Yes, the robots are coming. This eventuality is met with guarded excitement as everyone begins to understand a life where artificial intelligence is interwoven into aspects of our lives, subtlely at first, but proving its invaluable significance in this new world.
We are clearly witnessing traditional models forced to acquiesce to emerging solutions – that are faster, more accurate, and able to remember rules and apply them to each successive task. The face of the new applications goes beyond functional. They are in a constant state of on-the-job training, failing AND learning as new data points are introduced.
We were pleased to interview with Dennis Mortensen, Founder and CEO of x.ai, a personal assistant in the form of artificial intelligence, which schedules meetings for you. You may know these assistants as Amy or Andrew Ingram.
In this episode I spoke to Dennis about:
- The rise of cognitive computing and artificial intelligence, threatening traditional business models.
- x.ai: how it all began –> with the pain of meeting scheduling.
- The reality: AI will deliver net job growth, NOT losses.
- The importance of understanding language and being singularly focused.
- The vision for x.ai in the next 5 years?
You can listen to the podcast here or on Libsyn
Continuous disruption is compounded by a clear shift to solutions that mimic human functions
I heard this from two guys at Facebook the other week: Where mobile applications were all the rage less than a decade ago, the world is overrun by a glut of applications. With over 2 million mobile apps in market and over 100 billion downloads from the Apple App Store alone, users clearly do NOT require more apps.
The shift is to now develop bot technology to enhance the experience among existing applications.
Where we’ve witnessed Microsoft’s chatbot, Tay steal a page out of the Trump playbook and run amuck with with her racist tweets, the more placid versions of Siri and Cortana have clearly not being leveraged to their fullest potential. Dennis believes there is a generational “difference” in how we use AI. GenZ are heartily embracing voice interface, compared to the rest of us who prefer to type. It’s probably the reason why my kids’ spelling leaves a lot to be desired… but I digress.
The abrupt changes we experience today will be considered normal in the years to come. Dennis notes that people living in today’s world do not see the full impact of this transition,
Perhaps the Siris and Cortanas aren’t supposed to be the Oracles that have the answers to any and every question that we have with the ability to do any job I want done. Perhaps they are just horizontal AIs that serve as a memory or enabler… So when I want a meeting scheduled, I’ll have Siri reach out to another AI like Amy Ingram (an agent who is specifically good at one thing) to complete the task. Siri, in this regard becomes the assisting agent reaching out to an army of other agents. Each of those agents will perform plethora of tasks that are today, human tasks.
We will end up with an intelligent agent marketplace and x.ai will be one of them.
Meeting Scheduling: The “necessary evil”
Consider this: Every one has a to do list. There are tasks that may not be worthy of your time. Setting up meetings is one of them.
Setting up a meeting is easy if everyone has agreed to a time and location on the first email. That is rarely the case. x.ai realizes the “annoying ping pong” of emails generated in setting up meetings. Imagine if this one task were removed…
When we initially set up a time for this podcast, I was introduced to Andrew Ingram (the male scheduling assistant). I was actually fooled into thinking that Andrew was a human. When he added me on LinkedIn, I initially believed he was a Principle at x.ai. He understood what I said (and yes, I threw him a few curve balls), and within a few emails we had the schedule set for the interview.
Andrew passed my Turing Test. I keep referring to Andrew as “he”. It’s clear to me I don’t perceive him as anything other than human.
People don’t want to schedule meetings. Period.
Dennis contends that in 2012 alone, he had set up over 1200 meetings. Many companies have been trying to invent applications to improve the scheduling headache through web services or plugins like Doodle. However, the market did not want the “help” – they wanted to do without it completely.
Today, human personal assistants do this already. The solution was to replicate the value of a personal assistant, “who does this very important task that disappears from your to do list”.
The Science around x.ai: Dealing with Human Ambiguity
Success for x.ai means that it becomes increasingly important to understand language, especially the context within the email exchange. While there is a clear objective, the assistant needs to effectively navigate the conversation and have the creativity to accomplish that objective.
In contrast to building an application, x.ai is an invisible software that exists in dialogue only. Hiring Amy comes with a level of trust. As Dennis points out,
YOU need to hand over the job of setting up a meeting to US and leave us be. We come back and tell you that we have solved for it.
Natural Language (NLP) continues to evolve. The more dynamics in discussion presented to Amy or Andrew Ingram, the more they learn the ramifications of potential failure.
We continue to fight this near 100% accuracy game where we cannot have anything go wrong otherwise meetings just don’t happen.
For Amy and Andrew Ingram, doing their job well means effectively dealing with human ambiguity and intention in language. e.g. Someone sends an email at 2AM to Andrew asking to push a scheduled meeting tomorrow back by 1/2 hour. For humans, tomorrow means “after I’ve slept”. So receiving an email at 2AM makes sense. However, for a machine, 2AM is technically the next day so asking to schedule “tomorrow’s meeting” is incorrect. The machine must learn this type of “ambiguity,” which, according to Dennis, is “all over our system”.
x.ai picked a very specific vertical. It’s important to master one task…
We are fortunate that we are so specific and well defined. This means: 1) Initial Request 2) Negotiation 3) Insert into my calendar.
It is the only way to do it because we effectively reduce variability and variety.
AI and the Impending Doom of Mass Unemployment
Dennis recently published this article on LinkedIn: AI will Deliver Net Job Growth by the Next Presidential Election
He challenged the notion of job loss as a result of automation. While automation may assist in improving production and speed to market, 0% of full jobs could be fully automated. The World Economic Forum (WEF) suggested that, in fact, the impact was far less than expected. As Dennis points out,
The gist: AI is not the hydra-headed beast that demolishes jobs, as it tramples through the global economy. Nor is it akin to a targeted assault on white collar workers—finally making even educated workers irrelevant.
The WEF attributes a small negative impact on employment growth to AI (about -1.6%) between 2015 and 2020. Moreover, many of the technologies that will deliver job growth, according to this analysis, will most certainly be integrating AI, which yields net positive employment growth.
And while there are insinuations about the vulnerability of some professions which require less cognitive function, we are still in a wait and see mode.
There is also an optimistic side to this scenario: the real possibility of reduced stress, increased leisure and a more balanced life.
What this also allows is the creation of new jobs. For x.ai, this means AI trainers, new positions that include a combination of QA, labelling and annotation. They also created a new role, the AI Interaction Designer, who designs the dialogue and determines the data input and the fitting responses. That person will train the system to be more empathetic. These are all convergent roles that have never existed before.
Privacy and the Use of Human Data
Most people realize that there may be a real risk when data becomes increasingly analyzed and contextualized. As systems begin to learn from the nuances of language, they will get better in developing understanding of human intention. There is some time before we need to worry about reaching the level of Artificial General Intelligence or “thinking machines”. For now, AI, provides seamless, effective solutions to improve human productivity and decision making.
Is there worry that these systems will take advantage of our data?
x.ai has access to information in scheduling related emails and Dennis asserts they are in the business of creating an agent”– a utility so good that you are willing to pay for it.” x.ai doesn’t mine the user inbox. The system is called when it is specifically asked to perform a task. It’s at that moment that they are exposed to the information.
When Amy and Andrew Ingram Grow Up…
Dennis notes that today there are 10 billion formal meetings being set up in the US alone. Where will x.ai be in 5 years?
I want to chase that 10 billion and once and for all, take the pain away for everyone. I want everyone to have an Amy and Andrew so then the dialogue is an exchange, not between man and machine, but an internal negotiation. That eventually increases happiness in the network.
The time people will save in real time will free up additional time to do more meaningful things. And that I find a worthwhile chase.
About Dennis Mortensen: Dennis is a pioneer and expert in the analytics, optimization and big data space and has been since its inception. He is also a fully-fledged entrepreneur and successfully delivered a number of company exits including one to Yahoo! Dennis is also an accredited Associate Analytics Instructor at the University of British Columbia, and the Author of Data Driven Insights. Follow @Dennis Mortensen and @xdotai on Twitter.
About Humanity in Data: Data isn’t only a story of numbers … or information … or insight. We forget, it is also about the people who interpret that data. Looking for Humanity in Data seeks to explain the ways in which behavioral data represents the mind and emotions of those who generate it. We do this by engaging with data science specialists and practitioners in this space. Our purpose is to translate from the scientific to the every day impact of technology on the individual. We’ll discuss topics about information collection, privacy and governance, increased contextualization and what is at stake for everyone.