AI, or Artificial Intelligence, will soon power most machines to various degrees. These machines could be software running on the cloud, or hardware in the form of home appliances, transport and medical devices. But there are levels of intelligence. Some machines will be truly intelligent whereas others will be merely 'smart'.
Your washing machine can be remote-controlled by a smartphone app. 200 years ago that would have been cutting edge AI but now that is mundanely routine. It would be a bit smarter if it had a material sensor that automatically set the washing cycle depending on the type of clothing you threw in. But it would be really intelligent if it could sift through information on the Net and determine that the washing powder you use is detrimental to the environment (without being programmed to do so), or suggest natural alternatives to the synthetic clothing you've been using (available at a 30% discount from www.say-no-to-plastic-clothing.com).
Machines are programmed by humans to achieve certain tasks. Smart machines hook up to the internet to continously upgrade themselves and adapt themselves better to their users. Intelligent machines are the ones that are able to learn from their users and environment and autonomously write new software (or sub-routines) that can execute new tasks or enhance the tasks they were originally programmed to carry out.
Artificial Intelligence can help streamline some ecommerce processes that would otherwise take too much of your time, or provide solutions that would normally not be available with current technology. Image recognition and voice recognition is already here and improving in leaps and bounds. Here are some more areas that AI can help automate:
1. Customer Service: Instead of waiting around to talk to a customer representative, an AI chat engine can interact with your customers 24/7 and help identify the purpose of their visit and find out more about the kind of products and services they are looking for (as opposed to you making assumptions limited by what you are trying to sell).
2. Information Architecture: An AI could propose an information structure for your site that will help you organise or classify your products in a way that customers can find them easily.
3. Automate Content Generation: Generate content for your site or app from a set of parameters provided by you.
4. Optimise Individual Salesflow Understand individual customers and generate a hyper-focussed customer experience that cuts out unnecessary ecommerce 'noise'.
5. Optimise Platform User Experience Analyse customer purchase patterns and then suggest general user experience or site design optimisations that can help increase sales.
All this can be achieved to varying degrees with current levels of technology. But take the first one - real-time customer service. An artificial intelligence that can converse with a human customer in real-time and make sense would have to be incredibly powerful. Natural language processing requires boatloads of cross-referenced data and some very powerful algorithms to make sense of the subtle nuances of human speech. Chatbots can mimic this process and look up keywords in an FAQ database and print out prepared statements with some results. That is not AI, that is hard-coded smartness.
An AI needs to learn and grow and create new tasks for itself. To empathise, have a sense of humor and understand human motive is a complicated task.
Lets see how a human mind works.
Rule #1: Ensure your survival.
Rule #2: Pursue pleasure, so long as it does not conflict with Rule #1.
These two rules drive everything we do. The human mind has devised various routines to execute these two prime directives and we constantly keep developing new sub-routines on a nearly daily basis to achieve these two objectives.
So we get jobs, grow food, get healthcare and make long term plans to ensure our survival. When this primary process is more or less under control we go pub-hopping or play football or go skydiving. Sometimes we get super smart and eat chocolate, thus executing both rules with one action.
The human mind keeps learning and keeps problem solving with amazing versatility. The human brain is one device but it can keep learning new skills. You perform a job, then go look up recipes for Thai curries on the net and then go and learn how to play a musical instrument. The human brain has evolved into quite a fascinating mechanism. You can go club an animal over the head and eat it afterward, or build a rocket that can carry your species to another solar system.
It is a complex decision-making engine that manipulates an ever-changing environment and tackles new challenges to address those two basic rules.
An AI, similarly, can be given a set of operating rules, a decision-making engine, a learning engine that stores and cross-references data and some sort of rewards mechanism that helps it manipulate data and complete a task through trial and error (good dog, bad dog).
Most computer applications have a hard-coded set of instructions that enable data storage and display (an excel sheet, or an information website) or manipulate data (online banking) or enable human interaction (Skype / social media / chat). There are armies of human programmers that keep upgrading those instructions on a periodic basis and develop new instructions that execute new tasks.
An artificial intelligence, however, needs to have a much more smarter architecture that helps it to operate autonomously. Autonomy for an AI means the ability to learn and improvise, which in machine terms would involve creating sub routines automatically when the situation demands it.
Besides decision-making an AI also needs to have a learning engine. This helps it to collate data about its working environment and use that data to arrive at more qualified decisions. You can tell an AI the difference between a cat and a dog and that they are both carnivorous. The AI then sees a zebra, identifies it as an 'animal' but then observes its eating habits and figures out that this one is an altogether new category of animal called 'herbivore'. Then it discovers cows and donkeys and horses and starts putting them all into this new category.
Now there are millions of animals and birds and trees and man-made objects in the world. It would be a crazy task to manually label all that data and store it in a database for an AI to use. May as well give this job to the AI and let it capture and classify data and knowledge about the physical world on its own. The AI would get this data from its interactions with humans, other AIs and 'sensory' inputs such as image and video.
In ecommerce terms this means that the AI figures out that a 'tshirt' belongs in the category of 'clothing' and a 'desk' belongs in 'furniture'. Manually creating this database would take eons but if the AI could do this on its own, that would be great. A learning engine like this would help the AI to learn about the physical world, about human customers, their geography, culture possibly, spending patterns and more. And this is where ecommerce automation can begin.
So these are some of the challenges lying ahead of us and we are beginning to tackle some of them. Want to join in? Here's a little bit more about how we are doing it all >