One of the things I’m looking forward to seeing at next week’s NVIDIA GTC event is an update on their Conversational AI efforts.
I’m fascinated with this because I was a Competitive Analyst for ROLM Systems and part of my responsibility was IVR (Interactive Voice Response Units). IVRs were the predecessor to the bots that constantly call and annoy us trying to sell things we don’t need, like extended automobile warranties.
The goal then and now is to create something that can interact with humans with voice that is indistinguishable from another human. At an earlier GTC, back when they did these events in person, I hosted a panel and told the story of how an advanced IVR system based on IBM’s Watson was so good in practice that one caller started trying to hit on it.
This example showcases both the problem and power of an excellent conversational AI. The problem is it will undoubtedly be used for spam calls. Still, the promise is we will get a far better experience interacting with companies that currently use the old, outdated IVR technology we’ve come to hate.
There is a more significant change coming through and highlighted in the Huawei analyst call this week, and that is that this technology could spell the end of the Smartphone.
The Coming Conversational AI Revolution
Since the beginning of the Computer Age, we’ve primarily been interacting with our creation backward. Instead of teaching computers how to best interact with us, we’ve created systems that forced us to learn how to interact with them. Even when voice is used, most systems only recognize command words they have been programmed to accept again, forcing us to comply in a way defined by our creation.
Well, Conversational AI fixes this mistake. While initial work is still pretty generic, eventually, these machines will be able to focus on what makes us individually unique and alter their conversational behavior to address that uniqueness better.
This change will be massively disruptive because people who have no computer skills will increasingly compete with those who use computers. That could make a whole crapload of existing computers and IT people obsolete.
Think of Data Scientists; they often perform the role of translating what line executives want to be done into computer systems to accomplish that goal. But suppose computers can directly understand and implement what the executive wants, given Data Scientists generally don’t excel interpersonally.
In that case, the result should quickly be better than what those executives get now. We do have a considerable data scientist shortage, and clearly, there will remain other tasks this skill set can do, but this alone should change how you staff up in the future.
Outside sales jobs will increasingly be better done by these conversational AIs, which will know a great deal about the sales prospect and will be able to craft customized offers at scale without charging a commission uniquely. Call Center personnel would also be on the shortlist to be replaced by these systems that will never get angry, never be tired, and be less likely to drive away a client.
High-profile authors may no longer need ghostwriters. They’ll give the conversational AI an outline for the story, set parameters tied to who the audience is, and write the column, article, or book taking into account the uniqueness of the targeted audience better than most full authors do today.
But this same capability can, and likely will, be used to spam out sales calls, attempt to defraud us at scale, and more effectively phish for otherwise secure information. These systems will know a great deal about us and use that information to convince us we know them far better than a person could today and learn from other systems what kind of arguments will be more effective in getting our approval.
We will undoubtedly need a new kind of AI defense for this risk, much like looking at tools to deal with Deep Fakes. When you combine those two technologies, things get scary or exciting. This same technology could allow us to talk to a believable simulation of people that have passed on. I know I’d love to be able to talk to some of my grandparents again.
Conversational AI has great promise and represents significant risks. Like any technology, it can be used for good or ill.
For good, it could create deeper relationships with our customers, close more sales with a massively lower cost of sales (computers don’t yet play golf, get paid commission, or have substantial expense accounts). For evil, these things would make incredibly effective phishing platforms and find ways to cheat people globally.
And, as I noted, this technology could also connect us to long-dead relatives or even the founders of your company or country. For instance, we often talk about what the founders wanted; this technology layered on top of the information we have on those founders could reliably give us those answers. Imagine a realistic and vetted simulation of George Washington laying into some of the folks currently in congress.
In the end, conversational AI may represent the most significant interface change since the beginning of the computer age. We aren’t nearly ready.