Cognitive Computing In The Next Decade Of AI
The year 2020 is upon us, and I am hearing across the news that this will be the next big decade in major technology advancement for artificial intelligence (AI). For example, HealthData Management reports that 93% of respondents are expecting an augmented workforce, with people, robots and AI working closely together. AI is expected to touch all domains of not only healthcare, but financial services, supply chain management, and even law. I am particularly excited about the next advancement of AI known as cognitive computing.
Perhaps Dharmendra Modha, IBM fellow, framed it best when he said, “Cognitive computing goes well beyond artificial intelligence and human-computer interaction as we know it – it explores the concepts of perception, memory, attention, language, intelligence, and consciousness. Typically, in AI, one creates an algorithm to solve a particular problem. Cognitive computing seeks a universal algorithm for the brain. This algorithm would be able to solve a vast array of problems.”
The term is not new; it was first discussed in the 19th century in George Boole’s The Laws of Thought. Jerome Pesenti explained it well in his 2014 TED Talk when he described a cognitive AI capability where the system uses natural language processing (NLP) to understand human speech and combines it with image recognition for reading lips to enable greater accuracy. This is an example of combining multiple AI technologies to enable machines to become more like humans — this is what cognitive computing means to me.
As we look forward to the next decade, how will AI and cognitive computing enable advancements in the area of augmented expertise of humans and machines (including robots) working better together? As a person who has been writing and building software systems for over 25 years, my favorite advancement is that humans no longer need to look at “dashboards” to interact with software. We freely talk to software now, as seen with Alexa and Siri. This is becoming so commonplace that most children know how to interact with AI software with speech today.
Cloud computing “big data” pipelines are becoming the norm, with technologies like Kafka, noSQL databases and serverless cloud computing that signal a shift to containers and microservices. This enables a more elastic and cost-effective cloud platform for AI and cognitive computing.
For example, an oracle in ancient times was someone who offered advice or prophecy, thought to come from some divine being. In the cloud AI world, software can easily gather many years of data for trends and patterns, analyze the state of and relationships between entities, and use a variety of predictive machine learning models to predict the future.
Time-series databases are now becoming a very popular infrastructure for this type of analysis. The intersection of this time-oriented data analysis, combined with a “human expert in the loop,” can enable a new era of cognitive computing to solve many tough problems we face in society. Just like with a child, a human expert can tell the software when it is correct and when it makes errors — so it learns from its mistakes and becomes smarter. In cognitive computing, the machine will use multiple data sources and the output of multiple layers of analysis to help solve a problem.
The next decade will bring rapid advances in the way we build software. Instead of a focus on logic/rule building (deterministic programming), we will see a focus on data tagging and data cleansing. In AI it is all about the data — good, clean data and lots of it. We are already seeing a shift in “offshore” contracting in software to a focus on AI data tagging/labeling and building up massive cloud infrastructures for data pipelines.
Technology like Kubernetes is advancing cloud computing — or as some call it, “cluster computing” — to enable DevOps to more easily and rapidly stand up and tear down cloud computing AI infrastructure.
