THOUGHT LEADERSHIP

Wendy M. Pfeiffer: Well, it’s further back than you might think. Even though AI in the 1980s was really only surfacing itself in the Hollywood movies of the time, I was actually involved with some real software engineering on the front line at the cutting edge of AI development. I worked for a company called ILOG that was based in France and these guys really had some practical application use cases of AI in C++ class libraries that other companies could use to augment their software. Although ILOG was much later acquired by IBM in 2009, we were sowing the seeds of AI tools that would both “augment and enable” software a long time before it has enjoyed its currently more widespread popularization and proliferation. 

Advances in processing power, compute power, storage, and so on over the last 20 years have of course enabled us to work with vastly more complex math, which has allowed us to do really cool AI tricks that even manifest themselves in mobile apps. Sometimes it’s just that small amount of ML on a mobile device that can make all the difference. It all comes down to the math, now that we have the datasets to really leverage AI inside new coding methodologies, I actually hope that machine intelligence will start to outgrow the mathematical boundaries it stems from.

Nutanix Q&A bot: Inside Nutanix, your team has implemented the X-Bot toolset as one of its first AI projects, tell us more about how it was developed and how smart it is.

Wendy M. Pfeiffer: We did a very “hybrid IT thing” with this product. By that I mean, we built some parts of it where we knew we had the right in-house coding skills, plus we also integrated other parts in where we could see that we could bring in third party intelligence and get to an even higher plane.

X-Bot is informed by a quantitative measurement that we call  “first time right” (FTR). It is software that knows the optimal order in which things should happen. Just like I, as a human being, know the correct order and optimal workflow for getting dressed (underwear, pants, socks, and then shoes), children initially need to be taught those things. 

So we took that notion of quantitative workflow mechanics into X-Bot, but we also engineered it to ensure it was following the optimal interaction design. X-bot has a qualitative element that incorporates our measured Net Promotor Score, or NPS. This means that we are able to optimize not just what the AI does, but also how well the AI to human interaction works. 

Human beings are not very good at context switching; they prefer a more narrow focus, whatever the workflow task is in question. We need to be able to give our workforce the chance to work in a more direct line, that way they will ultimately design more appropriate machine interactions. Machines themselves are also getting better at refining their code and getting better at learning and augmenting what they do, so a virtuous circle is possible here if we’re smart about the way we use AI.

Nutanix Q&A bot: Why do you think some CIOs are holding back in terms of their implementation of AI into live production environments – is there still some level of mistrust at the user level, or does the total technology proposition fail to appear as robust as it needs to be?

Wendy M. Pfeiffer: The kinds of Machine Learning (ML) and Natural Language Processing (NLP) that companies need to really use in production are the ones that understand the nuances of the way we work. They need granular tuning that works on a user-agnostic, internationally culturally aware basis, and they need to be able to utilize unique company training and interaction data that makes the difference between people liking it and adopting it, or people hating it. The resulting AI technologies will flourish in live production environments for any CIO.

Nutanix Q&A bot: What kind of real kickstart do you think AI could use to promote its adoption and ubiquity? 

Wendy M. Pfeiffer: I’m part of an organization called Association for Inclusive AI (AIAI), and the work here is part of what’s really going to make adoption more widespread. We know that the social cohort of individuals building AI engines is not broad enough. This group has until now been too narrow in socio-economic terms, in cultural scope, in gender and neural diversity, and even in terms of age spread. As a simple example, think about the fact that some older people talk with a slower cadence, as do some people with disabilities. How can we expect NLP interfaces to work effectively for mission-critical and life-critical applications unless we build them with training data that includes a total lack of bias for speech cadence?

Nutanix Q&A bot: What really makes AI work for human beings intelligently, practically, and pragmatically?

Wendy M. Pfeiffer: What I truly love is consumer technology (even though I work for an enterprise IT organization)  and right now I often think about the fact that, during this time of global pandemic, when we all pivoted to work from home, we started being more reliant on home technologies and consumer-grade tech. All that technology had, in so many ways, done a better job of making itself invisible to us. I think this has shown us how enterprise technologies have been left behind in some sense. This is why digital transformation has gone from the aspirational to the foundational–that is, we need this stuff to work fast, first-time, intuitively, and intelligently.

Going forward, the best AI will be able to deliver granular and nuanced capability that people don’t even notice. That’s a trait you can apply to all the best ML-enhanced technologies that already exist – I mean, when you drive your car, do you think about how the fuel injection system is working? When you travel by airplane, do you sit there and think about the thousands of course-corrections each hour that are executed by the autopilot? When you use a word processor, do you stop to think about the application parallelism that is allowing you to both type and, concurrently, get alerted to spellcheck errors? Making infrastructure invisible is what we’re talking about here, and it is perhaps no coincidence that this has been embedded in the DNA of Nutanix from the start.

Nutanix Q&A bot: Your “last dinosaur valley”  theory is pretty special. Can you explain what it means in the context of the conversation here?

Wendy M. Pfeiffer: Well, thanks. It’s a little original theory of mine rather than any piece of palaeontological science, but I think it speaks volumes.

The last dinosaur valley was populated with herds of Diplodocus, Triceratops, Velociraptors, and of course a few Tyrannosaurus Rex. They were all quite happy but eventually started feeling pretty hot and hungry. They also started getting troubled about all the little pesky mammals running around bashing sticks and stones together. Although completely disconnected from the rest of the planet, they finally got the big wake up call when the asteroid hit. But by then it was too late.

IT organizations today need to remember those last dinosaurs and make sure they operate sustainably, resilient to the impact of coming asteroids. Whether it is the impact of social media, the advent of cloud, the birth of microservices, the AI renaissance, or a non-tech factor like the next global pandemic (god forbid), it’s a question of being aware of what’s already around the corner, and tuning ourselves and our technology ecosystem to sustain us through whatever comes our way. 

We don’t need to focus on getting back to that happy valley and making sure that we have enough resources to sustain us. Instead, like the modern day crocodile, we need to focus on thriving where we are in the new world and building new lives where necessary. The underlying substrate of the world (and the topography of technology) around us has already changed, so we need to look at what we need for the future and what will no longer serve us. The future can be a happy place and a more intelligent place, but we need to be smart about how we get there.