1 00:00:00,000 --> 00:00:09,400 Welcome to Cannabis Data Science right before July 4th. 2 00:00:09,400 --> 00:00:15,840 So we'll keep it light, we'll keep it fun, and then I think we'll walk away at the end 3 00:00:15,840 --> 00:00:22,760 of the day with a few insights because I think the data is so rich that every time we go 4 00:00:22,760 --> 00:00:25,840 out panning, we come back with gold. 5 00:00:25,840 --> 00:00:32,920 So sure enough, got a couple gold nuggets today to share with you and point you in the 6 00:00:32,920 --> 00:00:42,000 direction of some rich gold veins so you can go mining for gold of your own. 7 00:00:42,000 --> 00:00:48,040 Till we jump into this, if any of you have anything to share, so I know it was a long 8 00:00:48,040 --> 00:00:56,160 day for you Jerry, anything on your mind as far as future directions for the group? 9 00:00:56,160 --> 00:01:04,960 So next month is a new month, a new quarter, so perhaps we can turn a new leaf and see 10 00:01:04,960 --> 00:01:07,280 a new fiscal year for some people. 11 00:01:07,280 --> 00:01:08,280 Exactly. 12 00:01:08,280 --> 00:01:16,160 So we can see what new topics may be. 13 00:01:16,160 --> 00:01:21,440 I posted a pretty extensive business model on Slack. 14 00:01:21,440 --> 00:01:24,640 I had a conversation with Charles about it. 15 00:01:24,640 --> 00:01:30,960 I'd like to have one-on-ones with everybody just to see what their thinking is on what 16 00:01:30,960 --> 00:01:32,080 I'm proposing. 17 00:01:32,080 --> 00:01:38,600 I do have to say right out front that I am 100% opposed to getting involved in NFTs or 18 00:01:38,600 --> 00:01:44,280 crypto in any way, form, or any way, shape, or form, and I just won't do it. 19 00:01:44,280 --> 00:01:46,320 I will not be involved in that. 20 00:01:46,320 --> 00:01:54,640 It's a very unstable foundation to build on. 21 00:01:54,640 --> 00:02:04,200 Well to each their own, so for those who are new, what is the crypto, the NFTs, essentially 22 00:02:04,200 --> 00:02:11,520 I was trying to find a mechanism that people can get paid directly for their data and their 23 00:02:11,520 --> 00:02:12,520 algorithms. 24 00:02:12,520 --> 00:02:19,920 Essentially, the downside of analytics is I'm essentially a middleman. 25 00:02:19,920 --> 00:02:26,360 If people want to list their data through their algorithms, they'll have to essentially 26 00:02:26,360 --> 00:02:28,040 contract with analytics. 27 00:02:28,040 --> 00:02:35,960 We all have to get those listed through a mechanism like one of your traditional payment 28 00:02:35,960 --> 00:02:42,760 providers, and it's just a lot of hassle. 29 00:02:42,760 --> 00:02:47,600 What ends up happening is I'll list my data and I'll list my algorithms, but then I don't 30 00:02:47,600 --> 00:02:50,520 think other people get to participate. 31 00:02:50,520 --> 00:03:03,600 The idea is what if people could somehow post their data and then their data remains private 32 00:03:03,600 --> 00:03:12,920 unless it gets purchased by a particularly interested party. 33 00:03:12,920 --> 00:03:18,680 In my exploration of what is going on in the data space, there's a cool company, Ocean 34 00:03:18,680 --> 00:03:19,680 Protocol. 35 00:03:19,680 --> 00:03:25,520 It looks like they've set up a data market. 36 00:03:25,520 --> 00:03:32,800 Like I said, I'm not necessarily sold on it, but it just looks like a way that you could 37 00:03:32,800 --> 00:03:34,760 post a dataset. 38 00:03:34,760 --> 00:03:41,040 I believe it would be encrypted, so I don't think anyone would have access to it unless 39 00:03:41,040 --> 00:03:45,560 they essentially buy from you. 40 00:03:45,560 --> 00:03:54,560 I think it's still the early stages for this technology. 41 00:03:54,560 --> 00:04:05,600 I just see the way the tide is turning. 42 00:04:05,600 --> 00:04:13,160 I don't know if these giant middlemen have much of a long future. 43 00:04:13,160 --> 00:04:24,360 Do all these private companies where the executives walk away at the end of the day with a lot 44 00:04:24,360 --> 00:04:30,360 of money and then maybe they're using interns, not paying the interns, who knows what's going 45 00:04:30,360 --> 00:04:31,360 on? 46 00:04:31,360 --> 00:04:36,000 Is that a sustainable business model? 47 00:04:36,000 --> 00:04:40,000 It's been sustained for hundreds of years. 48 00:04:40,000 --> 00:04:41,000 Exactly. 49 00:04:41,000 --> 00:04:48,280 That's the point is we've had to put up with this for hundreds of years and thank God we 50 00:04:48,280 --> 00:04:54,800 now finally have technologies that may finally allow- 51 00:04:54,800 --> 00:04:58,080 There's cutting edge and there's bleeding edge. 52 00:04:58,080 --> 00:04:59,720 You're on the bleeding edge. 53 00:04:59,720 --> 00:05:00,720 True. 54 00:05:00,720 --> 00:05:11,040 I tell you, the amount of theft, crypto is for criminals. 55 00:05:11,040 --> 00:05:16,940 What you're talking about is a totally unregulated system in which you have no protection whatsoever. 56 00:05:16,940 --> 00:05:22,880 Having been burned, having had identities theft occurred by my involvement in crypto, 57 00:05:22,880 --> 00:05:27,000 the downside is unacceptable. 58 00:05:27,000 --> 00:05:33,360 I'll just push back in that as a small business owner, I already feel like I'm 100% unrepresentative. 59 00:05:33,360 --> 00:05:37,560 It's not like I even have any legal mechanism. 60 00:05:37,560 --> 00:05:41,720 The court systems are so biased. 61 00:05:41,720 --> 00:05:44,280 It's basically money talks. 62 00:05:44,280 --> 00:05:52,000 I've gone through the wringer and I've seen under the curtain, money talks, especially 63 00:05:52,000 --> 00:05:55,000 in that realm. 64 00:05:55,000 --> 00:06:00,480 The problem with crypto is that there's nobody to talk to. 65 00:06:00,480 --> 00:06:02,640 I don't want to go on too much. 66 00:06:02,640 --> 00:06:03,960 I have a problem with my bank. 67 00:06:03,960 --> 00:06:06,440 I go to my banker. 68 00:06:06,440 --> 00:06:11,400 Like I said, if this is something like you're interested in checkout ocean protocol, I'm 69 00:06:11,400 --> 00:06:18,320 not necessarily tying analytics to the crypto space at all. 70 00:06:18,320 --> 00:06:22,600 However, I do explore cool new technologies. 71 00:06:22,600 --> 00:06:32,840 It's something to explore, but I'm not going to put effort into a business that's paying 72 00:06:32,840 --> 00:06:37,760 me in a mechanism that I have no faith in. 73 00:06:37,760 --> 00:06:38,760 Then don't. 74 00:06:38,760 --> 00:06:39,760 Okay. 75 00:06:39,760 --> 00:06:40,760 I'll see you later. 76 00:06:40,760 --> 00:06:41,760 Okay. 77 00:06:41,760 --> 00:06:50,680 That's sort of the way it is. 78 00:06:50,680 --> 00:07:08,600 That's sort of my attitude now is I'm seeing the industry getting captured by big players. 79 00:07:08,600 --> 00:07:13,280 We'll actually talk specifically about that today. 80 00:07:13,280 --> 00:07:16,640 As I said, I'm not even really here to talk about cryptocurrency. 81 00:07:16,640 --> 00:07:18,120 I'm here to talk about statistics. 82 00:07:18,120 --> 00:07:21,760 Why don't we change gears? 83 00:07:21,760 --> 00:07:23,480 That's what I know at the end of the day. 84 00:07:23,480 --> 00:07:25,360 That's what we're in the business of doing. 85 00:07:25,360 --> 00:07:30,600 We aggregate cannabis data and provide statistics as a service. 86 00:07:30,600 --> 00:07:34,840 I was looking at a mechanism for other people to get paid. 87 00:07:34,840 --> 00:07:37,840 That's the way I currently do this. 88 00:07:37,840 --> 00:07:43,200 I actually kind of like that it's controversial because I don't think if you're not stepping 89 00:07:43,200 --> 00:07:46,760 on toes, then I don't think you're making progress. 90 00:07:46,760 --> 00:07:52,760 I think Canlytics does an awesome job at lighting people's pants on fire. 91 00:07:52,760 --> 00:07:58,000 I'm proud for that and going to keep paving the way forward. 92 00:07:58,000 --> 00:08:00,640 I think we're in good territory. 93 00:08:00,640 --> 00:08:22,920 I'll go ahead and share my screen and you two will be in for a couple golden nuggets. 94 00:08:22,920 --> 00:08:29,800 I saw an article the other week where there was a cannabis company. 95 00:08:29,800 --> 00:08:36,560 So where does the story begin here? 96 00:08:36,560 --> 00:08:40,960 So there's Respect My Region, which is a... 97 00:08:40,960 --> 00:08:47,520 I don't want to miscategorize them, but I guess they're like a culture, branding, marketing 98 00:08:47,520 --> 00:08:50,640 company out of Seattle, Washington, I do believe. 99 00:08:50,640 --> 00:08:55,880 Hey, John, we're off to a rocky start, but we're going to just keep paving the way forward 100 00:08:55,880 --> 00:08:56,880 here. 101 00:08:56,880 --> 00:08:59,080 But long story short. 102 00:08:59,080 --> 00:09:04,080 Okay, there's noise, there's company Jungle Boys coming into... 103 00:09:04,080 --> 00:09:12,200 Okay, so what was controversial about the Jungle Boys before I go further? 104 00:09:12,200 --> 00:09:14,320 If I may ask. 105 00:09:14,320 --> 00:09:20,880 Just over on the East Coast, there's a really big push against MSO taking over all of these 106 00:09:20,880 --> 00:09:26,440 smaller licenses that were originally awarded to smaller companies. 107 00:09:26,440 --> 00:09:32,640 Just like from a consumer standpoint, you've got dwindling choices whenever you've got 108 00:09:32,640 --> 00:09:36,840 fewer and fewer companies running the market. 109 00:09:36,840 --> 00:09:45,160 Well, Kimberly, hopefully you'll stay to the end because I do have something to say to 110 00:09:45,160 --> 00:09:46,160 that. 111 00:09:46,160 --> 00:09:54,880 So you kind of, I guess, cut straight to the point there. 112 00:09:54,880 --> 00:09:55,880 My apologies. 113 00:09:55,880 --> 00:09:57,400 Well, it's not that... 114 00:09:57,400 --> 00:10:03,080 No apologies there, but if you're all on board, let me go through... 115 00:10:03,080 --> 00:10:20,400 Well, yeah, we'll save it for the end here because it's sort of a long story to tell 116 00:10:20,400 --> 00:10:21,400 here. 117 00:10:21,400 --> 00:10:22,400 Okay, so it's controversial. 118 00:10:22,400 --> 00:10:34,040 So we've got a company from the West Coast entering into the Florida market. 119 00:10:34,040 --> 00:10:42,760 Okay, so please stay tuned to the end because we'll kind of get to that. 120 00:10:42,760 --> 00:10:46,120 So I'm going to sort of walk through the story. 121 00:10:46,120 --> 00:10:49,240 So we'll kind of keep that in the back of our mind. 122 00:10:49,240 --> 00:10:53,360 Now we'll sort of just enter into the realm of statistics. 123 00:10:53,360 --> 00:10:57,120 We'll touch on statistics here for a little while and then sort of circle back with our 124 00:10:57,120 --> 00:10:58,120 story. 125 00:10:58,120 --> 00:10:59,120 So essentially... 126 00:10:59,120 --> 00:11:00,120 All right, now... 127 00:11:00,120 --> 00:11:04,400 Okay, so cannabis industry... 128 00:11:04,400 --> 00:11:07,120 Okay, not great. 129 00:11:07,120 --> 00:11:13,480 Now we run into the machine learning and also not necessarily great. 130 00:11:13,480 --> 00:11:19,080 You see, okay, you've got people and what you see a lot is people are kind of doing 131 00:11:19,080 --> 00:11:25,440 away with the statistics that underlies all this and we may kind of need to get back to 132 00:11:25,440 --> 00:11:27,240 the theory. 133 00:11:27,240 --> 00:11:38,560 So three big problems and these may even kind of even be similar overlapping problems with 134 00:11:38,560 --> 00:11:40,800 the cannabis industry in its own way. 135 00:11:40,800 --> 00:11:42,760 So three big problems. 136 00:11:42,760 --> 00:11:44,160 So dependency. 137 00:11:44,160 --> 00:11:49,280 So depending on your data, you may get an entirely different outcome. 138 00:11:49,280 --> 00:11:50,840 Consistency. 139 00:11:50,840 --> 00:11:55,480 Depending on how you go about training your models, you may get an entirely different 140 00:11:55,480 --> 00:11:56,880 outcome. 141 00:11:56,880 --> 00:11:59,720 Finally, transparency. 142 00:11:59,720 --> 00:12:02,400 Does anybody even really know what's going on? 143 00:12:02,400 --> 00:12:06,200 I see a lot of hand waving. 144 00:12:06,200 --> 00:12:14,800 I see people just saying, oh, we're going to use model X and then I look into it and 145 00:12:14,800 --> 00:12:24,160 maybe the model is appropriate but it should be standing on some theoretical foundations 146 00:12:24,160 --> 00:12:25,160 here. 147 00:12:25,160 --> 00:12:26,160 Cool. 148 00:12:26,160 --> 00:12:33,480 Sorry, I'm sort of all over the map but I'll tie it all together. 149 00:12:33,480 --> 00:12:38,680 Now, okay, so now where are we going with this? 150 00:12:38,680 --> 00:12:47,280 Okay, so we've got a lot of reviews data and we could actually use machine learning to 151 00:12:47,280 --> 00:12:54,080 get a measure of, for example, people's emotion. 152 00:12:54,080 --> 00:12:59,960 How do people feel about Jungle Boys? 153 00:12:59,960 --> 00:13:04,240 How engaged are people with Jungle Boys? 154 00:13:04,240 --> 00:13:09,840 When they write about Jungle Boys, are they depressed or not? 155 00:13:09,840 --> 00:13:13,680 Does this correlate with someone's personality? 156 00:13:13,680 --> 00:13:16,920 How intense are they about this? 157 00:13:16,920 --> 00:13:23,520 So for example, I just learned about this company honestly just in the past couple of 158 00:13:23,520 --> 00:13:32,080 weeks so my intensity may not be the same as someone who's known about the company for 159 00:13:32,080 --> 00:13:34,040 a longer time. 160 00:13:34,040 --> 00:13:38,320 Firstly, polarity, positively or negatively. 161 00:13:38,320 --> 00:13:45,440 Do you have a negative view of Jungle Boys or a super positive view of Jungle Boys? 162 00:13:45,440 --> 00:13:56,680 So essentially, these are all metrics that we can start to sort of tie into this concept. 163 00:13:56,680 --> 00:13:59,480 So basically, where was I going with this? 164 00:13:59,480 --> 00:14:06,560 Sorry, I got a little frazzled at the beginning but I should be able to bring it all together. 165 00:14:06,560 --> 00:14:14,560 Essentially, Respect My Region, they focus on culture and they say, oh, you know, culture 166 00:14:14,560 --> 00:14:15,560 is what matters. 167 00:14:15,560 --> 00:14:24,680 And what I say is what's cool is when we can tie the data into the culture, right? 168 00:14:24,680 --> 00:14:37,280 So instead of just going off of a narrative, so to speak, we could actually tie in some 169 00:14:37,280 --> 00:14:41,080 data with our narrative. 170 00:14:41,080 --> 00:14:46,680 So just going to show you the data real quick because it is cannabis data science after 171 00:14:46,680 --> 00:14:54,720 all and then we'll, as I said, we'll tie it back into the cannabis industry here at 172 00:14:54,720 --> 00:14:55,720 the end. 173 00:14:55,720 --> 00:15:02,760 Okay, so now that we're to the data, hopefully we can move through a little less frazzled 174 00:15:02,760 --> 00:15:05,360 here. 175 00:15:05,360 --> 00:15:20,120 Also, I may skip this little bit here at the beginning on personality but that's an aspect 176 00:15:20,120 --> 00:15:29,080 that I think we should look at in upcoming weeks is essentially the idea is different 177 00:15:29,080 --> 00:15:34,160 personality types respond differently to different stimuli. 178 00:15:34,160 --> 00:15:40,540 And so I was thinking cannabinoids are in fact stimuli so it would just be interesting 179 00:15:40,540 --> 00:15:49,440 to wonder if different personality types respond differently to different cannabinoids or terpenes, 180 00:15:49,440 --> 00:15:50,440 what have you. 181 00:15:50,440 --> 00:16:02,800 So that's sort of theoretical but something that can be done. 182 00:16:02,800 --> 00:16:10,400 The main thing I wanted to share with you today was these eight metrics that would be 183 00:16:10,400 --> 00:16:14,560 awesome to start collecting from reviews. 184 00:16:14,560 --> 00:16:27,360 And so while this is loading, I'll just share with you the work that can be done. 185 00:16:27,360 --> 00:16:31,640 And so we don't have to necessarily repeat ourselves, right? 186 00:16:31,640 --> 00:16:33,880 We don't want to reinvent the wheel each time. 187 00:16:33,880 --> 00:16:37,840 I was actually thinking about this analogy more. 188 00:16:37,840 --> 00:16:46,280 We just want to put some really cool wheels on a skateboard and go do some tricks. 189 00:16:46,280 --> 00:16:49,640 If the analogy is funny to you or not. 190 00:16:49,640 --> 00:16:56,400 But the long story short is there's a really cool company out there, Cintiq, so they have 191 00:16:56,400 --> 00:16:57,400 an API. 192 00:16:57,400 --> 00:17:04,560 I just emailed them to get an API key and they were pretty quick with the response. 193 00:17:04,560 --> 00:17:07,240 The API keys of course need to be kept private. 194 00:17:07,240 --> 00:17:09,920 I'm still reading their terms of service. 195 00:17:09,920 --> 00:17:17,400 So please feel free to read the terms of service yourself and apply for an API key if you wish. 196 00:17:17,400 --> 00:17:23,360 And then they have these metrics nicely packaged behind API endpoints. 197 00:17:23,360 --> 00:17:30,400 And so whether we can utilize theirs or they also have open sourced all of their code. 198 00:17:30,400 --> 00:17:39,920 So I was maybe going to see if we can't sort of like mix and match or just get all of the 199 00:17:39,920 --> 00:17:44,360 best parts and package them behind our own API. 200 00:17:44,360 --> 00:17:53,040 But the idea is and please check out their work because they've done really rigorous 201 00:17:53,040 --> 00:17:59,640 theoretical work, statistical work backing up their algorithms. 202 00:17:59,640 --> 00:18:03,440 So I think this is a really good starting point. 203 00:18:03,440 --> 00:18:07,480 But we don't necessarily just want to use this as a black box. 204 00:18:07,480 --> 00:18:10,880 So we don't necessarily want to use these as a black box. 205 00:18:10,880 --> 00:18:18,420 So over the I was going to try to do them all in one go but I was a little ambitious. 206 00:18:18,420 --> 00:18:21,640 But we can at least get through these first two. 207 00:18:21,640 --> 00:18:28,040 The idea is instead of just using their algorithm as a black box, which we can definitely do 208 00:18:28,040 --> 00:18:34,760 in production, it would be nice to kind of pick these apart to see what is going on under 209 00:18:34,760 --> 00:18:37,040 the hood. 210 00:18:37,040 --> 00:18:41,020 So we'll actually do that just right now. 211 00:18:41,020 --> 00:18:47,920 So we can go ahead and read in all of these strain reviews that we may have. 212 00:18:47,920 --> 00:18:55,120 And then the idea is there's a couple ways that we can go about grouping these. 213 00:18:55,120 --> 00:19:03,520 So for starters, if you were interested in getting a user profile, you could aggregate 214 00:19:03,520 --> 00:19:07,160 all of the reviews by user. 215 00:19:07,160 --> 00:19:10,440 And so I believe we began to look at that. 216 00:19:10,440 --> 00:19:17,160 But I just wanted to point out a little insight that we can find. 217 00:19:17,160 --> 00:19:20,840 And then I'll show you two new applications afterwards. 218 00:19:20,840 --> 00:19:24,400 So just part in the data loading. 219 00:19:24,400 --> 00:19:30,760 And basically the two other applications are we can also look at strains. 220 00:19:30,760 --> 00:19:35,120 And then finally, we'll look at brands, right? 221 00:19:35,120 --> 00:19:42,220 Because that was sort of where we entered the space was, you know, what's essentially 222 00:19:42,220 --> 00:19:46,080 the sentiment around trangle boards. 223 00:19:46,080 --> 00:19:53,640 And we'll just try to look at the data and then tie it in with the narrative. 224 00:19:53,640 --> 00:19:54,800 This is taking a hot minute. 225 00:19:54,800 --> 00:19:58,880 So if there's any other thoughts, comments, questions, I'd be happy to hear them. 226 00:19:58,880 --> 00:20:02,140 But then we can pay forward. 227 00:20:02,140 --> 00:20:08,240 From Kimmy, as you move forward using this data, I think something to really consider 228 00:20:08,240 --> 00:20:14,400 and something that I'm sure you probably already have is that there's a huge difference from 229 00:20:14,400 --> 00:20:20,400 state to state in all of these metrics. 230 00:20:20,400 --> 00:20:27,280 And every state uses different metrics to decide what products to make and how their 231 00:20:27,280 --> 00:20:29,720 patients are feeling about it. 232 00:20:29,720 --> 00:20:36,960 And just I think it's really important that any analysis that's done is done on, you know, 233 00:20:36,960 --> 00:20:43,800 like a sample of a certain state's data as opposed to a national sample. 234 00:20:43,800 --> 00:20:46,400 I definitely agree with you. 235 00:20:46,400 --> 00:20:53,400 And it's pretty easy to find the state variances, right? 236 00:20:53,400 --> 00:20:56,600 They vary on many different metrics. 237 00:20:56,600 --> 00:21:01,320 So you're 100% correct. 238 00:21:01,320 --> 00:21:09,360 I guess the two things I would say is one, I guess, please just, I guess, acknowledge 239 00:21:09,360 --> 00:21:11,000 the shortcomings of this. 240 00:21:11,000 --> 00:21:14,440 So please take everything as a grain of salt. 241 00:21:14,440 --> 00:21:17,120 Hopefully the statistics can still be reused, right? 242 00:21:17,120 --> 00:21:22,160 You can still use the same statistics with your better data. 243 00:21:22,160 --> 00:21:29,200 So that's sort of the idea is, you know, the statistical models can be used as tools, but 244 00:21:29,200 --> 00:21:33,040 they're only as good, your outputs is only as good as the data. 245 00:21:33,040 --> 00:21:40,800 So we'll just keep in mind that this data is, we'll just assume it's the worst. 246 00:21:40,800 --> 00:21:44,240 So let's not weigh in too, too much. 247 00:21:44,240 --> 00:21:49,520 Or let's just not really weigh our results at all. 248 00:21:49,520 --> 00:21:56,760 So this is more just a demonstration of the statistics versus trying to draw conclusions, 249 00:21:56,760 --> 00:21:58,260 what we're doing today. 250 00:21:58,260 --> 00:22:05,080 And then second, I think, so I think that's all I think I have to say, other than I think 251 00:22:05,080 --> 00:22:12,720 you're spot on in that if you did have a review data set, and you knew where the reviews were 252 00:22:12,720 --> 00:22:18,080 coming from, so you knew like sort of the user location, so you knew like state by state, 253 00:22:18,080 --> 00:22:21,240 then that would be incredibly interesting. 254 00:22:21,240 --> 00:22:25,400 And then, so for example, we're about to do sentiment analysis. 255 00:22:25,400 --> 00:22:32,840 And so it'd be interesting to do a sentiment analysis of say, jungle boys, state by state, 256 00:22:32,840 --> 00:22:39,840 so that way you could actually kind of parse out if sentiment's different than Florida 257 00:22:39,840 --> 00:22:45,400 versus on the West Coast. 258 00:22:45,400 --> 00:22:51,360 That's actually a brilliant thought that you brought up. 259 00:22:51,360 --> 00:22:58,800 So as I actually like to, I kind of realized this and kind of want to drive this home is 260 00:22:58,800 --> 00:23:06,240 all of your questions, comments, thoughts, ideas, these all help move the ball forward. 261 00:23:06,240 --> 00:23:19,760 So you know, I may be punching keys, but that's only part of the picture. 262 00:23:19,760 --> 00:23:23,920 So you can't complete the puzzle without every piece. 263 00:23:23,920 --> 00:23:29,480 And so all the questions, all the feedback, all the listening ears, those are all critical 264 00:23:29,480 --> 00:23:30,480 pieces. 265 00:23:30,480 --> 00:23:32,880 Okay, cool. 266 00:23:32,880 --> 00:23:38,080 So without droning on, let's just go ahead and get into it because I think the narrative 267 00:23:38,080 --> 00:23:43,800 here is almost more interesting than the data, but I'll at least show you how the statistics 268 00:23:43,800 --> 00:23:44,800 is done. 269 00:23:44,800 --> 00:23:53,880 But long story short, you can aggregate all the reviews for all the different users, some 270 00:23:53,880 --> 00:23:54,880 of them are anonymous. 271 00:23:54,880 --> 00:24:05,880 So you know, luckily, they've used pretty, I guess, anonymous names. 272 00:24:05,880 --> 00:24:10,760 Right, we don't really want to be dealing with actual people's names here. 273 00:24:10,760 --> 00:24:22,480 But I guess we'll exclude the anonymous ones and just do a little bit of cleaning up on, 274 00:24:22,480 --> 00:24:26,040 maybe just do a little bit of cleaning up on the reviews. 275 00:24:26,040 --> 00:24:36,280 And basically, what I've done here is basically for each user, I've just taken five random 276 00:24:36,280 --> 00:24:37,400 reviews. 277 00:24:37,400 --> 00:24:46,640 So the idea is, you can kind of see, start to parse out if there's any differences by 278 00:24:46,640 --> 00:24:47,640 user, right? 279 00:24:47,640 --> 00:24:55,200 And so the idea is, if every user is exactly the same, it's probably not going to be meaningful 280 00:24:55,200 --> 00:24:57,560 to do statistics with. 281 00:24:57,560 --> 00:25:02,000 And so that's something a professor told me is, right, you kind of want to look for variants, 282 00:25:02,000 --> 00:25:03,000 right? 283 00:25:03,000 --> 00:25:08,320 So the more variants you see, the more applicable statistics are. 284 00:25:08,320 --> 00:25:15,560 So we can see, okay, this is pretty simple, right? 285 00:25:15,560 --> 00:25:21,080 Different users use different amounts of words. 286 00:25:21,080 --> 00:25:28,640 And so the idea is, you can start to find little intricacies like this. 287 00:25:28,640 --> 00:25:41,840 And then it's essentially many of these facets that are being used in these AI models. 288 00:25:41,840 --> 00:25:47,840 So I was looking at their paper, and they basically just use hundreds of metrics like 289 00:25:47,840 --> 00:25:48,840 this. 290 00:25:48,840 --> 00:26:00,360 So they just use things like word count, just everything under the sun, sentence length, 291 00:26:00,360 --> 00:26:02,640 grammar, word choice. 292 00:26:02,640 --> 00:26:05,360 So they just use all of these metrics. 293 00:26:05,360 --> 00:26:07,680 And then they have a training data set. 294 00:26:07,680 --> 00:26:12,600 So they have someone who wrote an essay and took a personality test. 295 00:26:12,600 --> 00:26:17,960 And then they just basically fit a really, really good regression model. 296 00:26:17,960 --> 00:26:25,480 So they just basically try a bunch of these factors and just see how this helps predict 297 00:26:25,480 --> 00:26:28,640 personality. 298 00:26:28,640 --> 00:26:37,040 So that's sort of what's going on under the hood is, not only do they look at word count, 299 00:26:37,040 --> 00:26:44,920 but this is just basically a metric, and then this is used for prediction purposes. 300 00:26:44,920 --> 00:26:50,800 But before we get to the abstract, let's look at a few more concrete things we can do with 301 00:26:50,800 --> 00:26:52,280 this. 302 00:26:52,280 --> 00:26:59,440 So the idea is, okay, well, you can take out... 303 00:26:59,440 --> 00:27:03,080 So these are these people's essays. 304 00:27:03,080 --> 00:27:12,560 So you can take out all of the filler words, so all the this, the, a, is... 305 00:27:12,560 --> 00:27:20,640 And then what you're left with are essentially all the words that different users chose to 306 00:27:20,640 --> 00:27:22,360 use. 307 00:27:22,360 --> 00:27:26,920 And this is, once again, quite fruitful because... 308 00:27:26,920 --> 00:27:31,160 So this is, I think, kind of where the personality comes from is, right? 309 00:27:31,160 --> 00:27:37,120 There's sort of a correlation between the words people use and perhaps the personality. 310 00:27:37,120 --> 00:27:41,200 But that's sort of rushing ahead way to the end. 311 00:27:41,200 --> 00:27:46,920 The idea now is, what's just sort of the positivity? 312 00:27:46,920 --> 00:27:48,920 So you can kind of see... 313 00:27:48,920 --> 00:27:49,920 Question? 314 00:27:49,920 --> 00:27:59,280 Hey, Keegan, are these essays all for this, written for the same prompts? 315 00:27:59,280 --> 00:28:05,560 Well, the training data set, you mean? 316 00:28:05,560 --> 00:28:11,120 Well, the word count comes from essays, right? 317 00:28:11,120 --> 00:28:14,400 What's the prompt for the essays? 318 00:28:14,400 --> 00:28:18,360 In this case, my essay are strain reviews. 319 00:28:18,360 --> 00:28:22,560 So I just compiled five strain reviews. 320 00:28:22,560 --> 00:28:28,760 So these aren't actual essays. 321 00:28:28,760 --> 00:28:35,440 I'm just more thinking about this is like a corpus of somebody's work. 322 00:28:35,440 --> 00:28:41,720 So people are writing a page and a half with your work of strain review? 323 00:28:41,720 --> 00:28:44,360 That's pretty amazing to me. 324 00:28:44,360 --> 00:28:46,800 I mean, some people like to talk. 325 00:28:46,800 --> 00:28:47,800 Geez. 326 00:28:47,800 --> 00:28:53,000 Actually, we'll actually have those exact statistics here in one second. 327 00:28:53,000 --> 00:29:03,760 And so the idea is perhaps if somebody's writing longer reviews about your strain, that may 328 00:29:03,760 --> 00:29:05,760 be a good sign. 329 00:29:05,760 --> 00:29:09,880 But we can start to tie it all together. 330 00:29:09,880 --> 00:29:16,040 But here, I'll hurry up and get to the strains since I think maybe the more interesting part 331 00:29:16,040 --> 00:29:17,040 here. 332 00:29:17,040 --> 00:29:25,120 But basically, the one little nugget here that I found was, OK, so just to kind of show 333 00:29:25,120 --> 00:29:35,200 you how this works, you can kind of just do a polarity score of different words. 334 00:29:35,200 --> 00:29:42,840 The best is like, so best is positive. 335 00:29:42,840 --> 00:29:50,680 And then I think the idea is the very best. 336 00:29:50,680 --> 00:29:58,760 The algorithm can essentially kind of determine that the very best is more positive than the 337 00:29:58,760 --> 00:30:00,600 best. 338 00:30:00,600 --> 00:30:02,600 And the very worst. 339 00:30:02,600 --> 00:30:06,040 Let's see if they can handle that. 340 00:30:06,040 --> 00:30:08,240 And so then, oh, this is the very worst. 341 00:30:08,240 --> 00:30:11,940 OK, you get a negative score. 342 00:30:11,940 --> 00:30:16,800 So once again, we don't want to reinvent the wheel. 343 00:30:16,800 --> 00:30:21,680 So this is a point where we're going to stand on the shoulders of giants. 344 00:30:21,680 --> 00:30:27,000 We're going to essentially have to assume that the people who've done this work on the 345 00:30:27,000 --> 00:30:31,760 polarity scores did a rigorous job. 346 00:30:31,760 --> 00:30:36,760 And from briefly looking under the hood, it looks pretty rigorous. 347 00:30:36,760 --> 00:30:39,200 It looks like there was... 348 00:30:39,200 --> 00:30:48,120 I don't want to speak too much about it, but essentially it looks like a lot of psychologists 349 00:30:48,120 --> 00:30:50,000 have sort of been working on this. 350 00:30:50,000 --> 00:30:54,200 But once again, if this is something that you're particularly interested in, please 351 00:30:54,200 --> 00:30:58,760 poke under the hood at that point. 352 00:30:58,760 --> 00:31:01,760 Is there a question? 353 00:31:01,760 --> 00:31:04,560 Yeah, hi. 354 00:31:04,560 --> 00:31:09,840 Not to go too deep under the hood, but doesn't context matter when writing? 355 00:31:09,840 --> 00:31:22,400 I'm sorry, doesn't context matter with the polarity scores? 356 00:31:22,400 --> 00:31:25,600 Can you speak a little more to that? 357 00:31:25,600 --> 00:31:26,600 In case you use... 358 00:31:26,600 --> 00:31:29,200 Like it's not the very worst? 359 00:31:29,200 --> 00:31:30,960 Yeah, I'm not really... 360 00:31:30,960 --> 00:31:39,520 I mean, from the keywords you had selected, it looks like classic, muddy, other words, 361 00:31:39,520 --> 00:31:51,160 they could be used in different contexts to give a different score, that I would believe. 362 00:31:51,160 --> 00:31:56,080 Is that not something that could change the data at least? 363 00:31:56,080 --> 00:32:00,920 It sounds like that lack of specificity could be adding noise to your data, Emmanuel. 364 00:32:00,920 --> 00:32:03,120 I'm not sure if that's what you were trying to say. 365 00:32:03,120 --> 00:32:10,040 Yeah, is that something that we should be giving it a like? 366 00:32:10,040 --> 00:32:13,840 If I'm interpreting this correctly, the way I would... 367 00:32:13,840 --> 00:32:20,200 Sorry, basically saying a user may use the word gas to mean good, but that's not captured 368 00:32:20,200 --> 00:32:21,600 in the data. 369 00:32:21,600 --> 00:32:22,600 Exactly. 370 00:32:22,600 --> 00:32:23,600 Exactly. 371 00:32:23,600 --> 00:32:28,960 And so I think this is where you can basically enhance the model. 372 00:32:28,960 --> 00:32:32,560 So I do believe maybe... 373 00:32:32,560 --> 00:32:38,040 You may need to dig into the documentation, but I would like to believe that this is sort 374 00:32:38,040 --> 00:32:40,840 of the baseline. 375 00:32:40,840 --> 00:32:51,000 I think it's just using a baseline word mapping that's pretty basic, like maybe 16,000 words 376 00:32:51,000 --> 00:32:56,920 where they've got best and good and all the real rudimentary words. 377 00:32:56,920 --> 00:33:04,120 And I think you could essentially improve upon it somehow if you had your own mapping 378 00:33:04,120 --> 00:33:09,400 between gas is a good word in the cannabis space. 379 00:33:09,400 --> 00:33:14,200 If I may interject? 380 00:33:14,200 --> 00:33:22,000 I think in trying to get that more specific sample, that's when you would be really digging 381 00:33:22,000 --> 00:33:25,280 in deeper to see what words mean what. 382 00:33:25,280 --> 00:33:27,080 I think that's where you're going to be getting that context. 383 00:33:27,080 --> 00:33:31,720 You're going to have to look at each sample differently because some people might be using 384 00:33:31,720 --> 00:33:32,720 the word gas. 385 00:33:32,720 --> 00:33:35,320 Some people may be using the word fire. 386 00:33:35,320 --> 00:33:38,200 A word for bad out here on the East Coast is boof. 387 00:33:38,200 --> 00:33:40,000 I don't know if people use that. 388 00:33:40,000 --> 00:33:44,680 So it's something that I think you'll really have to get into when you're looking at your 389 00:33:44,680 --> 00:33:45,680 data set. 390 00:33:45,680 --> 00:33:50,520 And then you're just going to have to dig into the data and see what words you're actually 391 00:33:50,520 --> 00:33:51,520 seeing. 392 00:33:51,520 --> 00:33:53,320 Spot on. 393 00:33:53,320 --> 00:34:02,800 And I may have to refer to the prior week or I'll have to share some code with you because 394 00:34:02,800 --> 00:34:08,800 you can find the most frequent words that are used. 395 00:34:08,800 --> 00:34:15,400 So a strategy would be, okay, what words are people frequently using? 396 00:34:15,400 --> 00:34:21,840 And then you may want to go and kind of code those positive or negative. 397 00:34:21,840 --> 00:34:24,260 And then see if you can't supplement your algorithm. 398 00:34:24,260 --> 00:34:30,920 So that way if people are using the word fire, gas, or lit, or what have you, some of this 399 00:34:30,920 --> 00:34:42,080 slang in the cannabis space, it may be worth your while to hand code 100 words or what 400 00:34:42,080 --> 00:34:43,200 have you. 401 00:34:43,200 --> 00:34:46,240 But I guess it's just sort of a cost benefit thing. 402 00:34:46,240 --> 00:34:53,240 You can always keep making your model better, but I guess there's only a marginal benefit 403 00:34:53,240 --> 00:34:55,360 and there will be a marginal cost. 404 00:34:55,360 --> 00:34:57,400 Hey, Keegan. 405 00:34:57,400 --> 00:34:58,400 Please. 406 00:34:58,400 --> 00:35:09,280 Along this line, why don't you quantify a couple of words that are recognized kind of 407 00:35:09,280 --> 00:35:10,640 culturally? 408 00:35:10,640 --> 00:35:15,720 And yeah, boof is an interesting word. 409 00:35:15,720 --> 00:35:23,040 It may, I think it typifies kind of not so good attenuated cannabis. 410 00:35:23,040 --> 00:35:28,260 It's going to be tough maybe to link that into the strain reviews and the content. 411 00:35:28,260 --> 00:35:34,240 But words like dank, words like gas could easily correlate with strains. 412 00:35:34,240 --> 00:35:36,300 We kind of know what that is. 413 00:35:36,300 --> 00:35:42,600 So if you simply did a word, a review count that includes words like dank and gas and 414 00:35:42,600 --> 00:35:49,800 see if it maps to what we know to be the strains associated with that, that would be an interesting 415 00:35:49,800 --> 00:35:54,120 initial validation of this approach. 416 00:35:54,120 --> 00:35:59,480 I think we can do it pretty quickly here. 417 00:35:59,480 --> 00:36:03,480 Let's just do it. 418 00:36:03,480 --> 00:36:11,880 And in fact, you should maybe do dank versus gas and see what's used more. 419 00:36:11,880 --> 00:36:20,720 I think at that point we would be answering a different question, but I'm all with it. 420 00:36:20,720 --> 00:36:26,320 So maybe the question is slightly different, but I'm trying to approach how to validate 421 00:36:26,320 --> 00:36:31,600 a language processing on a set of reviews. 422 00:36:31,600 --> 00:36:38,640 And I go immediately to do strain specific review words like what we just mentioned, 423 00:36:38,640 --> 00:36:43,400 correlate with strains that we know would be in that bucket. 424 00:36:43,400 --> 00:36:46,120 I hope that's clear. 425 00:36:46,120 --> 00:36:51,800 I think you're on to something, John, and this is what I mean by every time you go panning 426 00:36:51,800 --> 00:36:56,160 for gold, you almost can't help but walk away with something. 427 00:36:56,160 --> 00:36:57,680 So I think you're right. 428 00:36:57,680 --> 00:37:04,360 I think if you essentially looked at the most frequent words, you've kind of got to filter 429 00:37:04,360 --> 00:37:10,720 out with just some that bubble to the top like great and strain. 430 00:37:10,720 --> 00:37:13,960 But I think there's something to that. 431 00:37:13,960 --> 00:37:22,400 And that's specifically what the personality prediction algorithm, that's one of the key 432 00:37:22,400 --> 00:37:25,520 tools that it leverages. 433 00:37:25,520 --> 00:37:32,480 I think the psychologists do believe there's a systemic difference between the word choice 434 00:37:32,480 --> 00:37:38,080 that people use. 435 00:37:38,080 --> 00:37:44,440 And if they use different words for different strains, then that would be quite defined. 436 00:37:44,440 --> 00:37:49,520 But here, I'll go ahead and get to the strains since I think that's the more interesting 437 00:37:49,520 --> 00:37:50,520 part here. 438 00:37:50,520 --> 00:38:00,200 But the idea is the application for this would be if you were going to say you were running 439 00:38:00,200 --> 00:38:07,800 a dispensary, you had a bunch of users, well, you could basically try to ask for their reviews 440 00:38:07,800 --> 00:38:12,600 and the idea is you could see if somebody left a review. 441 00:38:12,600 --> 00:38:16,040 So here somebody just leaves a random review. 442 00:38:16,040 --> 00:38:26,160 Well, instead of just saying, oh, is this a good or a bad review, you can actually measure 443 00:38:26,160 --> 00:38:37,360 that and you can say that this actually wasn't actually the best review in the world. 444 00:38:37,360 --> 00:38:46,400 So here this is a long one, but as far as intensity goes, it's sort of in the 23rd percentile. 445 00:38:46,400 --> 00:38:54,200 And if you just start to pick at this, it didn't affect my mood at all. 446 00:38:54,200 --> 00:39:02,000 I explored sativas and I still do, but I need a good indica. 447 00:39:02,000 --> 00:39:04,640 Whatever that means. 448 00:39:04,640 --> 00:39:07,760 So we may need to find it different. 449 00:39:07,760 --> 00:39:14,360 But the long story short is, let's see if we, so this looks like a positive one. 450 00:39:14,360 --> 00:39:23,360 You can basically determine, and once again, take this as all a grain of salt. 451 00:39:23,360 --> 00:39:29,560 It's not the best data, it may not be the best data in the world, but the idea is if 452 00:39:29,560 --> 00:39:35,600 you've got a corpus of reviews, you can then start to rank them. 453 00:39:35,600 --> 00:39:44,880 So these actually may be user specific, but you can basically find out is this user above 454 00:39:44,880 --> 00:39:54,840 average happy or less than average happy. 455 00:39:54,840 --> 00:40:10,440 And then the real quick application is basically the idea is, probably a lot of people are 456 00:40:10,440 --> 00:40:14,600 going to be happy with this, but the idea is if you're running a company, you want to 457 00:40:14,600 --> 00:40:20,360 identify people who are unhappy and see if you can't rectify that. 458 00:40:20,360 --> 00:40:22,600 So here you've got a bunch of reviews. 459 00:40:22,600 --> 00:40:35,240 Well, here you can basically pick out the most negative review, if that was something 460 00:40:35,240 --> 00:40:36,240 you wished. 461 00:40:36,240 --> 00:40:40,120 Let's see if we can't print this out somehow. 462 00:40:40,120 --> 00:40:51,440 Okay, so here's the most negative review. 463 00:40:51,440 --> 00:40:59,840 And just from reading through this, this is almost something that if you're a dispensary, 464 00:40:59,840 --> 00:41:04,440 if you were able to catch this and get back in touch with this person, you may be able 465 00:41:04,440 --> 00:41:10,520 to salvage their customer relationship. 466 00:41:10,520 --> 00:41:27,160 Because it sounds like they maybe had a mediocre time, got me slightly high, lasted less than 467 00:41:27,160 --> 00:41:33,840 an hour, great pain relief, but gives me bad munchies. 468 00:41:33,840 --> 00:41:44,400 So if for whatever reason you're able to, I don't know, provide a strain that provides 469 00:41:44,400 --> 00:41:49,280 less munchies, which more research would need to be done there, then perhaps you could help 470 00:41:49,280 --> 00:41:52,120 this consumer out. 471 00:41:52,120 --> 00:41:59,120 But I'll quit worrying with that because I think this is actually the more interesting 472 00:41:59,120 --> 00:42:02,840 part since this is where most of your questions have been so far. 473 00:42:02,840 --> 00:42:04,560 So that's cool. 474 00:42:04,560 --> 00:42:05,720 We've done that. 475 00:42:05,720 --> 00:42:10,760 You can group reviews by user. 476 00:42:10,760 --> 00:42:16,640 Well now we can actually group reviews by strain. 477 00:42:16,640 --> 00:42:25,320 And we actually have an enormous amount of reviews for some of these strains, I realize. 478 00:42:25,320 --> 00:42:28,000 So some we may not have as many. 479 00:42:28,000 --> 00:42:33,960 So just for this analysis, I don't know if this is pertinent to do or not, so I'll let 480 00:42:33,960 --> 00:42:42,120 you be the judge, but I restricted the analysis to those that have more than 30 reviews, and 481 00:42:42,120 --> 00:42:48,680 then I selected 20 random reviews of those. 482 00:42:48,680 --> 00:42:56,720 So the idea is that way we're kind of comparing comparable bodies of work. 483 00:42:56,720 --> 00:43:01,920 That way the corpuses are about the same length. 484 00:43:01,920 --> 00:43:07,400 I don't know if that matters or not, but I just kind of was thinking that you may get 485 00:43:07,400 --> 00:43:16,040 a different reading from 600 reviews than you would from 20. 486 00:43:16,040 --> 00:43:22,120 So please explore to see if this is pertinent to do or not. 487 00:43:22,120 --> 00:43:30,120 But the idea is, okay, just like before, we can basically create essays for all the different 488 00:43:30,120 --> 00:43:31,680 strains. 489 00:43:31,680 --> 00:43:38,560 So these are basically 20 random reviews for these strains. 490 00:43:38,560 --> 00:43:46,280 And so then, well, just like we said before, you can actually start to see which strains 491 00:43:46,280 --> 00:43:50,560 people are talking more about. 492 00:43:50,560 --> 00:44:01,160 And so this just was sort of a fun idea, but one of the effects was talkative. 493 00:44:01,160 --> 00:44:09,600 So it would just be kind of funny to see if the effect talkative had any correlation to 494 00:44:09,600 --> 00:44:16,080 the number of words in a strain's review. 495 00:44:16,080 --> 00:44:25,680 And then now we can get to the fun part is you can actually look at all the words that 496 00:44:25,680 --> 00:44:28,960 people are using for different strains. 497 00:44:28,960 --> 00:44:34,960 And so, John, this is actually right up your alley where you can start to see are people 498 00:44:34,960 --> 00:44:44,200 using gas or fuel a lot for particular varieties. 499 00:44:44,200 --> 00:44:47,280 And today we're just sort of looking at positive, negative. 500 00:44:47,280 --> 00:44:58,800 And so what you can do is you can actually rank the reviews for these strains from positive 501 00:44:58,800 --> 00:45:00,360 to negative. 502 00:45:00,360 --> 00:45:09,260 And so what this got me thinking about is you have these cannabis cups where everybody 503 00:45:09,260 --> 00:45:12,280 brings in their variety. 504 00:45:12,280 --> 00:45:19,760 I haven't ever been to one, so I'm just kind of speaking from what I'm imagining that they 505 00:45:19,760 --> 00:45:21,120 would be like. 506 00:45:21,120 --> 00:45:27,240 But I imagine people bring in different varieties and a lot of people try them and they vote 507 00:45:27,240 --> 00:45:30,000 on the best one. 508 00:45:30,000 --> 00:45:36,620 Well the idea here is now you don't even necessarily have to have people vote because that may 509 00:45:36,620 --> 00:45:40,300 even be sort of an imperfect measure. 510 00:45:40,300 --> 00:45:46,080 If you just ask somebody to rate a strain one to five, they may just give everything 511 00:45:46,080 --> 00:45:48,000 a five. 512 00:45:48,000 --> 00:45:50,640 So that may not even be the best measure. 513 00:45:50,640 --> 00:45:57,160 But what you could do instead is you could have everybody write a review about the strain 514 00:45:57,160 --> 00:46:05,560 and then you could see how positive they are and then you could see who the winner is. 515 00:46:05,560 --> 00:46:09,880 Drum roll please. 516 00:46:09,880 --> 00:46:27,000 And so with the random sampling, we have the strain with the most positive reviews being 517 00:46:27,000 --> 00:46:30,360 conspiracy kush. 518 00:46:30,360 --> 00:46:41,080 I just wanted to show you if you don't do the, so you may actually just want to actually 519 00:46:41,080 --> 00:46:45,520 just see which strains are maybe just in the top percentile. 520 00:46:45,520 --> 00:46:47,760 So let's see if we can't do that real quick. 521 00:46:47,760 --> 00:46:54,360 Just say which strain is greater than the quantile 90th percentile. 522 00:46:54,360 --> 00:46:57,920 I think this will work. 523 00:46:57,920 --> 00:46:58,920 Cool. 524 00:46:58,920 --> 00:47:10,200 So now you can basically get your top 10 favorite strains out there. 525 00:47:10,200 --> 00:47:18,640 As I said, this is the first time I put this list together, but I think this is valuable. 526 00:47:18,640 --> 00:47:31,880 So if you're going to go grow cannabis, maybe you want to start with super blue, super skunk 527 00:47:31,880 --> 00:47:39,200 or blueberry cheesecake, chocolate, lemon. 528 00:47:39,200 --> 00:47:40,800 What if it's supposed to be lemon? 529 00:47:40,800 --> 00:47:48,840 But you may want to see if you can't start with the strain that people already kind of 530 00:47:48,840 --> 00:47:50,880 like. 531 00:47:50,880 --> 00:47:53,160 Cool. 532 00:47:53,160 --> 00:48:06,560 So that'll sort of tie us in nicely to the last part here and then we'll get back to 533 00:48:06,560 --> 00:48:09,320 the cannabis industry. 534 00:48:09,320 --> 00:48:17,000 So once again, just coming at this from a statistical lens with full knowledge that 535 00:48:17,000 --> 00:48:25,680 this data set one is imperfect, two, it doesn't take into consideration the location of the 536 00:48:25,680 --> 00:48:29,640 users and all of this. 537 00:48:29,640 --> 00:48:35,580 Many, many potentials for downsides, but the idea is if you've got your own body of reviews, 538 00:48:35,580 --> 00:48:39,560 you can hopefully use these exact same statistics. 539 00:48:39,560 --> 00:48:40,640 Cool. 540 00:48:40,640 --> 00:48:53,160 So first things first, just going to rate every single review from positive to negative. 541 00:48:53,160 --> 00:49:03,600 This is something that, well, I guess we haven't really proven that it works yet, but we may 542 00:49:03,600 --> 00:49:09,160 never really prove that it works, but the idea is hopefully this can be a tool for you 543 00:49:09,160 --> 00:49:10,160 to use. 544 00:49:10,160 --> 00:49:17,120 So if there's any way that you can use these tools to make better business decisions or 545 00:49:17,120 --> 00:49:25,200 decisions as a consumer, then more power to you in my book. 546 00:49:25,200 --> 00:49:32,760 But the idea is, okay, we can actually see if there are any reviews here for the Jungle 547 00:49:32,760 --> 00:49:33,760 Boys. 548 00:49:33,760 --> 00:49:38,120 And pardon me that I didn't realize that it was with an S. For some reason I thought it 549 00:49:38,120 --> 00:49:43,840 was with a Z, but I've got multiple spellings here. 550 00:49:43,840 --> 00:49:46,720 Just pardon the code. 551 00:49:46,720 --> 00:49:57,320 But the idea is we've got tons of reviews, so it would be nice if, so this isn't probably 552 00:49:57,320 --> 00:50:01,840 a large enough subsample to draw any meaningful conclusion. 553 00:50:01,840 --> 00:50:05,360 We only have seven samples. 554 00:50:05,360 --> 00:50:14,200 But we do have seven reviews where people specifically mentioned the Jungle Boys. 555 00:50:14,200 --> 00:50:21,280 So some gelato, orange cookies, and wedding cake. 556 00:50:21,280 --> 00:50:31,120 And once again, don't read too much into this because I think you would have to do a difference 557 00:50:31,120 --> 00:50:43,200 of means tests to see if, what I'm going to do here is we've got our rankings of all the 558 00:50:43,200 --> 00:50:46,080 reviews. 559 00:50:46,080 --> 00:50:52,400 These are all the reviews positive to negative. 560 00:50:52,400 --> 00:50:59,280 And the average is around 0.67. 561 00:50:59,280 --> 00:51:09,920 So the nominal value, like where you actually are, in my opinion doesn't really matter too, 562 00:51:09,920 --> 00:51:11,000 too much, right? 563 00:51:11,000 --> 00:51:15,280 Because most people are having a positive experience. 564 00:51:15,280 --> 00:51:22,360 The idea is are you above average or below average? 565 00:51:22,360 --> 00:51:36,000 So basically what you can do is you can basically calculate the average intensity of the Jungle 566 00:51:36,000 --> 00:51:37,920 Boys reviews. 567 00:51:37,920 --> 00:51:42,440 So here are their products. 568 00:51:42,440 --> 00:51:47,680 You see this one really high. 569 00:51:47,680 --> 00:51:51,120 This one really low, right? 570 00:51:51,120 --> 00:51:54,920 0.51 is below average. 571 00:51:54,920 --> 00:51:59,400 But the idea is, oh, and look at this one, 0.32. 572 00:51:59,400 --> 00:52:02,960 That one's well below average. 573 00:52:02,960 --> 00:52:09,340 But the idea is if you were going to take the average of all the reviews, and so if 574 00:52:09,340 --> 00:52:18,680 you, as N goes large, you want your sample size to get large. 575 00:52:18,680 --> 00:52:24,960 And I'll back it out because conditions matter. 576 00:52:24,960 --> 00:52:27,520 And what do I mean by conditions matter? 577 00:52:27,520 --> 00:52:33,760 Well, what if there's a difference, a systemic difference by state? 578 00:52:33,760 --> 00:52:41,680 So what if the people that gave the low reviews, what if they're in a particular state? 579 00:52:41,680 --> 00:52:46,000 And the people who gave the high reviews are in a different state? 580 00:52:46,000 --> 00:52:51,800 Well, that would have potentially meaningful implications. 581 00:52:51,800 --> 00:52:57,360 So as Kimberly pointed out, that matters. 582 00:52:57,360 --> 00:52:59,080 We don't have that data. 583 00:52:59,080 --> 00:53:06,760 And we're just taking an unconditional average, which as I pointed out, hopefully in many 584 00:53:06,760 --> 00:53:14,280 meetups before, the more conditions you can add on your averages, typically the more interesting 585 00:53:14,280 --> 00:53:15,640 your statistics are. 586 00:53:15,640 --> 00:53:22,440 I think you could write paper upon paper just by taking conditional average after conditional 587 00:53:22,440 --> 00:53:26,520 average and then doing a difference in means. 588 00:53:26,520 --> 00:53:32,520 There's no need to go do machine learning models when you can just do a difference in 589 00:53:32,520 --> 00:53:36,120 means test, in my opinion. 590 00:53:36,120 --> 00:53:42,960 But I don't know why I come at it with that perspective. 591 00:53:42,960 --> 00:53:45,480 But I do. 592 00:53:45,480 --> 00:53:55,600 But long story short is, if you were going to take the average sentiment of the Jungle 593 00:53:55,600 --> 00:54:02,400 Boys reviews, at first glance, it does look like they may be above average. 594 00:54:02,400 --> 00:54:05,320 As I was saying, we only have a sample size of seven. 595 00:54:05,320 --> 00:54:11,320 So we probably couldn't actually draw any statistical significance from that. 596 00:54:11,320 --> 00:54:18,280 But the idea is you'd want to get as many reviews as you can, get as good quality reviews 597 00:54:18,280 --> 00:54:25,480 as you can, and repeat the analysis yourself. 598 00:54:25,480 --> 00:54:29,360 So that's what I had to say about the statistics. 599 00:54:29,360 --> 00:54:37,720 Next week, we'll finally get to personality, maybe look at some other applications of natural 600 00:54:37,720 --> 00:54:38,720 language processing. 601 00:54:38,720 --> 00:54:44,680 Because as I said, there are some other ones we can do here. 602 00:54:44,680 --> 00:54:48,920 The aspect of extraction is where we basically find effects. 603 00:54:48,920 --> 00:54:56,800 We can find the headache, whether it cured a headache or caused a headache. 604 00:54:56,800 --> 00:55:00,480 The personality prediction, I think, is going to be cool. 605 00:55:00,480 --> 00:55:09,320 Depression, I think, is going to be really cool to study, because that's something that 606 00:55:09,320 --> 00:55:16,760 people even talk about as one of the reasons why they may use cannabis. 607 00:55:16,760 --> 00:55:20,200 They may be seeking to alleviate depression. 608 00:55:20,200 --> 00:55:30,480 So what would be incredibly helpful, interesting, would be to if you had a body of reviews by 609 00:55:30,480 --> 00:55:38,480 a user that are time stamped, then it would be interesting to see if their depression 610 00:55:38,480 --> 00:55:42,000 is changing over time. 611 00:55:42,000 --> 00:55:49,240 So maybe their reviews may indicate that they're really, really depressed. 612 00:55:49,240 --> 00:55:55,720 One, maybe their depression may wane, or maybe their depression may increase. 613 00:55:55,720 --> 00:56:01,600 The idea is if there's any way you had access to user reviews that are time stamped, say 614 00:56:01,600 --> 00:56:09,880 you're a dispensary, then one, you could monitor to see if any user is becoming unusually 615 00:56:09,880 --> 00:56:15,880 depressed, which that may be a, you wouldn't want that. 616 00:56:15,880 --> 00:56:23,400 And then two, if somebody is actually getting depression, they're alleviated, you may want 617 00:56:23,400 --> 00:56:31,600 to see what strains they're using, how much are they consuming, how are they using cannabis 618 00:56:31,600 --> 00:56:34,400 to help them out in their life. 619 00:56:34,400 --> 00:56:42,280 So I think that's an incredibly, incredibly cool one to study. 620 00:56:42,280 --> 00:56:43,280 Engagement measures. 621 00:56:43,280 --> 00:56:46,200 What reviews are you using here? 622 00:56:46,200 --> 00:56:50,680 Is this what we call the 42,000 set? 623 00:56:50,680 --> 00:56:51,680 Yes. 624 00:56:51,680 --> 00:56:52,680 Yes. 625 00:56:52,680 --> 00:56:57,480 So the one we've been working with for a number of weeks. 626 00:56:57,480 --> 00:56:58,480 Exactly. 627 00:56:58,480 --> 00:57:02,840 I think we should now call it like 22,000 set. 628 00:57:02,840 --> 00:57:11,080 But that's what I was saying is like, almost like every time we go out panning for gold, 629 00:57:11,080 --> 00:57:16,720 it's like there's just always like, and I'm saying like, this isn't necessarily gold because 630 00:57:16,720 --> 00:57:21,320 this isn't statistically significant, but I'm just saying it's a demonstration of how 631 00:57:21,320 --> 00:57:24,600 you can use the model to find gold. 632 00:57:24,600 --> 00:57:29,400 So I think that's gold in and of itself. 633 00:57:29,400 --> 00:57:35,200 But the point being is it's almost like every time you go and pan through this cannabis 634 00:57:35,200 --> 00:57:42,960 data, it's almost like you can't help but discover these interesting ways to look at 635 00:57:42,960 --> 00:57:47,800 the data. 636 00:57:47,800 --> 00:57:56,560 So can I make a, have you looked at the username count? 637 00:57:56,560 --> 00:58:05,480 In other words, most of the, you have, if I remember correctly, about 350 anonymous 638 00:58:05,480 --> 00:58:07,200 that you throw out. 639 00:58:07,200 --> 00:58:12,520 And then as you move down, the next most frequent is around 70 odd. 640 00:58:12,520 --> 00:58:18,640 And then there's a whole series of 60s and 50 reviews. 641 00:58:18,640 --> 00:58:22,200 And then of course it goes all the way down to ones. 642 00:58:22,200 --> 00:58:31,400 And my sense is what one could make a good living with is look at, take maybe the top 643 00:58:31,400 --> 00:58:40,480 20 most frequent reviewers and look at that against the top 10 most frequent strains or 644 00:58:40,480 --> 00:58:46,640 something and see what kind of a living you can make. 645 00:58:46,640 --> 00:58:51,480 It might be better to subset it and work accordingly. 646 00:58:51,480 --> 00:58:53,400 And you're definitely right, John. 647 00:58:53,400 --> 00:59:02,600 And the way I would describe this is there may be some knowledge hidden in these consumers 648 00:59:02,600 --> 00:59:04,120 that respond a lot. 649 00:59:04,120 --> 00:59:10,800 So if they just leave one review, it's hard to really know what to make of that. 650 00:59:10,800 --> 00:59:14,520 I mean, there's definitely potential insights there. 651 00:59:14,520 --> 00:59:17,600 So I don't think you should just throw it away. 652 00:59:17,600 --> 00:59:24,320 As for the analysis today, I looked at the people who left more than eight reviews, which 653 00:59:24,320 --> 00:59:28,800 I think was around 90 people. 654 00:59:28,800 --> 00:59:34,040 And once again, that's just the data that we had here. 655 00:59:34,040 --> 00:59:43,280 And also I think I may be running... 656 00:59:43,280 --> 00:59:52,040 So for example, if you take the reviewer who has left 70 reviews and their handle is Chill 657 00:59:52,040 --> 01:00:05,280 Panda and you look at their energetic uplifted response versus their relaxed sleepy response, 658 01:00:05,280 --> 01:00:09,520 you find that they flip into the inverted response class. 659 01:00:09,520 --> 01:00:10,960 It's real clear. 660 01:00:10,960 --> 01:00:12,320 Others don't. 661 01:00:12,320 --> 01:00:19,760 And so my intent was to go through some of the top reviewers and see how successful I 662 01:00:19,760 --> 01:00:23,360 can be at assigning them to response class. 663 01:00:23,360 --> 01:00:28,040 But if you're going to be dealing with this globally, I think you got to stratify your 664 01:00:28,040 --> 01:00:36,840 users, your reviewers into the appropriate response class before you draw too many conclusions. 665 01:00:36,840 --> 01:00:37,840 I hope that's clear. 666 01:00:37,840 --> 01:00:41,680 I can show you guys in a couple of weeks as I progress through that. 667 01:00:41,680 --> 01:00:43,840 I'll have some figures for that. 668 01:00:43,840 --> 01:00:51,560 You're definitely clear, especially in that we need to now control for different users. 669 01:00:51,560 --> 01:00:58,000 And this is why I really wanted to move into personality was because those are five metrics 670 01:00:58,000 --> 01:01:05,120 we can add on a user by user basis. 671 01:01:05,120 --> 01:01:11,880 If you have any other ways to add user specific data, then you can supplement there. 672 01:01:11,880 --> 01:01:19,760 I'll create this data set or this response for the top reviewers and see how well I can 673 01:01:19,760 --> 01:01:22,200 assign them to response class. 674 01:01:22,200 --> 01:01:28,460 And then if you want to see how that matches against any personality testing algorithms, 675 01:01:28,460 --> 01:01:29,460 go forth and prosper. 676 01:01:29,460 --> 01:01:34,640 I think that would be probably a good challenge. 677 01:01:34,640 --> 01:01:41,000 Whenever you're doing these kinds of analyses, you want to tell a story that hangs on something 678 01:01:41,000 --> 01:01:42,000 else. 679 01:01:42,000 --> 01:01:48,200 Otherwise, it just hangs out in the breeze and it doesn't go anywhere. 680 01:01:48,200 --> 01:01:51,240 You got to tie it back to a story. 681 01:01:51,240 --> 01:01:52,280 Exactly. 682 01:01:52,280 --> 01:01:55,920 And speaking of which, hopefully I can get through this. 683 01:01:55,920 --> 01:02:01,200 So I think my computer may be about to run out of juice here, but hopefully I can at 684 01:02:01,200 --> 01:02:07,360 least tell you the final bit of the story here to bring this all home. 685 01:02:07,360 --> 01:02:14,200 So where's this all going with the sentiment analysis? 686 01:02:14,200 --> 01:02:28,280 Well, so basically there was the frustration at the beginning that I've talked in with 687 01:02:28,280 --> 01:02:33,120 essentially legacy growers in Florida. 688 01:02:33,120 --> 01:02:40,840 And I made the mistake of saying, oh, things are going pretty good in Florida, huh? 689 01:02:40,840 --> 01:02:48,920 And they said, yeah, if you're a leaf or truly, it's going pretty well. 690 01:02:48,920 --> 01:02:53,600 And so what do they mean by that? 691 01:02:53,600 --> 01:03:05,480 Well, if you see who's licensed to operate in Florida, you'll see there's less than 692 01:03:05,480 --> 01:03:11,960 two dozen or so companies. 693 01:03:11,960 --> 01:03:23,320 So I try to be a glass half full person and say, hey, at least some's better than none. 694 01:03:23,320 --> 01:03:32,480 But then my economist gets in me and then I start seeing like, oh, like in Oklahoma, 695 01:03:32,480 --> 01:03:38,600 you've got hundreds of different retailers throughout the state. 696 01:03:38,600 --> 01:03:40,280 And so what did they mean by what they said? 697 01:03:40,280 --> 01:03:43,760 Oh, it's pretty good if you're a leaf or truly. 698 01:03:43,760 --> 01:03:47,520 Well, it's like, oh, like this is kind of what they meant. 699 01:03:47,520 --> 01:03:56,640 It's like, you know, all of these could potentially be independent companies, right? 700 01:03:56,640 --> 01:03:59,440 And that's exactly what you see in Oklahoma, right? 701 01:03:59,440 --> 01:04:09,080 Instead of 24 companies with hundreds of locations, you actually see hundreds of businesses, right? 702 01:04:09,080 --> 01:04:12,600 And I think that's so, right? 703 01:04:12,600 --> 01:04:17,480 So I think we now need to look at this data and see, okay, what's actually going on here 704 01:04:17,480 --> 01:04:20,000 in Florida. 705 01:04:20,000 --> 01:04:30,120 And I think that's sort of what some of the criticism is, is they say like, what's going 706 01:04:30,120 --> 01:04:39,880 on because I was actually speaking with an investigator at the Bureau of Cannabis Control 707 01:04:39,880 --> 01:04:41,400 in California. 708 01:04:41,400 --> 01:04:44,440 And I said, you know, everyone's talking about diversion. 709 01:04:44,440 --> 01:04:45,880 You know, what's it like? 710 01:04:45,880 --> 01:04:52,000 You know, is it like a box of eight pens that fell off the back of a truck? 711 01:04:52,000 --> 01:04:57,640 And he said, no, I went to a location that was supposed to have a million of eight pens 712 01:04:57,640 --> 01:05:01,280 and there were none there. 713 01:05:01,280 --> 01:05:08,080 And he said he can and did pass that information up the line, but that's all that he really 714 01:05:08,080 --> 01:05:09,720 could do there. 715 01:05:09,720 --> 01:05:17,960 So basically what I think people are frustrated about is there are supposedly agencies regulating 716 01:05:17,960 --> 01:05:18,960 this. 717 01:05:18,960 --> 01:05:28,680 However, everybody knows that the cannabis is being diverted and sold illegally. 718 01:05:28,680 --> 01:05:33,800 Yet we're somehow supposed to pretend, you know, that's not happening. 719 01:05:33,800 --> 01:05:43,880 And what we observe is right in Florida, there's many poor or working class farmers that would 720 01:05:43,880 --> 01:05:47,800 love a chance to grow cannabis. 721 01:05:47,800 --> 01:05:55,960 And they see their government agency, you know, award the contract to a company from 722 01:05:55,960 --> 01:06:04,200 the West Coast, where in my opinion, I think we need to change the null hypothesis here, 723 01:06:04,200 --> 01:06:05,200 right? 724 01:06:05,200 --> 01:06:11,320 Instead of the null hypothesis being there's no diversion, can you prove diversion? 725 01:06:11,320 --> 01:06:17,280 As far as California goes, in my opinion, I think the null hypothesis should be there 726 01:06:17,280 --> 01:06:22,120 is no legal cannabis in California. 727 01:06:22,120 --> 01:06:25,080 Please provide evidence if there is. 728 01:06:25,080 --> 01:06:31,120 This is where I am not trying to point fingers because I think everybody's just following 729 01:06:31,120 --> 01:06:32,640 their incentives. 730 01:06:32,640 --> 01:06:36,160 I don't fault anyone on the West Coast. 731 01:06:36,160 --> 01:06:42,360 In fact, people on the East Coast may not have cannabis if it was not for their efforts. 732 01:06:42,360 --> 01:06:44,440 They're following the incentives. 733 01:06:44,440 --> 01:06:47,640 They're doing what the incentives lie. 734 01:06:47,640 --> 01:06:50,920 So I don't think anybody should be frustrated with Jungle Boys. 735 01:06:50,920 --> 01:06:55,600 And in fact, it looks like people generally really like Jungle Boys and they really like 736 01:06:55,600 --> 01:06:57,160 their products. 737 01:06:57,160 --> 01:07:06,880 So if there's anyone to be frustrated at, I think it would be the Florida Department 738 01:07:06,880 --> 01:07:12,560 of Agriculture and every single politician in Florida. 739 01:07:12,560 --> 01:07:19,640 And I would now flip the null hypothesis here, where the null hypothesis would be there's 740 01:07:19,640 --> 01:07:25,700 no fraud, there's no corruption, there's no money being changed under the table. 741 01:07:25,700 --> 01:07:31,240 And I'm not making any accusations here, but I would flip the null hypothesis. 742 01:07:31,240 --> 01:07:43,300 And I would say, I want to see evidence that things are legitimate because ask me what 743 01:07:43,300 --> 01:07:46,480 my null hypothesis is. 744 01:07:46,480 --> 01:07:51,880 And so that's sort of, I think, what people's frustration is in Florida. 745 01:07:51,880 --> 01:07:59,280 I think they don't know who to be frustrated with. 746 01:07:59,280 --> 01:08:02,820 They want to be able to participate. 747 01:08:02,820 --> 01:08:08,300 They just see everything kind of get handed over and it's like they don't want to be 748 01:08:08,300 --> 01:08:14,820 frustrated at Jungle Boys because they may identify with Jungle Boys. 749 01:08:14,820 --> 01:08:18,000 And as I said, I don't necessarily think they're doing anything wrong. 750 01:08:18,000 --> 01:08:20,160 That's just where the incentives are. 751 01:08:20,160 --> 01:08:29,200 So basically, if you're in Florida and you're frustrated that you can't join in on the fine, 752 01:08:29,200 --> 01:08:35,240 I think you kind of got to look at your politicians. 753 01:08:35,240 --> 01:08:44,400 And that's sort of where I kind of just have kind of just kept following this. 754 01:08:44,400 --> 01:08:46,640 And I kind of identify with the frustration. 755 01:08:46,640 --> 01:08:49,960 And I'm not trying to step on anyone's toes. 756 01:08:49,960 --> 01:08:51,720 I'm out here on the West Coast. 757 01:08:51,720 --> 01:08:57,280 I am sure I can assume some West Coast cannabis in my day. 758 01:08:57,280 --> 01:09:02,160 So no hard feelings to anyone. 759 01:09:02,160 --> 01:09:08,080 Really I'm just saying like, hey, I think if there's any frustration, I think we may 760 01:09:08,080 --> 01:09:11,280 need to have some frustration with our politicians. 761 01:09:11,280 --> 01:09:15,500 And in my opinion, vote for someone younger, right? 762 01:09:15,500 --> 01:09:17,020 We need a solution. 763 01:09:17,020 --> 01:09:22,580 So we looked at the data and I'm not trying to be ageist or anything, but we did look 764 01:09:22,580 --> 01:09:34,960 at the data and the younger crowd is generally more on board with cannabis being permitted. 765 01:09:34,960 --> 01:09:37,040 So that's all I would necessarily do. 766 01:09:37,040 --> 01:09:41,920 Like if there were two candidates running against each other, I don't even know if I 767 01:09:41,920 --> 01:09:43,240 would look at political party. 768 01:09:43,240 --> 01:09:50,200 I would just go with who's the younger chap or gal. 769 01:09:50,200 --> 01:09:52,400 So I don't know. 770 01:09:52,400 --> 01:09:53,560 Those are my thoughts. 771 01:09:53,560 --> 01:09:58,240 I'm not really trying to have this be a political at all. 772 01:09:58,240 --> 01:10:01,520 I just kind of share with some of the frustration. 773 01:10:01,520 --> 01:10:07,840 But at the same time, maybe this is something to be celebrated. 774 01:10:07,840 --> 01:10:08,840 I don't know. 775 01:10:08,840 --> 01:10:10,400 I'll kind of pass this on to you, right? 776 01:10:10,400 --> 01:10:13,200 Because should we be a glass half full? 777 01:10:13,200 --> 01:10:17,360 Like is it awesome that Jungle Boys are coming to Florida? 778 01:10:17,360 --> 01:10:28,880 Like are they knocking down the wall or will the Florida growers get their chance at this? 779 01:10:28,880 --> 01:10:30,320 So that's my spiel. 780 01:10:30,320 --> 01:10:32,320 Those are my thoughts. 781 01:10:32,320 --> 01:10:34,320 I'll pass it back to you. 782 01:10:34,320 --> 01:10:40,760 So today was already a controversial day, so may as well make it 100% controversial. 783 01:10:40,760 --> 01:10:45,640 But any thoughts, comments, questions before we call it a day? 784 01:10:45,640 --> 01:10:46,640 Thank you. 785 01:10:46,640 --> 01:10:47,840 That's really good to know. 786 01:10:47,840 --> 01:10:49,840 So I don't even know what the Jungle Boys are. 787 01:10:49,840 --> 01:10:51,760 So I'm going to Google that. 788 01:10:51,760 --> 01:10:52,760 And I just left Florida. 789 01:10:52,760 --> 01:10:53,760 I'm four hours from Massachusetts. 790 01:10:53,760 --> 01:11:00,280 And as soon as I get back, I'll get settled and then I'll start getting re-engaged. 791 01:11:00,280 --> 01:11:02,040 Thank you so much, Keegan. 792 01:11:02,040 --> 01:11:03,040 Thanks, Candace. 793 01:11:03,040 --> 01:11:06,560 And you're the person who's in Florida. 794 01:11:06,560 --> 01:11:11,520 So don't let me tell you what you should and shouldn't do in your state. 795 01:11:11,520 --> 01:11:12,520 This is up to you. 796 01:11:12,520 --> 01:11:13,520 I agree with you. 797 01:11:13,520 --> 01:11:14,520 I agree with you. 798 01:11:14,520 --> 01:11:20,040 I think what frustrates me with Florida is I can grow weed up in Massachusetts. 799 01:11:20,040 --> 01:11:21,040 No problem, right? 800 01:11:21,040 --> 01:11:23,800 Because I kicked all those pharmaceuticals, right? 801 01:11:23,800 --> 01:11:25,920 The Parkinson disease pills, everything. 802 01:11:25,920 --> 01:11:27,780 You saw the list, right? 803 01:11:27,780 --> 01:11:31,960 And so it frustrates me that I just can't grow in Florida. 804 01:11:31,960 --> 01:11:35,920 I have to actually buy medical marijuana. 805 01:11:35,920 --> 01:11:38,960 But I'm grateful for that, that it's available. 806 01:11:38,960 --> 01:11:41,680 But no, you rock. 807 01:11:41,680 --> 01:11:42,760 I think it's great. 808 01:11:42,760 --> 01:11:59,440 Thank you so much for what you do.