Chatbots

 

  1. What are chat­bots?
  2. How do chat­bots work?
  3. Use cases

 

 

What are chatbots?

 

Chat­bots are com­pu­ter pro­grams that use NLP to inter­acts with users in a human-like fashion. Inter­ac­tions usual­ly take place in a web inter­face or a chat pro­gram such as Face­book Mes­sen­ger. Some chat­bot sys­tems run in pro­grams such as slack or by SMS. The aim of a chat­bot is to mimic the respon­ses and reac­tions of a human being.

The term “chatbot2 ori­gi­na­tes from Chat­ter­Bots, TinyMuds, and the Turing Test: Ent­e­ring the Loeb­ner Pri­ze Com­pe­ti­ti­on, which was published b Micha­el L. Maul­din in 1995. It is a dimi­nu­ti­ve form of the word “chat­ter-bots”.

Chat­bots have exis­ted in some form sin­ce the 1940s. Howe­ver, it has only been in the last two deca­des that they have grown in popu­la­ri­ty.

One of the first milestones in chat­bot deli­very was the publi­ca­ti­on in 1950 of Com­pu­ting Machine­ry and Intel­li­gence by Alan Turing. The ide­as dis­cus­sed wit­hin the papers for­med the foun­da­ti­on of much of what we under­stand and how we think today about chat­bots. Cen­tral to the paper is the now-famous Turing Test, which still remains the go-test assess­ment most know for AI.

The Turing Test is whe­re a human sits down and types ans­wers into a com­pu­ter. The ques­ti­ons are ans­we­red. If a com­pu­ter ans­wers the ques­ti­on and the human can­not tell that they are speaking to a com­pu­ter, then the machi­ne has pas­sed the test.

It took four­te­en years for a pro­gram to pass a ver­si­on of the Turing Test. That was in 1964 when Joseph Wei­zen­baum began working on ELIZA at MIT’s Arti­fi­ci­al Intel­li­gence Labo­ra­to­ry. ELIZA was released in 1966 and was writ­ten in Weizenbaum’s MAD-Slip pro­gramming lan­guage. In 1974, Wei­zen­baum published Com­pu­ter Power and Human Rea­son: From Judgment to Cal­cu­la­ti­on, in which he expo­un­ded on his thoughts about arti­fi­ci­al intel­li­gence.

Ano­t­her com­pu­ter pro­gram around that time that achie­ved some pro­mi­nence was Ken­neth Colby’s PARRY, which was published in 1972. PARRY was desi­gned to mimic the actions of a human with para­no­id schi­zo­phre­nia. In its year of publi­ca­ti­on, PARRY took part in a demons­tra­ti­on at the Inter­na­tio­nal Con­fe­rence on Com­pu­ter Com­mu­ni­ca­ti­ons whe­re it con­ver­sed with ELIZA. Even­tual­ly, PARRY would pass a ver­si­on of The Turing Test.

Later chat­bots to have pas­sed The Turing Test inclu­de 1989’s Jabber­wa­cky, which attemp­ted to incorpo­ra­te machi­ne learning, and 1995’s ALICE.

How do chatbots work?

 

Ear­ly chat­bots reli­ed on very-basic struc­tures in order to simu­la­te a chat and con­se­quent­ly were neit­her flu­id nor able to accu­ra­te­ly simu­la­te a human inter­ac­tion.

Modern chat­bots, howe­ver, rely on arti­fi­ci­al intel­li­gence and machi­ne learning. This means that the bots can take what is being of them, extract from it the most-sali­ent infor­ma­ti­on, and search wit­hin an organisation’s data and archi­ves to find the most-rele­vant ans­wer.

Modern chat­bots often rely on inter­faces such as Face­book Mes­sen­ger, Slack, or SMS.

Use cases

 

Good for rou­ti­ne ques­ti­ons, mea­ning that HR can bet­ter deploy resour­ces to more-com­plex que­ries.

Chat­bots are wide­ly used in retail and other custo­mer ser­vice roles, ans­we­ring rou­ti­ne que­ries and pro­vi­ding ans­wers. They are also used wit­hin lar­ge orga­ni­sa­ti­ons to help employees navi­ga­te HR and cor­po­ra­te poli­ci­es.

Recent rese­arch found that while 37 per cent of respondents were hap­py to use chat­bots for get­ting quick ans­wers in an emer­gen­cy, 35 per cent said that they would use a chat­bot to resol­ve a com­p­laint or pro­blem, or to get detail­ed ans­wers or explana­ti­ons. Among tho­se sur­vey­ed, 35 per cent belie­ved chat­bots could ans­wer com­plex ques­ti­ons, with 37 per cent thin­king that they could give detail­ed, expert ans­wers.

The­re are nume­rous bene­fits for chat­bot imple­men­ta­ti­on. The Har­vard Busi­ness Review found that auto­ma­ting ans­wers could be vital to retai­ning busi­ness: it found that if the­re was a five-minu­te delay in respon­ding after a sales lead reached out, the odds of con­ver­ting that lead decrea­sed ten­fold.

Deloit­te esti­ma­te cost reduc­tions of 15–90 per cent with the intro­duc­tion of chat­bots.

 

Refe­ren­ces & PDF:

  • https://jorin.me/chatbots.pdf
  • https://www2.deloitte.com/content/dam/Deloitte/nl/Documents/deloitte-analytics/deloitte-nl-chatbots-moving-beyond-the-hype.pdf
  • https://intelligent-information.blog/wp-content/uploads/2017/09/A-Primer-AI-and-Chatbots-in-Technical-Communication.pdf