Don't Expect Chat Bots to Get Anywhere Anytime Soon
Despite the blind rush, we still have lots more research to do before we start developing really useful chat bots.
There's a new craze going on. If you've read around lately, you might have noticed that chat bots are all the rage. They are going to replace contact center agents. They are the new mobile app. Every business must have a chat bot.
I have to say, I can't understand this blind rush towards bots.
Bots of today are broken. They are like toddlers in a world of grownups -- they do funny things and are just more entertaining and cute than useful and efficient. To get a glimpse of what I mean, just watch this obvious prank someone decided to pull over Amazon Echo and Google Home:
Everyone and his dog is running to implement chat bots that solve our financial problems, book us vacations, go to court for us to dispute parking tickets, or purchase pizzas with our preferred toppings. But can we slow down a bit and see what is even available to us at this point in time?
We need four ingredients to build a useful chat bot:
1. Top-notch speech-to-text engine
- Necessary if we want to start with a natural conversation
- Needs to be able to autocorrect based on context
- Needs to support the languages we want to be able to handle
2. Natural language parsing and understanding
3. Ability to understand the concept of context
- Between messages in the same conversation
- Where exactly is the user?
- What did the user just do?
- Does the action relate to something on the user's calendar?
- Did the user ask me something in our previous interaction that I should add to my list of considerations?
- What day is it? Any specific holiday? Black Friday maybe?
4. Flexibility to learn automatically
- Our services and websites constantly change. Chat bots need to be able to take these changes into account
- Guess what? If everyone is having the same problem today, the chat bot might as well get to the point a lot faster than run me around in circles
Each one of these ingredients require hard-core analytics and artificial intelligence (AI) capabilities. And we're only scratching the surface of what this really entails.
Why do I know we're not ready for chat bots? By that video up there. The one where the Amazon and Google virtual assistants are each locked in an endless loop answering the same stupid calendar question and, by that, invoking the other to do the same. They lack the intelligence to break the loop and move on.
And yes. We can be sure that Google will fix this behavior in the next week or month, but that's going to be just another if/then/else thing in the code. This nice prank shows exactly what's missing in bots, that being real intelligence... the depth to understand meaning.
Chat bots can be great for menial tasks -- automating stuff that you can write down in a script with little to no deviation. But they're still not there once things begin to go off-script or become too general in nature.
Can you fill in that gap? Google is ahead of the pack when it comes to AI, but Google Now, Allo, and Google Home still have their own issues with understanding and context.
Despite the big shift towards AI and chat bots in the market today, I feel we're still in the early phases of the R part of R&D and nowhere near something we can call a product.
Who should be doing chat bots then?
- Large players that want to become dominant in AI. Think Google, Apple, Facebook, Microsoft, etc.
- Small startups trying to innovate and bring a new technology to market. Most probably, however, the best ones will be acquired before coming out with something really effective
- Developers looking at bots that are less about AI and more about automation. If you have these in mind, then by all means, invest in them
But that whole story of chat bots replacing apps? I don't buy that. Not in the short or midterm.