...

What About Transforming From NLP to NLU?

Can you think of a world where complete and satisfactory human-computer conversations exist?

With the level of interest in natural language processing (NLP) and Natural Language Understanding (NLU) – the world we are imagining is not away.

The difference between NLP and NLU

In order to provide more convincing interactions with computers, researchers are designing intelligent systems using artificial intelligence (AI) to develop a better understanding of human actions and queries.

In 2018, Open AI (Open AI is an AI research and deployment company) used an ml technique (reinforcement learning) to teach agents to design their own language. They were given a simple set of words and the ability to communicate with each other. They were then given a set of goals that were best achieved by communicating with other agents. 

 

OpenAI Releases Jukebox AI, A Music Generating Neural Net

Hence on their own, agents develop ‘grounded’ language.

Now, what is ground language?

 

A network of artificial neurons learns to use human language

 

Human language is said to be grounded language. People like us, grasp the basic words and their meaning by interaction and experience– not by learning dictionary definitions. We have an understanding in terms of sensory experience -- for example, words like yellow, light, and below.

Ground language allows people to understand words and sentences in contextual form.

The opposite of a grounded language is called Inferred language. Inferred languages extract word meaning from the word itself and not what they represent. But AI techniques work on textual data therefore they lack in knowing what the word actually means.

Can Agents develop a language that we won’t understand?

 

AI Research in Germany

Yes, it can happen. Though the researcher gives agent simple English words, the agent endlessly turns to its own, unintelligent language. 

Recently researchers at Facebook, Google, and Open AI all experienced this kind of situation.

Agents are reward-driven technicalities. If there is no reward for using human language then the agents will develop a more efficient language for themselves.

Is it good or bad?

ETCentric Facebook AI research

When the Facebook Artificial Intelligence Research lab’s researchers designed chatbots to workout with each other using machine learning, they had to change one of their models otherwise the chatbot-to-chatbot conversation would have led to the divergence from human language. So the agents developed their own language for negotiation. They used a fixed supervised model instead.

There is only one problem - transparency. Machine learning techniques like- deep learning are black-box technologies. A lot of data goes into the AI. In this case ML technique - the neural network was used to train and develop its own rules. The model is then fed new data which is used to spit information out. The black box technology is used because it is very hard. It’s not impossible in complex models, to know how the AI gets the answers. If AI develops its own languages when talking to other AI, the transparency problem is generated. How can we fully trust AI when we don’t exactly know how it is making its decisions and what it is telling other AI?

But it does demonstrate how machines are redefining people’s understanding of so many things. The Facebook researchers concluded that it offered them a satisfying insight into human and machine language. The chatbots also proved to be very good negotiators, developing intelligent negotiating strategies and systems.

These new insights will lead to smarter chatbots that have a greater understanding of the real world and the context of human conversations and sentences.

Chatbots – The Future!!

Chatbots are a key technology in all true meanings; it could allow people to consume data analytics without realizing that’s what they’re doing.

Chatbots-the future

Chatbots create a human-like interaction platform that makes results accessible to everyone. The transformation of NLP toward NLU will have a lot of important implications for businesses and consumers alike.

Convincing human-computer dialogues/conversations will soon exist and will have applications in fields like medicine, law, sciences, sales, education. As the volume of unstructured information is continuously growing in an exponential way, humans will benefit from AI’s unique ability to help us make sense of all the possibilities.

 

payal_mahajan

A professional Coder with sheer knowledge in much technical domain. we very highly value her.