In Context

The next-generation communications blog from Nexmo

< Back

Bots and AI: The Current State of Technical Maturity

August 24, 2017 Published by

Bots have come a long way since the days of using chatbots in your IRC channel or instant messenger. But just how far have they come from a tech standpoint? Yes, every company and business media outlet is talking about them, but what are the real bits and what’s just hype? To help me separate the reality from the marketing, I spoke with CEO/Co-Founder of The Bot Platform Syd Lawrence and Microsoft Technical Evangelist Martin Beeby about the current state of bots and AI.

Watch the video of our conversation here, or scroll below the video to read the full transcript.

At what stage of technical maturity are bots and AI? (Full Transcript)

Sam Machin (Nexmo Developer Advocate & Alexa Champion): So one thing to think about is the current whole state of the union—what we mean by bots and AI and all these things. Lots of talk about it, but one thing I was thinking is where are we with the tech in terms of maturity? If we compare it to the mobile PDA space, are we at the early PalmPilot days, or the calculator watch days, or the iPhone days of it? Where is it?

Syd Lawrence (CEO/Co-Founder of The Bot Platform): I reckon we’re probably at the calculator watch stage. So while these things have been around for many years, not necessarily in the mainstream space, which is the biggest difference. Yeah, I reckon it’s probably calculator watch stage for the majority of people.

So there’s obviously been a lot of press over the last 12 months about these things being far better than they actually are. Thankfully, I think people are starting to understand more so what they are now. I mean, we’ve been building bots on IRC many years ago. And now it’s just that it’s on mainstream platforms. And they’re not really that different to what they used to be. It’s just that they’re on mainstream platforms.

we’ve been building bots on IRC [for] many years … now it’s just that it’s on mainstream platforms.

Martin Beeby (Technical Evangelist at Microsoft): I think that there is a slight bit of jump in terms of what they’re capable of compared to previously. And that’s the trying to figure out what a person is actually saying to a system. I mean, most systems, most web applications, most mobile applications have always driven people down a very specific menu-driven kind of way of operating an application. And I think with chatbots and with voice bots, we’re getting to a point where the user is able to explicitly ask for what they want. And we’re kind of getting to the point where we understand what the user is saying. We’re able to fulfill certain needs that user has at that point. And I think that’s where the real advancement is gonna come in with bots is when we start to understand more and more of what the user’s real intent is.

that’s where the real advancement is gonna come in with bots is when we start to understand more and more of what the user’s real intent is

At the moment, it’s a bit of a trick. We’re able to understand the intent relatively well with the cognitive services and various different artificial intelligence systems that most major companies are building. But I think over the next two or three years, we’ll start to see leaps and bounds in that area where we really truly understand what the user’s saying from freeform English rather than having to drive a user down a very specific menu structure.

Sam: Yeah, and I think that’s a good point. I think for me the big leap is the difference between something being command-line driven. So I mean, we’re all geeks; we’re used to typing in commands. And really, the first generation of stuff was a bot that sat in your IRC channel or your Instant Messenger. But fundamentally you were typing in a command as a string and you had to get it in a specific order, in a specific syntax and that kind of thing. And I guess the tipping point that starts to make it maybe a bit more AI is this idea that you haven’t got to structure your request. It can figure out from the keywords and things in what you’re asking for, isn’t it? But it’s still definitely early days, I think. We probably all agree to on that.

Martin: Well, a lot of the work that I’ve been doing recently, we build a lot of chatbots. But most of the customers which we’re speaking to want to create some kind of voice system or voice control on top of that chatbot as well as just creating plain chatbots. And so speech to text kind of systems I think are really at the sort of 2003 stage of touch. If I look at the touch metaphor, of how touch phones were in 2003, 2004 when they were all just capacitive screens and they were totally rubbish. And that you would press on them, and you would hammer different sections but you never really got the thing that you wanted.

And then all of a sudden the iPhone came out in 2007 and touch interfaces were forever changed. We’re in that 2004 moment with speech detection and speech intent, where things are kind of hit and miss. And yes, it sort of works but it sort of doesn’t as well. And it’s really frustrating and most users avoid it. But we’re going to get that pivotal point when speech and intent are fully understood by computer systems. And then we get to this magical moment where we’re then far more accepting of the technology. But at the moment, I think for the vast majority of the people they can be quite frustrating, these AI systems and these speech systems.

we’re going to get that pivotal point when speech and intent are fully understood by computer systems.

Sam: Yeah. I think it’s quite interesting. Just a really good example actually, the whole touch interface. You have the stylus, and it was really just a replacement for a mouse interface, which wasn’t the same thing. And if you were careful you could drive it. But for this stuff to become user-friendly, it’s got to be easy to do while multitasking—while you’re trying to get the shopping out of the car with one hand and then hold the door with the other and then you want to be able to shout at it. It’s no good recognizing you if I’m in nice, quiet room with a microphone and I walk up to the device and ask it in calm, plain English. It’s when it can work in that real-world scenario.

for this stuff to become user-friendly, it’s got to be easy to do while multitasking

Martin: Yeah, one of the big limitations at the moment of chatbots is this onboarding experience. How do you explain to the user what they can say to the bot, what sentences they need to do, how they structure things? And that onboarding experience is really painful. And it’s also because the bot doesn’t really understand free-form English. It doesn’t understand anything that the user could say to it. Only, generally speaking, it only understands a subset of English which it’s been programmed to. And I think to get these things really successful, we’re gonna have to get past that.

[Editor’s Note: Watch the full one-hour discussion on the state of AI bot technology.]

Tags: , , ,

Categorised in: , , ,

This post was written by