The AI and deep learning revolution that we are currently witnessing has suddenly given birth to a new generation of machine intelligence that has the capacity and capability to identify patterns in astronomical data sets with accuracy and efforts are being made to achieve intelligence that matches if not surpasses those of humans.
The type of Artificial Intelligence that we have achieved so far is Narrow Artificial Intelligence or Narrow AI. Narrow AI as the name may suggest is good at performing a single task, such as playing games such as chess, recognising images, recommending items for purchase, weather forecasting etc. Fancy terms such as Computer vision, natural language processing (NLP) are still at the current stage of narrow AI even though the future may seem to be very promising.
In simple terms, narrow AI works within a very limited context and cannot do tasks beyond for what it was created. For example, an image recognition AI cannot order food for you as for that we need to create another AI. For this reason, we sometimes call Narrow AI as Weak AI. Weak means its limited in its use and not that its incompetent. The routine work that is getting replaced by AI and threatens to displace many humans is all because of Narrow AI. The reason being Narrow AI is highly efficient and effective in identifying patterns and correlations much better than any humans can do.
But Narrow AI is no human level Artificial Intelligence, its miles behind! The human level of AI is called Artificial General Intelligence or AGI.
The more we try to achieve human level AI or AGI, the more we realise the miracle of nature, human brain is. Humans might not be able to identify patterns and correlations as computers do but human brain can think, plan and act. Juggle up multiple tasks at the same time and can even think on multiple issues simultaneously. To look at power of human brain just think for about a minute and you will realise that “the human brain is a most unusual instrument of elegant and as yet unknown capacity”.
Importance of Context in achieving Artificial General Intelligence.
In our endeavour to make AI applications replicate the power of human mind or move a step closer to AGI would be to provide them with context, providing the AI with related information which can be used to derive solutions to the problems. The more contextual information an AI gathers, the more effective it becomes. Contextual awareness is about making AI more compelling in the sphere of end user experience. Contextually aware AI can respond to the needs of customers by comparing their requests or problems to huge databases of layered neural information. In the consumer world, a contextually powered AI could remind you to make phone calls, stock your milk supply, or check your WhatsApp messages. Imagine a refrigerator that can order a replacement part and request a technician to your door long before you notice a fault, just because it’s contextually aware of common problems with the hardware, or a refrigerator that orders eggs because it knows you’re running out.
Graphs as provider of context to AI?
To a layman a graph is a systematic way of putting data related to facts, people, business, location, events and many other attributes including transactional and preferences, to create interconnected visual results that are more accurate and relevant.
Graphs are extremely powerful. When they are based on semantic standards, it is possible to relate knowledge to language in a structured way. In this way, language can provide a way into the graph. You can match words to the concepts represented by the graph.
Graphs also allow you to create structures for the relationships. You can tell a graph that a company have employee and employee can be a manager and two employees can be co-workers, and all of these are people. Providing such descriptive information allows new information to be inferred from the graph such as the fact that if two people have the same manager, they must be in the same group. Because they are pictorial, knowledge graphs are more intuitive. Not only People but machines can also immediately understand graphs. When a knowledge graph is drawn, it makes sense to most people and I am sure it will make more sense a very powerful sense to AI.
In addition, graphs are great at capturing scattered, unstructured, incomplete, and sketchy information, just like our brains. So, it is not that difficult to think and imagine how beautifully graphs can help AI achieve the next level of sophistication. As the requirements of machine reasoning and machine learning tasks become more complex, providing computers with context is not a simple exercise—it’s a complex problem for which computer scientists are experimenting with many different potential solutions graphs may be the best bet on which AI can rely.
As we move forward...
What do you think? I am sure there are use cases where graphs may be powering AI with context, but they may be few or at experimental stages. Would love to hear how graphs are helping AI with context in real world applications or this is something that is still in concept stage.
References
1.Why Knowledge Graphs Are Foundational to Artificial Intelligence. Jim Webber,March 2018 Datanami
2.Knowledge Graphs And Machine Learning -- The Future Of AI Analytics? Bernard Marr,Forbes June 2019.
3.The Knowledge Graph for Intelligent Systems By Haikal Pribadi Founder and CEO of GRAKN.AI