AI by the Cold Light of Day

It’s been a heck of a month for me, but one place where life is finally calming down a bit is Machine Learning. Whereas the last several quarters have seen a very frothy set of developments in the field, it finally feels like I’ve had the time and the space to get out of what-are-we-doing-right-now tactics and think a lot harder about strategy. Moreover, even though ML is advancing at a ridiculous pace, it still feels a bit like we’re in a slight pause, giving technology leaders time to think. That, and I’ve had the opportunity to spend a considerable amount of time this week reading, which always puts me in a good head space.?

So… a lot has happened… how am I thinking about advances in Machine Learning now I can look at things in the cold light of day?

Cognitive Automation

I think a watershed moment for me was a video by the Brookings Institution, where someone made the comment that we’ve experienced automation like this before - the difference was that as physical automation, and this time, to a debatable but non-zero degree, it’s cognitive. That probably feels worse, as our cognition is the last bastion of superiority we have over machines, but as Ray Dalio would say, let’s accept reality and deal with it.?

Our learnings with physical automation lead us back to a discussion of the Industrial Revolution. I’ve talked about that before, so I’m going to skip rather than repeat myself. However, when we think about automating some of the things we think of as a mental domain, it just feels worse, at least to me… it’s probably about what we get used to. My car seldom makes me feel like I have inferior legs, and OpenAI shouldn’t make me feel more stupid than usual. Rather, we need to think about this just like we do a car: what new opportunities does this vehicle open up for us? Through that lens, suddenly things get rather exciting.?

ML as a Moat

The question I was trying to answer is how to turn Machine Learning into a defensible moat. How can I use it in such a way that I build a long-term competitive advantage rather than think about it as the great equalizer?

That line of questioning reminded me of an earlier exercise in our "cognitive revolution" (as I shall now call this decade). At Florida Tech., as the Internet really took hold I started to notice a change in my students. My A+ students were still exactly that: sharp as a tack, eloquent, and driven. They had natural ability honed to a fine edge by a good work ethic. Nothing unusual to see there. However, the next tier of students down maybe had a little less ability, but the same innate drive.

My car seldom makes me feel like I have inferior legs, and OpenAI shouldn’t make me feel more stupid than usual. Rather, we need to think about this just like we do a car: what new opportunities does this vehicle open up for us?

These students closed their talent gap by becoming fantastic librarians of the knowledge that was out there on the Internet. That is, they used evolving technology to figure out how to bring the information they lacked to their fingertips in the most effective way possible. Because of these students, I realized that being able to really use the Internet effectively was the skill needed for the 2000’s - and that doing so well allowed for a compression of the talent gap. Many of these students launched themselves into very successful careers, and as I can see from LinkedIn are doing much better than fine.

The Way Forward

This line of thinking has given me the model of how to use our new LLM-based cognitive prosthesis. It’s not going to replace top-talent any time soon, but if you’re not lucky enough to be in that elite group of naturals (and even if you are, I think the lift provided by getting rid of automatable tasks is huge), you better learn how to get the best of these tools now.?

With that in mind, I once again go back to that whole defensible moat question. Here’s where I’ve landed.

• First, at least for now, I think your people remain your finest attribute. A good team is still a good team, and a great team is still a great team.? ML helps folks a lot, but isn’t going to replace raw talent any time soon. Lean in to hiring the very best you can, over and over again, and make sure your best talent doesn’t go walkabout.?Then make sure you are upskilling them in AI-based skills and how to leverage these new technologies.
• Second, and this is a big one, take a good look at the data you have that you think is world class. If you do X (whatever it is) better than anyone else in the world, you’re spinning off data that can be used to fine tune models to make those models better than those your competitors can make. Embrace your data, and stop letting it fall off the back of the conveyor belt. Figuring out how to capture and retain the data flowing through your systems is your new mission, and I promise you that “there is gold in those olive groves”.?
• Third, speed and agility are key. In 2030, there’s only going to be two kinds of cybersecurity companies. Those who truly embraced machine learning and those that sincerely wish they had. That’s it. That’s the reality you have to accept. Which side of that divide would you like to be on?
• Fourth, your brand and reputation are yours, for better or for worse. Keeping your brand’s appeal (and growing it) remains critical to building the overall success of your company. In a world where there’s likely to be a lot of noise in the “who’s actually doing this well?” signal, branding is massively important.?

The good news about all of this is that the defensible moat is still buildable - and the savvy among you will see the roiling of the waters in which we swim for what it is: opportunity. I hope that like me you’ve now had the chance to think things through and can see them by the cold light of day. Go grab that opportunity, and make your people win.?