AI/ML Business Model Dynamics
Artificial intelligence, machine learning, deep learning; basics, fundamentals, dynamic; CrowdStrike
There are many ways businesses can be successful. Some of the most successful businesses have similarities in their underlying business models.
As Nick Sleep and Zak called out in their Nomad Investment Partnership letters, scale economies shared is an example of such a theme. Walmart, Costco, Amazon, and Dell are some examples of great companies that flourished from this common concept of scale economies shared.
Underpinning the scale economies shared business model is the core, fundamental idea of being able to leverage size as an asset, such that as the company grows, their moat becomes strengthened.
While not necessarily having to adhere to the scale economies shared model, businesses focused on artificial intelligence (AI) and machine learning (ML) technologies benefit greatly as they grow.
In a business such as Costco or Walmart, the key resource in scale is sales volume. This leads to better negotiations, optimized supply chains, and other cost efficiencies.
In AI/ML businesses, the key resource is a rich data set which leads to a superior offering.
AI, ML, and DL Basics
“AI is the science and engineering of making intelligent machines.” — John McCarthy
Artificial intelligence (AI) is the ability of a computer or machine to mimic or imitate human intelligent behavior and perform human-like tasks. AI is the broadest term, including the ability to solve anything from narrow to general tasks.
Machine learning (ML) is a subset of AI that has the ability to automatically learn from data without explicitly being programmed or assisted by domain expertise. It is a technique for realizing AI. Given a set of tagged data with features, various algorithms can be used to train a model, evaluate performance and accuracy, and ultimately make predictions.
Deep learning (DL) is a subset of ML that has the ability to automatically discover the features or labels used for classification. This is the next evolution of machine learning.
ML Fundamentals
As noted above, machine learning allows a program to infer an outcome or prediction from a given data set, the training data.
Consider the following example from Towards Data Science.
This table identifies the type of fruit based on its characteristics:
As you can see on the table above, the fruits are differentiated based on their weight and texture.
The last row gives only the weight and texture, without the type of fruit.
A machine learning algorithm can be developed to try to identify whether the fruit is an orange or an apple.
After the algorithm is fed with the training data, it will learn the differing characteristics between an orange and an apple. The resulting model will assign weights, or importance, to each of the features (weight, texture).
Therefore, if provided with data of weight and texture, it can accurately predict accurately the type of fruit with those characteristics.
Evidently, the training data is of key importance. Training data set must be structured and tagged appropriately.
Algorithm developers should be cautious about overfitting their algorithm to the particular set. It is important to verify machine learning models with real situations, confirming the outcomes are in fact as expected.
Real world applications often include a confidence level with the output. If highly confident, it can pass without manual revision. If the confidence is low, a human may be asked to make the final decision.
Retraining the model is a critical iterative process; models can be continually improved by incorporating new data. With more data, the confidence level increases for any new scenario. As the confidence level increases, new rows of data start coming in automatically. The model is then retrained with the larger data set, effectively learning from each new data point. And the cycle continues.
Size as an Asset
Machine learning models naturally have a virtuous cycle.
The larger the data set, the better the model, the better the predictions. Repeat.
The model with the largest data that makes the best predictions will continue to attract more usage, or customers.
This cycle makes it such that the best model will most likely continue to improve at a faster rate than the second best model.
This leads to a winner-take-most market dynamic, where size is not a liability but a critical asset.
Case Study: CrowdStrike
Cybersecurity is and has been an important field for a long time. In the past, cybersecurity firms were seen as having a ceiling as to how large they could grow.
Consider the antivirus market in particular. Vendors would essentially gather lists of known attacks and vulnerabilities and distribute these databases to all the endpoints (computers, phones, etc) they sought to protect.
As a vendor grew and became successful, hackers targeted their solutions. Because the security technology was limited to point-in-time decisions, hackers would inevitably find holes to take advantage of and get through.
For these legacy antivirus solutions, size was an issue. It attracted attention which weakened their offering, making it difficult for cybersecurity firms to continue to scale.
Now, with machine learning, size has become an asset.
CrowdStrike is the leader in endpoint security, the next-gen antivirus solution built upon machine learning.
CrowdStrike’s solution leverages streaming to create contextual awareness.
They centralize signals from all of their connected endpoints into their ThreatGraph platform.
Their machine learning model highlights any suspicious activity or automatically prevents attacks when highly confident.
This streaming prevention solution gives organizations the ability to see how an attack is unfolding and then stop it with a very high degree of confidence.
This pattern-based, machine learning prevention, much more powerful than legacy solutions, can help block attacks the system had never seen before.
Rather than depending on an explicit list of attacks and point-in-time decision making, the ThreatGraph learns from many signals from many different devices.
As CrowdStrike adds more endpoints and/or more customers, the platform collects more signals onto their platform. More signals lead to a more accurate solution, which in turn leads to more breaches being stopped. This strengthens their offering, which attracts more customers. And the cycle repeats.
Closing
Machine learning is ultimately driven by data; the largest, most comprehensive data sets have a clear advantage.
Given the market dynamic, it seems to be a durable competitive advantage that is enhanced as these companies continue to grow.
--
Torre Financial is an independent investment advisory firm focused on emerging and established compounders.
Federico Torre
Torre Financial
federico@torrefinancial.com
https://torrefinancial.com
Disclaimer: This post and the information presented are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.