recently, Salesforce Chief Scientist Richard Socher has unveiled the grand challenge of them all: language. Here’s why.
LACK OF EMOTION
The current AI language doesn’t understand the emotional element in human communication. As a result, it doesn’t distinguish between anger and happiness, frustration and satisfaction, etc. It may reply to your statement, but it will a flat tone.
LACK OF CONTEXT
The current AI language doesn’t respond to questions in a contextual manner. It just does based on what’s built into it. So if you ask a machine contextual questions like the time when your plane will arrive, it will not distinguish between the aircraft and your parcel containing the carpentry tool – supposing that you’ve also ordered that tool for woodworking.
LACK OF COMPLEXITY
The idea for the AI language is to make it as complex as possible by allowing it combine simple tasks in order to accomplish more complex ones. The problem is, that hasn’t happened yet. Currently, AI language is set so that it can complete a series of simple tasks without the ability to complete complex ones.
LACK OF LEARNING
Related to lack of complexity, AI machines now do not have the ability to learn on their own. Humans still need to integrate additional programming into their systems in order for them to ‘acquire’ the ability to perform additional tasks. This entails that the human-machine communication is limited.
For Salesforce’s part, it has already created what it calls the Natural Language Decathlon. This is a program that seems to be promising answer to AI’s ability to answer questions with accuracy. However, NLD can only answer the 10 of the toughest tasks in Natural Langue Programming. In the end, Socher admits that developments in the area of AI language is still at its infancy stage.