AI In Action: AI and Cannabis
AI's use in medical applications is well known. We learn from Meni Morim, CEO of Namaste Technologies, how AI can benefit cannabis users.
- By James E. Powell
- October 22, 2019
Upside: We don't often associate AI with the cannabis industry. Tell us a little about what Namaste does and what drove your company to use AI.
Meni Morim: Namaste Technologies operates the largest global cannabis e-commerce platform, with over 30 websites in more than 20 countries. At Namaste, we're looking to provide an end-to-end service to our patients rather than just facilitate a purchase. Through AI technology and machine learning, we can assist customers from the moment they need to learn about cannabis and understand if it's right for them.
What does your app, Uppy, do and how is AI incorporated into it?
Uppy uses AI technology to track customers' usage patterns, medical needs, and product reviews to offer personalized recommendations for medical cannabis customers. Users enter their usage data into Uppy, telling us what kind of cannabis they took, how much they took, and when they took it. Uppy analyzes this data to determine the frequency of usage and type of cannabis products that worked for the user.
Uppy knows which products worked because it follows up with the users to ask how well it worked or did not work. It also compares this data to other data from other users and calculates the products that are most likely to have good results for the patient. Uppy was developed with AI from the ground up.
Is Uppy brand new? How was Namaste making recommendations before?
Uppy was first released a little over one year ago. All recommendations were made by a human nurse practitioner before that.
Why use AI? Why not just use a set of pre-coded recommendations (if customer used strain "a" and had headaches, then recommend strain "b" instead) for example? What does AI do that pre-coding recommendations doesn't?
Pre-coding is a rule-based approach, and it works only for cases where there are limited sets of options. When more complex domains are in use, it is not possible to generate or create rules due to an enormous number of factors. Just because one product gives a person headaches doesn't mean another product will solve their problems.
Cannabis plants contain chemicals called terpenes, which some humans may be allergic to. Allergies and other adverse reactions can cause negative symptoms and side effects, such as anxiety or paranoia. It is also important to analyze the ratios of THC to CBD, which helps control the effects of cannabis. Lastly, it is important to know how frequently a person doses cannabis and how much they dose.
There are a lot of strain attributes, different user preferences, and interactions (e.g., duration of sessions or speed of consumption, consumption method, time patterns, etc.) that must be considered. We want to take into account all of those factors. However, they compose a very complex domain that can't be covered by a rule-based approach.
That being said, we do incorporate a rule-based approach to filter out strains. For example, we know that X strain is superior for Y condition, but the user doesn't like it and shows us by explicitly clicking a dislike button, so first we apply AI to generate a list of candidates, and then we apply rules to polish the results.
What are the benefits for the consumer?
A lot of people medicate with cannabis, but with so many different strains out there, patients don't have all the information to make the right purchasing decision. Consuming the wrong strain is dangerous and can lead to severe side effects like anxiety or paranoia. Through the personalized recommendations that Uppy provides, patients can discover the product that best suits their individual medical needs, rather than having to go through a lengthy and costly trial-and-error process.
AI is only as good as the data it's fed. What sources of data are you using?
Uppy allows patients to quantify what a specific strain does for them -- whether it is a positive or negative impact on their symptoms, the time of day they used it, or the quantity. On the back end, we collect all of this data anonymously and use it to train the machine learning models so that the app can tell a patient with specific symptoms which strain had the best results for people who also had those symptoms.
Given that some users feel AI is a bit mysterious, and given the sensitive nature of medical information (let alone cannabis use), did you face any privacy issues?
Users of Uppy provide their data voluntarily and can select which questions they'd like to answer. All the data is collected anonymously and kept within the company.
AI can be a big step for many enterprises.
Prior to joining Namaste, I was the founder and CEO of Findify, an AI-focused turnkey solution that enables e-commerce sites to personalize search results, recommendations, and product collections based on a user's specific behavior and proprietary machine learning technology. After Findify was acquired by Namaste, I worked to incorporate AI technology into the Namaste platform through Uppy.
Where else do you see AI being used within Namaste?
Through Uppy, we hope to convert app users to purchasers of products on our e-commerce platform, CannMart. Once we can really integrate Uppy into CannMart, you will be able to search for the strain that meets your needs, click on the recommendation, and it will direct you right to CannMart where you can make the purchase. At the same time, we'll be collecting quality, structured data for the machine learning algorithm.
James E. Powell is the editorial director of TDWI, including research reports, the Business Intelligence Journal, and Upside newsletter. You can contact him
via email here.