Artificial intelligence, machine learning, and deep machine learning; are they all the same?

Your ontological connection between Neural Networks as a subset of Deep Learning as a subset of Machine Learning as a subset of AI is excellent, however, there are competing models, such as how to put these into practice, especially Enterprise Products (including or perhaps especially Cybersecurity products). I was introduced to the HPAI model -- https://www.hpcwire.com/2016/11/10/hpc-meets-ai-creates-new-grand-challenges/ -- while at NVIDIA, and this model examines a different ontology: one where Neural Networks remain a subset of Deep Learning, but Deep Learning is instead tied to High-Performance Computing Optimization methods through Simulation. This is opposing (or colliding at opposite spectrums) with Machine Learning, which is more-so a subset of Search, Automation, and, yes, AI. For Cybersecurity products, one might use both models to determine affected outcomes. The book, How To Measure Anything in Cybersecurity Risk, is actuarial about measuring control effectiveness -- so I might recommend its approaches as a primary resource to compare products. However, just thinking in terms of this second model, HPAI, I see Cybersecurity products such as Hortonworks Cybersecurity Package (HCP, commercial), ASGARD (open-source?), Graphistry (commercial) or Apache Metron/Spot (open-source) as matching components to the entire HPAI model (instead of merely components of it). From merely an architecture perspective, I see Spark Streaming (e.g., Databricks) conforming to HPAI closer than, say, Hadoop (e.g., Cloudera, Hortonworks, MapR) or Cassandra (e.g., Datastax) alone -- not these can't be used in combination. Performance-wise, connecting Graphistry to both MapD (on the simulation side) and MapR (on the search side) might prove explosive, especially with equal parts in Spark Streaming in applications. Usability-wise, none of the outcomes from these Enterprise products can be harnessed (i.e., operationalized) by casual business users (only by data-science power users) unless there is also a data wrangling architecture, such as Trifacta. Regardless, most orgs are investing in DL/ML/AI (often through UEBA) where they are already invested otherwise. Invested in RSA, then Securonix. Demisto/Phantom, Exabeam. Microsoft, Varonis. Hortonworks, their own HCP (along with Metron and NiFi). HPE, the acquired Niara. Elastic, the acquired Prelert. Splunk, the acquired Caspida. Only Hortonworks is the full HPAI actualized, though. You'll see constant mention of machine-learning concepts from Palo Alto Networks, Tanium, and Splunk (especially in combination with the Accenture Cyber Defense Platform), but you could say anything has machine learning or AI -- these include hundreds of programming concepts and techniques since the 70s that made websites like Google and Amazon possible even in the late 90s. Hopefully my added discussion provided some clarity, and if it did not, then I suggest a quick read of -- https://www.sans.org/reading-room/whitepapers/critical/applying-machine-learning-techniques-measure-critical-security-controls-37247 -- for thoroughness.

This is great, thanks for sharing!

We should focus on the outcomes that AI can deliver. https://www.dhirubhai.net/feed/update/urn:li:activity:6297446830267596800

Interesting writeup.

I need to come up to speed on this to be able to comment, and my main objective would be in how it can be used to further predictive and detection applications.