Chatbots are everywhere at the moment. It began as something of a novelty back when Apple introduced the world to Siri and allowed us to control our phones just by speaking to them. You needed to follow specific phrases, have no hint of an accent and be in silent surroundings, but now bots are gaining real momentum as we allow assistants into our homes in the form of Amazon’s Alexa and Google’s aptly named Assistant.
This latest evolution of assistants are far more powerful than their original equivalents. Their voice recognition accuracy is vastly improved, as is their ability to process different phrases that have the same intent. Without wanting to be disparaging however, these bots still aren’t really doing anything cool – and by cool I mean anything that we couldn’t do ourselves as humans. They can check the weather forecast for you or turn on your favourite radio station, but I can equally do that with my thumb. In a way they are like working dogs. They make your life easier with loyalty and obedience but if you ask them to solve any genuine problem such as the fastest way to deliver your birthday party invitations or which of your sales team can make that last minute meeting with a hot new lead, then all you’re going to get back is a blank look and a wagging tail.
All’s not lost, however. This isn’t a fundamental limitation of chatbots, just that their various functions such as sending messages to your friends or turning off the bedroom lights are the obvious first steps for bots which reside in your phone or a box in your living room.
At this point you may be thinking, “Satalia does optimisation, why are they talking, or even thinking about bots?”. Well, it just so happens that optimisation is one of those cool things that people can’t really do. In fact, it isn’t just difficult for humans to find optimum answers, but it can be equally difficult for people to describe the problem before the algorithms like those in Satalia’s Solve Engine can even get to work on solving it. If we can get a bot to perform this mathematical description for us then it will have transcended the loyal puppy into something fundamentally more useful for a much broader range of people. Optimisation for the masses!
Grand plan – but is that actually possible?! Well, yes and no. The Holy Grail in natural language for optimisation is a system that can receive a problem described as you would describe it to a friend, and then fundamentally understand the problem in order to start assigning objectives, variables, constants and constraints based on that problem’s underlying structure. This is the process which our optimisation experts at Satalia go through when developing our large consultancy projects, but sadly these superstars can’t yet be replaced by intelligent robots.
Fortunately though, once a problem has been defined mathematically, that structure can be reused. For example, trying to find out how to spend the most time possible at the top 10 landmarks in London and how to spend the least time possible on the icy roads when visiting your friends and family at Christmas time are modelled in exactly the same way, just with different numbers in place. That’s not necessarily to say it is easy to reuse these problem templates, but with some ingenuity and reliable code it is certainly possible. Basically, we can get our optimisation experts to do this modelling heavy lifting on behalf of the natural language pipeline which can, in turn, simply ask users for the relevant information in a conversational and easier to understand way. This works on exactly the same premise as your phone’s assistant filling in the relevant information to be able to send a message, but now it’s doing something really cool.
Discussions of the singularity and whether there will ever be such a thing as generalised artificial intelligence aside, if we build chatbots intelligently then we are already at a stage when they can augment our human ability to solve some of the world’s hardest problems.