7 Key Differences Between Chatbots and Intelligent Personal Assistants

 - June 25, 2018

The use of Chatbots (CB) and Intelligent Personal Assistants (IPA) is exploding.

There are now over 300,000 CBs on Messenger – up  from 100,000 a year ago – and the IPA market is growing at 36.7% and is expected to reach $12.3B by 2024.

As the market expands so does the intensity of the discussion about the defining characteristics of each type of system.

There are a number of credible sources that claim that there are no material differences or that one is a subset of the other.  While it’s true that the lines are blurring and new subclasses are emerging, it isn’t accurate to say that they are the same thing.   There are key differences and it’s important to understand the distinctions.

 

This post discusses seven key differences and explains why they matter. Of course it’s possible to find CBs that act a lot like IPAs and IPAs that smell like CBs, but the descriptions in this post relate to the typical use of CBs and IPAs.

The remainder of this post looks at the differences as they exist today and focuses on differences that are likely to continue into the future regardless of changes in technology.

Seven Key Differences

Many of the articles that discuss differences focus on distinctions in terms of technology and capabilities, however, this often obscures the most important and clear distinctions:

 

1. Intent

The discussions that focus on technical differences often miss the single most clear distinction between CBs and IPAs: they play for different teams.

In general CBs are intended to serve companies and similar organizations.  The may well “serve the user” in the sense that they give the user important information or provide some service, but their purpose is to serve the company.  Any assistance they provide to the user is a byproduct of their efforts to achieve corporate objectives.

The roles are reversed for IPAs.  They exist to serve the user.

An IPA may well do something that benefits a company – for example, booking a table at a restaurant – but any service they provide to a company is a byproduct of their efforts to serve the user.  Note that IPAs such as Alexa may well have an overall objective of selling products but those sales occur indirectly as the system goes about serving the needs of the user.

 

2. Location of Execution

The second characteristic is related to the first.  The classes differ in terms of where they operate.

Most CBs are corporate based and as a result CB’s tend to execute in office environments.  IPAs are consumer based and as a result most IPAs operate in a home or on a personal device.

 

3. Scope and Domain

CBs by their nature tend to deal with specific types of problems within limited domains.  For example, a customer support CB will generally deal only with customer support problems and only with problems that relate to the subject company’s products and services.

IPAs operate within an environment that is much broader and more open-ended.  Alexa currently has 15,000 specific skills and they vary from helping users plan a vacation to finding out facts about the Indian city of Dehradun.

Some very credible sources claim that this distinction is a myth, pointing out that CBs are growing more capable every day.  This assertion fails to recognize how CBs and IPAs are typically used today.  Certainly CBs could be extended to serve in broader areas but the typical CB in the current world deals in a much more restricted domain.

 

4.  User Knowledge

The effectiveness of IPAs depends on having access to specific knowledge regarding the user.  CBs may also have some user knowledge but it tends to be much less extensive compared to the knowledge maintained by IPAs.  For example, a typical CB may well know which of the company’s products the user owns whereas an IPA may have access to the user’s full contact list, their gmail history, and many other bits of information

 

5.  Persistence

IPAs tend to have a persistent relationship with their users.  CBs model service representatives who are picking up whatever call comes in whereas IPAs are in a long-term relationship with the user they serve

 

6.  Interface

Most CBs operate using message systems.  Many people think of CBs as running on services such as Facebook Messenger and Slack but the fact is that Line, Skype, WeChat, Spark, Telegram, and Viber also provide APIs for building CBs.

IPAs tend to have much richer interfaces using voice, text, images, and video to communicate with their users.

 

7.  Development Process

The last point of contrast relates to the process for developing each type of system.

It’s possible to go from 100,000 to 300,000 CBs in one year because it is a relatively straightforward process to develop a simple CB.  Development platforms such as Dialogflow allow developers to build a basic chatbot in a matter of hours.  A number of companies provide tools that they claim will allow users to develop CBs without even using code.  There are also a number of platforms for building CBs and many great companies that build CBs on behalf of customers.

Building an IPA is much more complicated and takes much more time.  The major players have invested hundreds of employees and many years in the development of their IPAs.

 

It is certain that CBs and IPAs will continue to evolve and accelerate into the future.  It will be interesting to see what develops and to see how the terminology changes as the systems evolve.

 

 

, ,
In Alexa, Chatbot, Intelligent Assistant, Uncategorized

Turing the Far Side of the Valley

 - June 10, 2018

The last couple of years have been very exciting when it comes to discussions of the uncanny valley.

The uncanny valley has been a topic of great discussion since the term was coined in 1970.  It has been described simply as “our strange revulsion toward things that appear nearly human, but not quite right.”

There has been a great deal of formal debate about why it occurs, how it affects different people, and about whether it actually exists. Regardless of that debate, it has been a very real topic for artists and visual developers for decades and has had an enormous influence on the visual representation of humans.

The Valley is to the visual word somewhat like the Turing test is to the world of AI.  It has stood as a great barrier that brought fear to the hearts of the creators who ventured too close to the edge.

 

Many Uncanny Valleys

Although the term is singular the reality is that there are many different valleys, or at least that the concept of the valley can appear in many different forms.  The original use of the term was in regard to robotics and it is still a hot topic in that arena.  The other prominent use of the term is in regard to visual humans – avatars, game characters, and CGI characters in film.

 

It is less common to hear about the valley as it relates to voice and audio but it still a very real and recognized effect.  Some have even suggested that the creepiness might be useful in works of horror.

 

Lastly, there are even cases where we see uncanny effects in real humans who have been altered to look like specific dolls and other characters.   For example, there are several people who have gone through extensive plastic surgeries to make them appear like the famous “Ken and Barbie” dolls.  For example, this image shows Rodrigo Alves who paid more than $500,000  with the goal of looking like the Ken doll.

 

The Creepy Valley

There is a lot of debate about the Valley but one thing is certain.  There is one word that is welded to the Valley.  That word is creepy.  Note that it is used in almost all articles that discuss the valley.  It has been used for decades to describe those unfortunate enough to reside there.  There are other words for sure.  Errie. Strange.  Bizzare.  But the word that comes back time after time is creepy.

 

All of which brings up a question:  what is the word the best describes the far side of the valley?

 

The Far Side of the Valley

 

There’s been an enormous amount of discussion about Google’s now-famous demo of a dead-on human voice calling to make a reservation at a hair salon.

 

Quite a bit of that discussion revolved around the question as to whether Google had demonstrated an AI application that actually passed the Turing test.  The chairman of Google’s parent company claims that it did, at least in a limited domain, but others say that it didn’t.  It may or may not have passed the Turing test but one thing is for sure:  it made the leap to the far side of the uncanny valley.  The voice and associated mannerisms were absolutely real; so real, in fact, that many writers are suggesting that it was faked.

 

It’s interesting to note how people reacted to hearing a voice from the far side.  There were some references to it being creepy, but that wasn’t the dominate reaction.  The dominate word for people’s reaction to this example was “scary,” or some variant of that word.  In fact, quite a few refered to it as horrifying.

 

There are a number of examples of things that are appearing on the far side of the valley. There are now super-realistic robots and avatars.  As people begin to become aware of the existence of these creations their initial reaction is often amazement – but over time the reaction that settles in is fear.

 

Many articles are appearing that express real concern about a future in which there are artificial people – in many different forms – that are indistinguishable from humans.   Of course much of this has been standard fodder for countless science fiction stories in the past.  A space alien lands on earth.  It is scary if he looks like some bizarre 12 headed hydra, and even scarier if he looks like The Terminator;  but what’s really horrifying is if he looks identical to a trusted friend and neighbor.

 

All of which seems to imply that the “standard valley claim”  may have been wrong all along.  For many years I – and many others – have been giving a standard pitch in which we describe the valley and then make the claim that, ‘It’s okay to be on the far side and produce characters that look fully human, and it is okay to be on the near side and produce characters that are clearly not human, but it is not okay to be in the valley between.”

 

But is it actually “okay” to be on the far side?

 

For many years when discussing the Valley we’ve warned people: “Thar be dragons there;”  but it now appears that the far side of the valley has it’s own dragons and they may be bigger and scarier than those in the valley.  It seems to be fine to produce fully realistic CGI characters in a film and perhaps even in a game.  The real fear arrises when the characters escape from the screen and begin to live among us.

 

We are entering an era where that scenario has the prospect of being real, albeit as a human-created character rather than an alien.   The google voice is horrifying.  Deep Fakes produces videos that are dead-on realistic fake videos. Projects like PAI are striving to produce highly realistic synthetic versions of people – including dead people.

 

These developments are fascinating but not necessarily benign.  It is probably time that as a society we begin to develop mechanisms that help us feel a bit more secure as we experience life on the far side of the valley.

 

 

 

 

 

 

 

 

 

 

,
In Digital Human, Humanoid Robot, Uncategorized

Teaching A New Dog Old Tricks

 - June 5, 2018

@Engaget recently published a very interesting article that discusses how Alexa is taking a step forward in its ability to actually do what a user is asking to do.

 

The most interesting part of the article is the first sentence:

 

Amazon is training its voice assistant to be able to tell which skill will most suit your needs if you have no idea which one to summon.

 

This and many other articles have described this as Alexa learning a “new trick.”  In reality it’s an old trick that this new dog is struggling to learn;  namely, the “trick” of simply understanding what someone is asking to do and doing it.

 

This is an example of an Assistant taking steps to begin to do what many people have always imagined they already did.  The truth is that this one sentence sums up one of the huge challenges that face developers of Voice Assistants and Intelligent Personal Assistants.

 

The sad fact is that although Alexa is a great system she really isn’t that intelligent.  The sadder fact is that Intelligent Assistants in general aren’t actually that intelligent.   They don’t really understand much about what the user is requesting.

 

IPA’s are conceptually simple.  They listen to an input, extract the intent and the objects of the intent, and then pass off to one of many “action execution modules” called skills that perform the requested function.

 

Fail, Funny, Stupid, Dumb, CrazyThat sounds simple but it’s actually extremely hard, mainly because really understanding free-flowing natural language is still over the horizon.  Real people use language that is vague, conversational, and heavily context-oriented. As a result one of the hardest problems in the world of IPA development is determining the intent.

 

The next hard problem after intent is deciding which skill to call to perform the intent. Alexa currently has more than 40,000 skills, many of which are not directly created or controlled by Amazon. Many of these skills have vague and overlapping capabilities.

 

The historic resolution of this problem has been to tacitly acknowledge that the new dog can’t actually learn old, basic human tricks and to pass the problem off to the user. Alexa (and other IPAs) can’t understand human language so she turns the situation around and insists that humans learn her language.

 

In the early days of Alexa the system was basically simply a voice interface that activated apps.  To make that work the user needed to know the name of the app that was to be kicked off, in much the same sense that a user who wants to hear No Tears Left to Cry needs to provide the name of the song.

 

Alexa advanced from kicking off specific apps to kicking off more general skills but note that the first line of the @Engaget article presents Alexa not as an assistant who understands people and performs tasks but rather as a mechanisms for the user to “summon” one of 40,000 mini assistants that perform specific tasks.  The user still needs to know the name of the skill or associated key words to get a reasonable result. It becomes harder and harder for the user to hold up their end as the number of skills explodes.

 

Amazon’s new feature, CanFulfillIntentRequest, will make real progress by going out to those 40,000 skills in an attempt to find one that can fulfill the request without the user having to be aware of the skills existence, capability, and name.

 

That’s a good thing but what we really need is for all the new dogs to learn the old tricks that we all expect them to have.  They need to understand real conversation.  They need to operate in context.  They need to deal with input that is vague, inaccurate, and changing.

 

That’s all a lot easier said than done.  It will be exciting to see where IPAs go in the future as various projects begin to work these fundamental problems.

 

 

 

 

 

, , , , ,
In Alexa, Fundamental challenges, Intelligent Assistant

Introduction to Fakefokes Blog

 - May 24, 2018

Hi.

My name is Dave Rolston and I’m the primary author for FakeFokes.

Thanks for visiting.

This post provides some information about my background, describes the intent and goals of the blog, and encourages your involvement in the future.

 

My Background:

I attended my first conference on Artificial Intelligence in San Francisco in the late 1960s when I was still in high school.  I was bitten by the AI bug at that conference and have been passionate about it ever since.

I eventually completed a PhD with an emphasis in AI and in 1988 published Principles of Artificial Intelligence and Expert Systems Development, one of the first best-selling books on AI.

I became the youngest fellow at Honeywell, Inc. and the first to be named in the AI domain.  In that and other roles I worked on several early research projects in natural language understanding, machine learning, robotics as well as other areas of AI.

I’ve been involved in AI programs ever since  but also developed a passion for 3D graphics, especially as it relates to creating avatars that reflect human emotions as well as other forms of “artificial people.”

The first half of my career was purely technical but along the way I discovered another passion-  a love  for creating businesses that apply technology in ways that helps people live richer and more interesting lives.

Over the last 20 years I’ve been CEO of four small tech companies and GM or three corporate divisions.  Most of my work during that time revolved around AI, 3D graphics, virtual and augmented reality, and robotics.

I’m currently CEO of Tirocorp where we’re working on a comprehensive intelligent personal assistant.

I lived in the Silicon Valley for about 30 years but these days I live near Austin, although I still spend much of my time in California.

 

You can check me out on Linkedin at: Dave Rolston

The following sections answer a few questions about the blog.

 

Why am I blogging?

I’ve written several blogs in the past.  I keep at it because I enjoy it.

I enjoy the opportunity to connect with people who have shared interests and I enjoy learning things from my interactions with readers.

I think of blogging as a deal between a writer and a group of readers.

My end of the deal is to work to provide information that is both interesting and useful, recognizing that different readers will have different views on what that means. My sense is that some readers will simply be looking for updates that keep them abreast of what’s happening in the target industries, those who are actually developing systems may be looking for pointers, and others will be looking for information that will help them understand how the target technology relates to business operations and profitability.

What I hope to receive from readers are comments, reactions, and input that will help me better understand each of these areas as well.

 

Which subject areas will be covered by this blog?

This blog focuses on artificial people, where that term is intended to describe mechanisms that simulate human forms or behaviors in one way or another.

We will be discussing several specific types of these synthetic mechanisms:

 

Chatbots / Intelligent Personal Assistants

Many different AI-based systems have been created that  strive for a human-level ability to engage in intelligent conversations.  Many of those systems are intended to assist their human users in one form or another.

One category of such systems, chatbots, includes assistants that are generally intended to help users with questions and requests that are related to a specific business topic.  Another category, intelligent personal assistants, include assistants that are usually targeted to provide information and help in a more general environment.

One key trend in this area is that the historical distinctions between these categories are beginning to fade.  That’s one of the trends that we will be discussing in this blog.

 

Digital humans / avatars

There are a number of different programs that are attempting to build graphic avatars that accurately represent human appearance, including detailed facial expressions and body movements.  These highly accurate avatars will be used to represent people in many different environments including social virtual reality, virtual worlds, etc.

Other programs are going beyond appearances and are creating digital human models the reflect not only human appearance and movement but also capture detailed information about the human systems.

 

Humanoid Robots

The ultimate goal for many synthetic human programs is to create a humanoid robot that is a physical entity that puts all the pieces together.  These programs envision a day when they can create robots that look like a human, move like a human, talk like a human, and think like a human.  Some of those programs are getting pretty close to doing just that.

 

Common Themes and Issues

Developers in each of these areas are dealing with very specific issues but there are also a number of themes and issues that cut across all the different categories.

For example, one such issue is the “uncanny valley” – the idea that entities that look almost human, but not exactly human, are creepy and scary and should be avoided.  The concept was first used in 1970 in regard to robots.

In reality there are now several different forms of the uncanny valley.  Almost-human avatars are still creepy but so are almost-human voices in chatbots, almost-human reasoning in digital assistants, and almost-human forms in robots

This blog will be looking at a number of these cross-discipline issues.

 

Who is the intended reader?

Of course I encourage anyone to read the blog and interact, I’m writing with three types of readers in mind:

  1. General interest – those who are simply interested in the general topic and read for personal interest or to remain up-    to-date on the topic in general
  2. Technical developers – those who are building artificial people
  3. Business people – those who are interested in learning more about how these capabilities can be applied in a business setting.

 

What kinds of content will be included?

The plan to post about twice a week.  The blog posts will provide information and comments that are oriented to the groups noted above:

  1. News and general developments– posts that focus on news items and developments  in the target industries.
  2. Technical information – posts that will be useful to developers who are working to create synthetic humans and to business managers that are looking to engage with appropriate development resources.
  3. Business applications – business-oriented posts that provide comments and information about how the target technologies can best be used in create business success.

 

What are the goals of this blog?

The long-term goal of the blog is to serve as a catalyst to help develop a community of people who are interested in this area and to give the members of that community a forum for exchanging ideas.

I have the following specific goals for the first year of this blog:

  1. 100 posts
  2. 500 regular readers
  3. 100 reader comments

I’d love to get your help in achieving these goals.

 

How can I get involved?

I welcome interaction with anyone who is interested in this area and I encourage comments on every post.  I try to respond to every comment within a day.

I also encourage suggestions for post topics and suggestions for guest bloggers.

Please feel free to email me at david.rolston@fakefokes.com.

I look forward to exploring this area together.

 

Dave Rolston

 

 

 

In Introduction, Uncategorized