Artificial Intelligence and Machine Learning Blogs
Explore AI and ML blogs. Discover use cases, advancements, and the transformative potential of AI for businesses. Stay informed of trends and applications.
Showing results for 
Search instead for 
Did you mean: 
Product and Topic Expert
Product and Topic Expert


Today, we meet them everywhere: Chatbots. Chatbots are conversational programs, answering user questions on websites, in apps or even at home like Amazon’s Alexa. And they can do even more for us: They can guide users and can even execute simple task, like starting a web search, turning up the volume or ordering a pizza. Chatbots are becoming virtual assistants in our daily lives. Many social networks opened their APIs to allow their users to interact with bots.

How Intelligent are Chatbots?

Even today, most chatbots are not intelligent at all. They do not understand the meaning behind a question nor do they do well in reasoning. Since the first chatbot was published in 1966 by Joseph Weizenbaum, most chatbots still use simple pattern matching and provide pre-prepared answers. Often it is the human willingness which helps traditional chatbots to pass the Turing test. In other words, many traditional chatbots only generate an illusion of understanding. Tutorials for building such chatbots can be found hundredfold on platforms like GitHub.

Advances in Machine Learning

Recent advances in machine learning and natural language processing allow for a new generation of chatbots, which can learn from interacting with users and do generate answers from knowledge sources such as structured text and documents. These AI chatbots do not only provide pre-prepared answers, but generate new knowledge themselves, and like IBM Watson can even beat Jeopardy champions. This brings us to a simple question: If chatbots know more than most humans beeings, why not learn from them? Why not use chatbots as teaching assistants in training and coaching?

Chatbots in Learning

A first AI chatbot for training students has been developed by Professor Ashok Goel. His teaching assistant Jill was based on IBM’s Watson and built to answer questions posted in a foum by students enrolled in a Georgia Tech’s online master program. Jill was able to answer thousands of questions in and most students did not notice anything unusual about their TA (teaching assistant). Goel anticipates that the use of AI in education will contribute to a style of learning that is “interactive and personal as well as immersive and social.” (source). We have now developed a model which describes the levels of learning facilitation by AI teaching assistants:

Levels of Learning Facilitation by AI

Leve 1: Welcome learners

The AI teaching assistant welcomes new learners with personalized messages.

Level 2: Recommend content and peers

The AI teaching assistant recommends learning content, next steps, peers, and experts for collaborative learning

Level 3: Answer questions

The AI teaching assistant answers typical questions posted by learners

Level 4: Set goals and monitor learning progress

The AI teaching assistant sets learning goals for learners, communicates them, and monitors the learning progress

Level 5: Provide feedback

The AI teaching assistant provides personalized feedback

Level 6: Individual learner coaching

The AI teaching assistant proactively coaches learners by analyzing individual learning needs, making individual recommendations and providing individualized feedback.


The higher the level of teaching assistance, the more AI and data is needed. This may be the reason, why most AI teaching assistants currently focusing on level 1 to 3, while real learning assistance starts at level 4. However, we are confident that with the advances in machine learning and big data analysis we will soon see first bots on higher levels.