A Sunny Guide to Chatbot History
Let's start a blog, we said.
Let's write about all the cool chatbots we are making and best practises we are starting, we said.
Let's add new content every day for 90 days to give it a jump start, we said.
Let's go on holiday, the wife said.
Here's my first day in sunny Mauritius:
Source: my un-sunned iPhone
Yes, that is the beautiful blue sky you can see out of the window of the business centre.
Yes, here I am desperately not willing to end our blog kill-streak.
So, forgive me if my writing on chatbot history comes across a little thin. This beautiful fresh fish is not going to eat itself.
Hey ho, less dilly dally. The quicker I write, the quicker my toes are in the sand.
First up, let's define what a chatbot is
A chatbot is a computer program you interact with via text or voice.
It is a way of accessing a service or a machine's power using by talking or typing. We also wrote a non-technical chatbot definition, read it.
Although they now use the latest in natural language processing (NLP) and broader artificial intelligence (AI), chatbots are nothing new.
Want to grab the PDF version of this article with all the info, graphs, and cool anecdotes? Download for free here.
As far back as 1950, Alan Turing and Joseph Weizenbaum had the foresight to realise that one day, communicating with a machine would be indistinguishable from talking to a human. They developed something called the Turing Test, a way to assess the natural language conversation level of a machine. This test is administered via text conversation, to remove the reliance on natural human-looking hardware. More about the Turing Test in my article 'What is the Chatbot Turing Test?'.
The next significant milestone in chatbot history is the development of Eliza in 1966. Eliza is a computer program that humans talk to via typing and text responses. It simulates the language of a psychotherapist and still works well even now, particularly when you talk about yourself and your life. You can still talk to Eliza if you like.
The first move into voice chatbots occurred in 1981 with the Jabberwacky project led by Rollo Carpenter. This bot was first seen as a Sinclair ZX81 program (there I was, thinking that copy-typing fruit machine games from a book was smart). This chatbot continued to be developed and ended up winning the Loebner Prize multiple times (awarded for creating the most human-like chatbot).
In 2000, Rober Hoffer was lead co-creator of the chatbot called SmarterChild. It lived on AOL (wow, flashback) and MSN Messenger and, in its heyday, grew to have 30 million users. The chatbot talked about lots of topics, from news and weather to giving current movie times. It could even act as a basic personal assistant using NLP technology.
In 2010, Apple entered the fray with Siri. Launched via every iPhone handset, Siri quickly became the sassy personal pocket assistant. The only way to interact with it was through text and voice. We all have Apple and Siri to thank for bringing chatbots into mainstream consumer awareness - paving the way for all us mere-mortal companies developing chatbots now.
In this instance, Google was a couple of years behind Apple with their Google Now product. Launched in 2012, and commonly found in Google search, it used NLP to answer questions and serve relevant search results. Google realised the future of search might be from people just asking with speech; typing is so 1990's.
Microsoft is a world leader in NLP; their flagship is the Xiaolce chatbot. With over 40 million users, the chatbot launched in 2014 as a Weibo extension. Xiaolce is kind of like a free therapist, she 'understands' and remembers user's moods ready for the next time she speaks to them - making sure she uses the correct level of empathy.
Amazon's turn now with Alexa in 2015. Who does not remember that glossy, shiny American family in the advert talking to Alexa via the Amazon Echo? Some 4 million sales later, Alexa is basically a world leading chatbot, admittedly, with some super-advanced soft and hard tech.
Straight out of the Halo video game - at least that is how I picture it - Microsoft launched Cortana in 2015. Initially living in Bing search and later on embedded in your PC's operating system, Cortana is an intelligent personal assistant who can set reminders, read all your emails and see all that juicy browsing history.
(Hmm, or is that Google? Or both? idk. Well, at least chatbots cannot see or hear what you are doing at home...o wait, Alexa.)
2016, Zuckerberg starts swinging by opening up a Facebook Messenger API for developers. Basically, every geek sat in their pants got all hot and sweaty and started creating rubbish chatbots. Yes, this the reason now everyone thinks chatbots are crap. Thanks.
Microsoft did also get one wrong. Their TAY Twitter chatbot, launched in 2016, was shut down pronto. It was designed to learn from inbound Twitter messages to build a version of human nature and 'personality'. In a nutshell, Twitter-dwelling people (is there a name for them? tweeters? twitterers?) taught TAY to talk positively about genocide, racism and Hitler - yes, whoops.
More recently, enormous advances in broader artificial intelligence and the more relevant domain of natural language processing have given chatbots 'almost' human-like behaviour.
Humans can now talk to machines and, if the scope is narrow enough, probably not even realise a computer talking back. Commercially viable business models and services which are wholly chatbot-driven are emerging, and businesses of all sizes are scrambling to adopt the technology (we're here to help).
Business Insider recently released research that by 2020, 80% of all businesses will have a chatbot. Also, according to research from Juniper, chatbots could help lower the costs of doing business by more than $8bn (USD) by 2022.
So, now you have had a brief, and sunny, look chatbot history - and realised the world has sat up to take notice - is it time you made contact with us?
I think you should, and I would be delighted to speak with you about your business. Just let me get home from 'holiday' first, ok?