At the risk of being reprimanded by Saul for an off-topic discussion, I’ll attempt to settle (or at least address) this discussion, and relate it to the initial purpose of this thread (i.e. Crowdstrike).
I think there is more agreement than it seems between buynholdisdead + rtichy and canonian. The disagreement stems from a dichotomy in the definition of AI. I hope I can provide a point of view that somewhat unites the divergence.
By the very complicated nature of defining AI, I like thinking of it as a spectrum (as canonian alluded) where humans and machines work together to generate insights. In short:
Assisted Intelligence
-Action performer: machine + human
-Decision maker: human
-Applications: following if/then decisions, filling-in forms, opening email attachments
-Examples: this is where RPA (robotic process automation) companies play, such as Blue Prism, and the soon-to-be-public UI path. Think of an Excel macro that can execute actions across your apps, not just within Excel
Augmented Intelligence
-Action performer: machine
-Decision maker: machine + human
-Applications: following commands via voice/vision, automating tasks with judgement, automating tasks that improve over time
-Examples: this is where algorithms get more complex, and where some of our companies play. A relevant example would be Crowdstrike’s Falcon’s use of anomaly detection - where it can detect and flag a vulnerability identified in my environment based on a similar one that occurred across the world a few days ago
Autonomous Intelligence
-Action performer: machine
-Decision maker: machine
-Applications: generating hypothesis, self-selecting options based on complex choices, having situational awareness
-Example: the use of autonomous intelligence is still nascent in commercial applications. Perhaps Deepmind’s Alphafold breakthrough is a decent notable example (a starting point to understand: https://www.youtube.com/watch?v=gg7WjuFs8F4&t=1s)
Another way to think about this is picturing a matrix, with “complexity” on one axis and “feedback loops” on another axis. Assisted intelligence is applicable on tasks of low complexity and strong feedback loops. Augmented intelligence becomes applicable on tasks of moderate complexity and strong feedback loops. Autonomous intelligence is beginning to become relevant on tasks with moderate feedback loops and high complexity.
There is a long way to go for high complexity tasks with low feedback loops to become relevant for AI. The challenge of strengthening feedback loops is contextual, and can be improved with the increase of digitization. Meanwhile, the challenge of complexity is technical, and will be improved by advancement in technology.
Looking at Crowdstrike’s open roles, you can clearly see that they have talented data scientists operating at the cutting edge of machine learning. For example, they’re looking for an intern to “deploy state-of-the-art machine learning classifiers that learn from extremely large amount of sample files”. There are 58 results when I search for “data science” jobs and the tech stack within those roles is as advanced (in ML) as it gets.
So, while marketers do love to tout “AI” across almost any digital business nowadays, it is fair to say that companies as advanced in data science and machine learning as Crowdstrike ought to get the benefit of the doubt. At least it’s fair to say that they are more advanced in “the spectrum” than most companies claiming the use of AI.
I apologize for the off-topic discussion, but “AI” is a prevalent theme for some of our companies and is thus an important topic for us to understand. I hope that this explanation is somewhat useful. If it is found to be too off-topic, please delete (no offense taken).