1 00:00:00,000 --> 00:00:12,000 What's new with you today? 2 00:00:12,000 --> 00:00:16,000 Um, been busy with GTC. 3 00:00:16,000 --> 00:00:18,000 What's new about that? 4 00:00:18,000 --> 00:00:24,000 GTC, the NVIDIA Machine Learning Conference. 5 00:00:24,000 --> 00:00:26,000 Oh wow. 6 00:00:26,000 --> 00:00:32,000 Tell me more about it. I haven't heard of it. So enlighten me. 7 00:00:32,000 --> 00:00:40,000 So, I've learned about this thing called, there's a program called Blazing SQL. 8 00:00:40,000 --> 00:00:47,000 And you can do SQL queries from CSV files and Parquet files. 9 00:00:47,000 --> 00:00:54,000 You don't actually have to load them into a database. 10 00:00:54,000 --> 00:00:59,000 Say that one more time. What are you doing with the CSVs? 11 00:00:59,000 --> 00:01:05,000 It's called, in the conference or the Blazing SQL? 12 00:01:05,000 --> 00:01:09,000 Um, this Blazing SQL that you're talking about. 13 00:01:09,000 --> 00:01:17,000 Blazing SQL is, it's supposed to be like an accelerated version of SQL. 14 00:01:17,000 --> 00:01:27,000 It can take advantage of your GPU, but it'll also allow you to do SQL queries directly from CSV files. 15 00:01:27,000 --> 00:01:30,000 Ooh. 16 00:01:30,000 --> 00:01:39,000 There are a couple data points in particular that I am looking for. 17 00:01:39,000 --> 00:01:48,000 And I was wondering if you may be able to help me get those out of those Washington State ones, because that Blazing SQL sounds like it's the right thing. 18 00:01:48,000 --> 00:01:59,000 The two data points I'm looking for are total waste, or, I'll write this down too. 19 00:01:59,000 --> 00:02:07,000 But, sorry to start writing you a laundry list, but essentially 20 00:02:07,000 --> 00:02:16,000 total waste by licensee by day. 21 00:02:16,000 --> 00:02:29,000 And there may be a couple different waste types, but essentially, so it would be nice to have the total for each type, but essentially just the daily totals. 22 00:02:29,000 --> 00:02:38,000 Because what would be nice would be, oh, there's someone, Josh. 23 00:02:38,000 --> 00:02:41,000 I think. 24 00:02:41,000 --> 00:02:45,000 Welcome, Josh. 25 00:02:45,000 --> 00:02:55,000 We're just beginning to talk about some high in demand data points, but I'm glad to have you join the group. 26 00:02:55,000 --> 00:02:59,000 Yeah. 27 00:02:59,000 --> 00:03:05,000 Sorry, my speakers on the wrong machine. That's interesting. 28 00:03:05,000 --> 00:03:09,000 Well, I guess we'll dive back into this data points in one second, Charles. 29 00:03:09,000 --> 00:03:12,000 I've still got them on my mind. 30 00:03:12,000 --> 00:03:18,000 I guess just welcome to the group, Josh. At the moment, we're just talking about 31 00:03:18,000 --> 00:03:25,000 the tools is discovered this tool blazing SQL. 32 00:03:25,000 --> 00:03:32,000 To basically help parse some of this Washington State cannabis data that we're working on. 33 00:03:32,000 --> 00:03:35,000 Just this big data dump. 34 00:03:35,000 --> 00:03:41,000 And some of the files are actually so large, it's tough to manage. 35 00:03:41,000 --> 00:03:53,000 So that's experiments with Dask also trying to get these files and stuff and 36 00:03:53,000 --> 00:03:57,000 trying to configure it. 37 00:03:57,000 --> 00:04:01,000 Dask. 38 00:04:01,000 --> 00:04:05,000 Sorry. 39 00:04:05,000 --> 00:04:18,000 Essentially, Charles, you've got this big data set there of the date of not even daily of just every observation there is. 40 00:04:18,000 --> 00:04:26,000 And so, you know, as the cultivators go about their daily business, they're recording their waste. 41 00:04:26,000 --> 00:04:31,000 And so I think you said you discovered that in harvest batches, which is creating 42 00:04:31,000 --> 00:04:35,000 waste from their harvest essentially. 43 00:04:35,000 --> 00:04:44,000 Right. 44 00:04:44,000 --> 00:04:48,000 If there's any money to essentially 45 00:04:48,000 --> 00:04:53,000 you may know more about this than I do some sort of aggregation query where you're 46 00:04:53,000 --> 00:04:56,000 summing up 47 00:04:56,000 --> 00:05:02,000 the waste by day and by licensee. 48 00:05:02,000 --> 00:05:06,000 If you could get that time series, 49 00:05:06,000 --> 00:05:11,000 that's actually what the company I was talking about better carbon solutions is looking for. 50 00:05:11,000 --> 00:05:14,000 They're trying to 51 00:05:14,000 --> 00:05:19,000 they're trying to break into like essentially the cannabis waste here in Washington. 52 00:05:19,000 --> 00:05:25,000 There's other in other states like California 53 00:05:25,000 --> 00:05:30,000 Colorado and even Oklahoma. 54 00:05:30,000 --> 00:05:34,000 Cannabis waste is actually 55 00:05:34,000 --> 00:05:37,000 the big 56 00:05:37,000 --> 00:05:45,000 it's a small, you know, it's a concern and there's a dozen or so companies dealing with 57 00:05:45,000 --> 00:05:49,000 though 58 00:05:49,000 --> 00:05:56,000 as they launch out, they're just going to need essentially a forecast of how much waste is even being produced. 59 00:05:56,000 --> 00:06:01,000 That way you can just sort of get a gauge on what you can do with it. 60 00:06:01,000 --> 00:06:07,000 And that's where the cannabis data science group can help out because 61 00:06:07,000 --> 00:06:15,000 one of the tools in my toolbox is forecasting. So give me a good time series and we can forecast it. 62 00:06:15,000 --> 00:06:18,000 So 63 00:06:18,000 --> 00:06:23,000 we've discovered this, you know, Washington State cannabis data dump 64 00:06:23,000 --> 00:06:25,000 through public records. 65 00:06:25,000 --> 00:06:35,000 And so now we're going to try to aggregate the waste by day. 66 00:06:35,000 --> 00:06:43,000 Can you say that one more time? 67 00:06:43,000 --> 00:06:53,000 I started working on the waste problem early on and then I kind of get sidetracked. 68 00:06:53,000 --> 00:06:56,000 Yeah, it doesn't have to be that one. 69 00:06:56,000 --> 00:07:06,000 Like I said, really any totals by day by licensee. 70 00:07:06,000 --> 00:07:13,000 The other total that I was interested in, it doesn't necessarily have to be by licensee, was 71 00:07:13,000 --> 00:07:22,000 I was essentially just wanting like a daily count of like lab tests, just to know how many lab tests are happening 72 00:07:22,000 --> 00:07:28,000 per day. And then once again, forecast that into the future. 73 00:07:28,000 --> 00:07:38,000 But yeah, you know, I need to publish that. What I found is that 74 00:07:38,000 --> 00:07:43,000 they don't record the lab tests 75 00:07:43,000 --> 00:07:47,000 like when they do them. 76 00:07:47,000 --> 00:07:52,000 I think maybe some places where you do it every day, 77 00:07:52,000 --> 00:08:00,000 it's a mark or something. It's not. 78 00:08:00,000 --> 00:08:08,000 That was that actually reminded me that was actually the data point. 79 00:08:08,000 --> 00:08:15,000 The data point we were looking for was a sales 80 00:08:15,000 --> 00:08:19,000 per lab result. 81 00:08:19,000 --> 00:08:24,000 I think the sales are more updated regularly. 82 00:08:24,000 --> 00:08:27,000 SQ502 data. 83 00:08:27,000 --> 00:08:32,000 And so if you could do like the SQL on like the sales, 84 00:08:32,000 --> 00:08:39,000 so basically just do sales for global lab result X 85 00:08:39,000 --> 00:08:47,000 and then A. I think it can be done because I think I've seen the data, but it's 86 00:08:47,000 --> 00:08:52,000 not the same as the data point. 87 00:08:52,000 --> 00:08:59,000 So the reason that data point is interesting is that gives you sort of if you average, 88 00:08:59,000 --> 00:09:04,000 if you take the average of the sales per lab result, 89 00:09:04,000 --> 00:09:10,000 that's basically the implied cost of a failed test. 90 00:09:10,000 --> 00:09:16,000 Because if you otherwise wouldn't have failed, then you would have on average, 91 00:09:16,000 --> 00:09:20,000 so it's a crude measure. 92 00:09:20,000 --> 00:09:26,000 But basically, you know, in economics, you try to measure all the costs, 93 00:09:26,000 --> 00:09:28,000 explicit and implicit. 94 00:09:28,000 --> 00:09:32,000 So we're trying to measure all the costs for cultivators. 95 00:09:32,000 --> 00:09:37,000 And one of their costs is the risk of failure. 96 00:09:37,000 --> 00:09:45,000 And the way you measure that cost is actually by their probability of failure. 97 00:09:45,000 --> 00:09:51,000 So that would just be the failure rate, you know, whatever percentage that may be, 98 00:09:51,000 --> 00:10:00,000 times the cost, which is going to be the average sales per lab result. 99 00:10:00,000 --> 00:10:07,000 So long story short is you can estimate the cost of failure. 100 00:10:07,000 --> 00:10:09,000 And I don't think anybody's estimated that. 101 00:10:09,000 --> 00:10:14,000 And I think they've maybe estimated that actually somebody's estimated that cost in California. 102 00:10:14,000 --> 00:10:17,000 But nobody's estimated that cost in Washington. 103 00:10:17,000 --> 00:10:26,000 And Washington has actually a lot less stringent regulations than California. 104 00:10:26,000 --> 00:10:31,000 So it would just be interesting to compare the costs here in California to, 105 00:10:31,000 --> 00:10:34,000 I mean here in Washington to California. 106 00:10:34,000 --> 00:10:35,000 Okay. 107 00:10:35,000 --> 00:10:36,000 That is related. 108 00:10:36,000 --> 00:10:40,000 I've been putting together a data set. 109 00:10:40,000 --> 00:10:46,000 I was going to put that like THC, like overclock, 110 00:10:46,000 --> 00:10:50,000 how it was rising and how that related to sales. 111 00:10:50,000 --> 00:10:55,000 Because I've been told that, you know, that's like a driving thing. 112 00:10:55,000 --> 00:10:58,000 What's a driving thing? 113 00:10:58,000 --> 00:11:04,000 The percentage of THC, like that's like a big selling thing. 114 00:11:04,000 --> 00:11:13,000 Going up, I'm wondering, is that like driving sales on? 115 00:11:13,000 --> 00:11:21,000 So just from the grapevine, I think, you know, some people are curious about that question. 116 00:11:21,000 --> 00:11:27,000 And so it's actually an easy research question to answer. 117 00:11:27,000 --> 00:11:31,000 I mean, you could do it more formally. 118 00:11:31,000 --> 00:11:47,000 But the real informal quick way you do it is you would just take a regression of sales on. 119 00:11:47,000 --> 00:11:54,000 Well, I think that's where you would do the sales per lab result. 120 00:11:54,000 --> 00:12:01,000 So you would look at, so your independent variable, 121 00:12:01,000 --> 00:12:09,000 or your dependent variable Y would be the sales per lab result. 122 00:12:09,000 --> 00:12:13,000 And then your independent variables, your regressors, 123 00:12:13,000 --> 00:12:19,000 would be the THC percentage of that lab result. 124 00:12:19,000 --> 00:12:25,000 And so then you could actually just do an ordinary least squares regression. 125 00:12:25,000 --> 00:12:32,000 And you would just see, okay, what's the coefficient on THC? 126 00:12:32,000 --> 00:12:39,000 You know, your hypothesis would be that it would be a positive coefficient. 127 00:12:39,000 --> 00:12:44,000 And the question is, what would be the magnitude? 128 00:12:44,000 --> 00:12:56,000 Is the magnitude even significant? 129 00:12:56,000 --> 00:13:00,000 And then, like I said, you could then do it, you could do fancier things. 130 00:13:00,000 --> 00:13:05,000 Like you could like take the log of both, I mean, that would almost be a given. 131 00:13:05,000 --> 00:13:12,000 You can take the log of both sides. 132 00:13:12,000 --> 00:13:15,000 And, you know, you could do fancier regressions, 133 00:13:15,000 --> 00:13:20,000 but often just the ordinary least squares is pretty informative. 134 00:13:20,000 --> 00:13:22,000 It's a good place to begin. 135 00:13:22,000 --> 00:13:27,000 That's my recommendation. 136 00:13:27,000 --> 00:13:32,000 If you scrape that data together, you could show you how to run the regression. 137 00:13:32,000 --> 00:13:40,000 But essentially, like I said, the two data points you need are sales per lab result. 138 00:13:40,000 --> 00:13:45,000 And then you would actually need the actual THC of that lab result. 139 00:13:45,000 --> 00:13:47,000 Right. 140 00:13:47,000 --> 00:13:50,000 So, yeah, it's just kind of linking the data. 141 00:13:50,000 --> 00:13:55,000 The lab results link to the inventory, which links to the sales. 142 00:13:55,000 --> 00:14:01,000 And so that's like the inventory and the sales are huge data sets. 143 00:14:01,000 --> 00:14:07,000 And it's trying to get it all into memory or at least swapped out the disk. 144 00:14:07,000 --> 00:14:12,000 And getting that all to join together has been a challenge. 145 00:14:12,000 --> 00:14:23,000 I was wondering if you could split up the data sets that in a way that wouldn't hurt the performance of any predictive model, 146 00:14:23,000 --> 00:14:32,000 like maybe yearly would be doable since I think if you have years of data, 147 00:14:32,000 --> 00:14:42,000 I think the year doesn't affect sales necessarily, but the month and day typically does. 148 00:14:42,000 --> 00:14:50,000 I think the problem is I could be wrong, but I think there's almost like unzipped. 149 00:14:50,000 --> 00:14:54,000 It could be like 100 gigabytes of sales data a month. 150 00:14:54,000 --> 00:14:56,000 And that could be wrong. 151 00:14:56,000 --> 00:15:02,000 But am I wrong? Was it like about that long? 152 00:15:02,000 --> 00:15:10,000 Yeah, it's over 100 gigabytes of data. 153 00:15:10,000 --> 00:15:14,000 I'm not sure if it's a total or a month, but... 154 00:15:14,000 --> 00:15:22,000 I don't even know how you run analytics on anything other than like a server, 155 00:15:22,000 --> 00:15:27,000 unless you can make summaries of the data over time. 156 00:15:27,000 --> 00:15:33,000 Well, that's my whole going down the Dask rabbit hole. 157 00:15:33,000 --> 00:15:42,000 That's supposed to solve that problem, and it probably does once you get Dask configured right. 158 00:15:42,000 --> 00:15:51,000 And I also found that Dask and Pandas runs differently on my Mac than it does on my Linux box. 159 00:15:51,000 --> 00:16:02,000 So I've been discovering things by running experiments on each machine. 160 00:16:02,000 --> 00:16:09,000 Well, it works, I guess. I've never used Dask. 161 00:16:09,000 --> 00:16:11,000 You've got the right idea though, Josh. 162 00:16:11,000 --> 00:16:21,000 Essentially, we're just trying to get like a time series of data, like summary data, metadata out of this, 163 00:16:21,000 --> 00:16:25,000 and just leave the bulk of it there. 164 00:16:25,000 --> 00:16:29,000 We're just looking for daily totals. 165 00:16:29,000 --> 00:16:32,000 Yeah. 166 00:16:32,000 --> 00:16:47,000 That's just sort of an ongoing project, but like Charles has said, that's sort of the question on everybody's mind is 167 00:16:47,000 --> 00:16:51,000 how much does THC affect sales? 168 00:16:51,000 --> 00:17:10,000 So like how much is like a 1% increase in THC? Like how much in dollars would that change at the retailer? 169 00:17:10,000 --> 00:17:21,000 Even if you did, if you had 0.1% differences in THC values, then what were you able to get the data for those specific ones? 170 00:17:21,000 --> 00:17:30,000 Then you could also do like a whole like 1000 values of different percentages of THC at most every day 171 00:17:30,000 --> 00:17:37,000 and probably have a small amount of data compared to 100 gigabytes per month. 172 00:17:37,000 --> 00:17:53,000 So that's also pretty good measurement that you can do. 173 00:17:53,000 --> 00:17:56,000 I kind of lost internet connection. 174 00:17:56,000 --> 00:18:03,000 I'm back. I lost connection. 175 00:18:03,000 --> 00:18:08,000 Josh, could you please repeat your last part? 176 00:18:08,000 --> 00:18:19,000 So I was thinking even if you had 0.1% scale of the amount of THC in marijuana, 177 00:18:19,000 --> 00:18:25,000 then you had sale data with respect to that, with respect to each of those products, 178 00:18:25,000 --> 00:18:34,000 you could at most have 1000 bins for each of those sales and then have that for every single day. 179 00:18:34,000 --> 00:18:39,000 And it would still be very small compared to the 100 gigabytes per month. 180 00:18:39,000 --> 00:18:43,000 And that would be 1000 points at most. 181 00:18:43,000 --> 00:18:53,000 I doubt I don't think you can have marijuana with 100% THC. That would be ridiculous. 182 00:18:53,000 --> 00:19:00,000 I think, Greg, is sort of the point you're going at. 183 00:19:00,000 --> 00:19:09,000 Are you just like looking for like observations that span like the spectrum of possibilities or? 184 00:19:09,000 --> 00:19:14,000 I think if you're saying THC is important for sale data, 185 00:19:14,000 --> 00:19:22,000 then that would be an important observation to get the analytics out of. 186 00:19:22,000 --> 00:19:29,000 If it's also time based, then those two could be correlated as well. 187 00:19:29,000 --> 00:19:38,000 Depending on the month, like Seattle, you have seasonal effective disorders so that you may have different audiences. 188 00:19:38,000 --> 00:19:43,000 You actually just hit on a good point. 189 00:19:43,000 --> 00:19:49,000 So that would actually you're actually talking about panel data. 190 00:19:49,000 --> 00:19:57,000 So what you have is observation Y of IT. 191 00:19:57,000 --> 00:20:08,000 So you'd have sales of lab result I, you know, at time T. 192 00:20:08,000 --> 00:20:11,000 So that's how you could improve the model. 193 00:20:11,000 --> 00:20:21,000 So essentially you just the easy way to do is just toss in a like a time effect. 194 00:20:21,000 --> 00:20:32,000 So just like a counter zero to N or zero to T for however many periods you have. 195 00:20:32,000 --> 00:20:45,000 So the regression and you control for time, there may be more elevators to do it. 196 00:20:45,000 --> 00:20:57,000 I think time is a good factor because that's actually something that I was looking at recently was literally prices in Oregon. 197 00:20:57,000 --> 00:21:01,000 Oregon makes their retail and wholesale prices available. 198 00:21:01,000 --> 00:21:08,000 And so you see you actually see prices, you know, go with, you know, dip way down over time. 199 00:21:08,000 --> 00:21:18,000 And actually recently they're actually kind of trending up a little bit, believe it or not. 200 00:21:18,000 --> 00:21:25,000 Yeah, I don't get to drive around much right now, but I don't drive around much. 201 00:21:25,000 --> 00:21:35,000 But before I could remember you drive down the street and you just see people with signs out signs out and, you know, the price just kept going down and down and down. 202 00:21:35,000 --> 00:21:42,000 But I haven't seen lately if it's like, you know, what people are advertising and. 203 00:21:42,000 --> 00:21:50,000 Well, do you want to see real quick that was sort of a when I put together for today. 204 00:21:50,000 --> 00:22:03,000 So cool. Yeah, I'll just show you real quick. 205 00:22:03,000 --> 00:22:09,000 So essentially, you know. 206 00:22:09,000 --> 00:22:14,000 Here's just, you know, total sales in Oregon. 207 00:22:14,000 --> 00:22:20,000 And then here are actually. 208 00:22:20,000 --> 00:22:32,000 And actually, if you want to just find it directly online. 209 00:22:32,000 --> 00:22:37,000 Oregon has an all right data dashboard here. 210 00:22:37,000 --> 00:22:46,000 The problem is, it's a little difficult to download data, but at least it's here. 211 00:22:46,000 --> 00:22:49,000 So data. 212 00:22:49,000 --> 00:23:05,000 I'll actually push this to the to the GitHub here. 213 00:23:05,000 --> 00:23:28,000 So I'll add I'll push some links here to the to the Oregon sales data. 214 00:23:28,000 --> 00:23:44,000 And then I'll just show you some of these prices open. So that's the other thing is I'm sort of recording some of these five million essential. 215 00:23:44,000 --> 00:23:51,000 So here you know you've got the Oregon data and so you've got. 216 00:23:51,000 --> 00:24:07,000 You know the wholesale and retail. 217 00:24:07,000 --> 00:24:11,000 So here's what I was sort of referring to. 218 00:24:11,000 --> 00:24:20,000 So you see, you know, like for concentrates, at least, you know, you've got the price drop by, you know, almost 50 percent. 219 00:24:20,000 --> 00:24:32,000 And now it's sort of the concentrates hovering around $20 a gram, you know, with maybe a slight uptick. 220 00:24:32,000 --> 00:24:41,000 So you're dropping from like $10 a gram, you know, down to about $5 a gram. 221 00:24:41,000 --> 00:24:45,000 But more actually like four and a half dollars. 222 00:24:45,000 --> 00:24:57,000 And then now it's sort of rising above to about five, five and a half dollars. 223 00:24:57,000 --> 00:25:10,000 So one thing that I was going to to introduce to you and then start calculating over this meetup in the next meetup was essentially inflation. 224 00:25:10,000 --> 00:25:15,000 And so, you know, that's an interesting economic variable. 225 00:25:15,000 --> 00:25:21,000 And I don't think people have really looked at it that much in the cannabis industry. 226 00:25:21,000 --> 00:25:27,000 I mean, I'm sure people have to a certain extent, however, I don't hear too much talk about it. 227 00:25:27,000 --> 00:25:34,000 So, you know, it's worth at least calculating at once just to see what's going on. 228 00:25:34,000 --> 00:25:47,000 So, you know, essentially we'll be calculating, you know, the price today, you know, minus the price yesterday divided by the price yesterday. 229 00:25:47,000 --> 00:26:06,000 I've realized the way that we want to do this properly is actually take a weighted average of price based off of the amount sold. 230 00:26:06,000 --> 00:26:13,000 So you'll want. Who was your question? 231 00:26:13,000 --> 00:26:34,000 Well, inflation is interesting because that reminded me, I think I heard someone arguing that Bitcoin wasn't actually inflating necessarily if you took some metrics of inflation with respect to the dollar, because the total amount of dollars in circulation was getting larger. 232 00:26:34,000 --> 00:26:38,000 Actually, just total amount of dollars available. 233 00:26:38,000 --> 00:26:46,000 But then it's hard to calculate the US dollars because is it available or is it in circulation? 234 00:26:46,000 --> 00:26:49,000 Or where is it? 235 00:26:49,000 --> 00:26:54,000 Well, that's it. 236 00:26:54,000 --> 00:26:56,000 That's a good question. 237 00:26:56,000 --> 00:26:59,000 I hadn't thought about that. 238 00:26:59,000 --> 00:27:02,000 I don't necessarily have a stable metric to compare this to. 239 00:27:02,000 --> 00:27:27,000 So just in itself, you know, the Bitcoin series is probably inflating because, you know, the price today, depending on how granular you want to measure price, maybe monthly or something, it depends on the time span you're looking at. 240 00:27:27,000 --> 00:27:29,000 I think that would be an interesting thing to measure. 241 00:27:29,000 --> 00:27:40,000 One thing I would just point out is I guess inflation is typically thought of as a like an aggregate of like all prices. 242 00:27:40,000 --> 00:27:42,000 So you'd be thinking of. 243 00:27:42,000 --> 00:27:43,000 Yeah. 244 00:27:43,000 --> 00:27:47,000 Typically a macroeconomic variable. 245 00:27:47,000 --> 00:27:56,000 I don't see any reason why you couldn't apply it to a single, single market. 246 00:27:56,000 --> 00:28:00,000 I think you'd have to have something like a cost of living index. 247 00:28:00,000 --> 00:28:04,000 And if that went up, that would be. 248 00:28:04,000 --> 00:28:09,000 I think that would actually be described as inflation by the legal definition. 249 00:28:09,000 --> 00:28:11,000 If I remember it correctly. 250 00:28:11,000 --> 00:28:15,000 Exactly. 251 00:28:15,000 --> 00:28:27,000 So basically, you know, what you're getting on here is like the basket of goods that the consumer has. 252 00:28:27,000 --> 00:28:33,000 So. 253 00:28:33,000 --> 00:28:51,000 So if you go, you know, if you go up here, you'll see, you know, out of the total sales, you know, let's say, you know, let's say flower is, you know, 60 million. 254 00:28:51,000 --> 00:29:04,000 Out of, you know, 100 million, and then we'll just say concentrates is, you know, 40 out of 100. 255 00:29:04,000 --> 00:29:13,000 Out of 100 million. 256 00:29:13,000 --> 00:29:27,000 So what we're basically saying here is, you know, the consumer's basket, you know, you're basically saying, you know, your basket. 257 00:29:27,000 --> 00:29:32,000 Is, you know, 60 percent. 258 00:29:32,000 --> 00:29:43,000 Flower and, you know, 40 percent concentrate. 259 00:29:43,000 --> 00:29:47,000 Sorry for the crude handwriting. 260 00:29:47,000 --> 00:29:55,000 So you've got 60 percent flower, 40 percent concentrate in just a consumer's average basket. 261 00:29:55,000 --> 00:30:04,000 So the way you would then measure inflation of a, you know, a cannabis basket. 262 00:30:04,000 --> 00:30:13,000 So, you know, for a consumer of cannabis, how, you know, what's the inflation in their cannabis basket? 263 00:30:13,000 --> 00:30:29,000 Well, you know, that would basically be, you know, 60 percent times, you know, you know, you know, five point five. 264 00:30:29,000 --> 00:30:31,000 Sorry for the. 265 00:30:31,000 --> 00:30:47,000 You know, and then you would just add, you know, the 40 percent of which they're spending, you know, 40 percent on concentrates. 266 00:30:47,000 --> 00:30:54,000 And then the concentrates is, you know, about twenty dollars. 267 00:30:54,000 --> 00:31:02,000 So basically what you're saying, you know, is basically at time t. 268 00:31:02,000 --> 00:31:12,000 You know, your price is, you know, point four to your price at time t is like five plus point six times. 269 00:31:12,000 --> 00:31:21,000 So, you know, so that so that's essentially what your price at time t is. 270 00:31:21,000 --> 00:31:29,000 And that this can be thought of. 271 00:31:29,000 --> 00:31:33,000 It's just a measure of like aggregate prices. 272 00:31:33,000 --> 00:31:38,000 And so then you kind of just see. 273 00:31:38,000 --> 00:31:43,000 Is, you know, do you would expect. 274 00:31:43,000 --> 00:31:55,000 Prices to be rising over time gradually, you know, maybe the economy is a whole, you know. 275 00:31:55,000 --> 00:32:05,000 I don't know what the inflation rate is off the top of my head, but, you know, maybe like one to three percent or so. 276 00:32:05,000 --> 00:32:13,000 So it would just be real interesting to see what the inflation rate is in the cannabis industry. 277 00:32:13,000 --> 00:32:19,000 And like I said, it's interesting. 278 00:32:19,000 --> 00:32:25,000 And you could almost do like a. 279 00:32:25,000 --> 00:32:33,000 Like a time series analysis and, you know, you basically see like. 280 00:32:33,000 --> 00:32:36,000 You know, is there like a difference? 281 00:32:36,000 --> 00:32:42,000 You like you could basically see like, oh, like, was there some sort of. 282 00:32:42,000 --> 00:32:48,000 You know, struck what you know, was there some sort of structural break in the market? 283 00:32:48,000 --> 00:32:53,000 Because it looks like things have kind of like stabilized. 284 00:32:53,000 --> 00:32:58,000 And like I said, they may even be kind of trending up. 285 00:32:58,000 --> 00:33:03,000 It's kind of hard to tell. 286 00:33:03,000 --> 00:33:12,000 I think a huge entry into the world and that forced. 287 00:33:12,000 --> 00:33:14,000 We have the same problem. 288 00:33:14,000 --> 00:33:18,000 We have the same problem with Chinese restaurants in this area. 289 00:33:18,000 --> 00:33:21,000 There was the dim sum wars. 290 00:33:21,000 --> 00:33:28,000 There are Chinese restaurants popping up like just, you know, like one right next to the other. 291 00:33:28,000 --> 00:33:30,000 And that forced prices down. 292 00:33:30,000 --> 00:33:34,000 And I saw the same thing with dispensers, like they were just popping up everywhere. 293 00:33:34,000 --> 00:33:38,000 And so maybe, you know, at some point, right, the market's going to be saturated. 294 00:33:38,000 --> 00:33:40,000 You can't open up any more. 295 00:33:40,000 --> 00:33:44,000 And, you know, so but originally it caused prices to drop. 296 00:33:44,000 --> 00:33:51,000 I think it's probably stabilized, especially with the pandemic, people are probably not opening up dispense, a lot of dispensary. 297 00:33:51,000 --> 00:33:58,000 So prices are probably starting to stabilize and they probably will start to rise again. 298 00:33:58,000 --> 00:34:05,000 Exactly. And that's what you're, you know, you see. 299 00:34:05,000 --> 00:34:16,000 That's why it will be nice to like look at some statistics and like do some trends. But I mean, it does look to me that maybe. 300 00:34:16,000 --> 00:34:23,000 Maybe starting in around June of 2019, things are gradual. 301 00:34:23,000 --> 00:34:26,000 I mean, prices are gradually trending up. 302 00:34:26,000 --> 00:34:30,000 But for the most part, it looks pretty stable. 303 00:34:30,000 --> 00:34:37,000 I mean, it looks compared to the prior two years. 304 00:34:37,000 --> 00:34:39,000 It's pretty stable. 305 00:34:39,000 --> 00:34:42,000 And. 306 00:34:42,000 --> 00:34:59,000 And in fact, that would actually be what economic theory would suggest is you would suggest that prices would would, for the most part, stabilize over time with a moderate amount of inflation. 307 00:34:59,000 --> 00:35:07,000 And just to kind of show you a bit more of the economic analysis that would essentially go into this. 308 00:35:07,000 --> 00:35:14,000 During this same time, keep in mind that. 309 00:35:14,000 --> 00:35:24,000 You know, one would, economics would suggest that, you know, prices, even in the cannabis industry, would be affected by the industry. 310 00:35:24,000 --> 00:35:31,000 And essentially, during this time, you know. 311 00:35:31,000 --> 00:35:38,000 See, it's real interesting, right? So you see this graph. 312 00:35:38,000 --> 00:35:40,000 And, you know, things are kind of stable. 313 00:35:40,000 --> 00:35:44,000 But at the same time, it doesn't even really look like. 314 00:35:44,000 --> 00:35:54,000 Cannabis prices are even correlated. I mean, at first glance with, you know, the federal interest rate. 315 00:35:54,000 --> 00:36:03,000 But once again, that's where we're going to use some statistics and see, you know, what what may be the relationship or they even correlated. 316 00:36:03,000 --> 00:36:12,000 And if so, to what degree. 317 00:36:12,000 --> 00:36:18,000 So long story short is the Federal Reserve kind of. 318 00:36:18,000 --> 00:36:21,000 So the way that the. 319 00:36:21,000 --> 00:36:26,000 The Federal Reserve policy generally works is. 320 00:36:26,000 --> 00:36:39,000 When the economy is expanding, they try to raise the interest rate as high as they really can. 321 00:36:39,000 --> 00:36:47,000 However, as they raise the interest rate, it sort of puts the brakes on the economy. 322 00:36:47,000 --> 00:36:53,000 And then whenever there's a recession, they basically drop. 323 00:36:53,000 --> 00:36:59,000 The interest rate to try to stimulate growth. 324 00:36:59,000 --> 00:37:15,000 What's sort of a criticism, I think, of economists is they basically say that the Federal Reserve isn't really able to raise interest rates high enough prior to a recession. 325 00:37:15,000 --> 00:37:30,000 So like prior to a recession, if they were able to get the interest rates up to like four or five percent, then, you know, when there actually is a recession. 326 00:37:30,000 --> 00:37:34,000 There they can actually drop it sufficiently. 327 00:37:34,000 --> 00:37:42,000 Because here you see the recession's hit and they've dropped it to zero. 328 00:37:42,000 --> 00:37:53,000 But that's sort of the problem is when you drop it to zero, there's nowhere there's nowhere to go. 329 00:37:53,000 --> 00:37:55,000 Interesting. 330 00:37:55,000 --> 00:37:59,000 Well, that's that's essentially what they do. 331 00:37:59,000 --> 00:38:18,000 They they have to essentially use creative or not really creative, but sort of technical monetary policy to to sort of simulate negative interest rates. 332 00:38:18,000 --> 00:38:31,000 So to a certain extent, you know, big banks during this time may actually be paying effective negative interest rates. 333 00:38:31,000 --> 00:38:36,000 But it's it's an indirect tool. 334 00:38:36,000 --> 00:38:52,000 So that's that's sort of the criticism is, you know, it would be nice if they could just do it directly, you know, and so they call it quantitative easing where they give the the banks sort of effective negative interest rates. 335 00:38:52,000 --> 00:39:03,000 But like there's like I said, there's criticisms that it doesn't actually act the same as just moving the interest rate down. 336 00:39:03,000 --> 00:39:15,000 So long story short is it would be ideal if they could just, you know, dip the interest rate down sufficiently, you know, without hitting zero. 337 00:39:15,000 --> 00:39:23,000 But, you know, to do that, they would have to gotten the interest rates up a little bit. 338 00:39:23,000 --> 00:39:38,000 So long story short, that's what's going on at the federal level. But, you know, here in the cannabis industry, we're going to see if this is even affecting prices or not. 339 00:39:38,000 --> 00:39:59,000 Because in sales, because here's where I here's sort of the economic theory behind what's going on is you're basically saying that output. 340 00:39:59,000 --> 00:40:17,000 So you're basically saying, you know, output will do inflation is inflation and, you know, the interest rate. 341 00:40:17,000 --> 00:40:28,000 So you're basically saying all of these are sort of related to each other. 342 00:40:28,000 --> 00:40:44,000 Just sort of to color code this just so you can kind of see. 343 00:40:44,000 --> 00:40:59,000 So basically, we've talked in the previous weeks about essentially in autoregressive process. 344 00:40:59,000 --> 00:41:22,000 So like this. Right. So that's just like an autoregressive process where, you know, what output depends on, you know, output from the previous period. 345 00:41:22,000 --> 00:41:31,000 Here, we're doing a vector autoregression. 346 00:41:31,000 --> 00:41:52,000 And so basically, we're just saying we're just going to estimate these three equations simultaneously. So we're basically just going to say. 347 00:41:52,000 --> 00:42:05,000 Yeah. So basically, you're just we're just saying, OK, output today depends on output yesterday inflation yesterday and interest rates yesterday. 348 00:42:05,000 --> 00:42:10,000 We will measure the coefficient on each of these. 349 00:42:10,000 --> 00:42:18,000 So we're not necessarily saying that output depends that much on interest rates, but we're going to find out. 350 00:42:18,000 --> 00:42:31,000 And similarly, you know, inflation depends on the amount produced as well as, of course, you know, inflation from the previous period. 351 00:42:31,000 --> 00:42:40,000 And then economic theory suggests it also depends on the interest rates. 352 00:42:40,000 --> 00:43:00,000 And then and then this last equation is interesting because this is essentially leads us to wonder. 353 00:43:00,000 --> 00:43:20,000 So this is sort of similar to this is remember earlier, we said the central bank base that the interest function of the output gap and inflation. 354 00:43:20,000 --> 00:43:32,000 So basically, this is, you know, they call it like the Taylor rule. 355 00:43:32,000 --> 00:43:42,000 And this isn't exactly the Taylor rule. You actually have to like I said, you'd actually have to do the output gap, which which is what we're going to get through. 356 00:43:42,000 --> 00:43:59,000 So the output gap is. What you would expect output to minus output actually is and. 357 00:43:59,000 --> 00:44:07,000 Oh, if you're wondering what you expect, we'll get to that in one second. 358 00:44:07,000 --> 00:44:18,000 Well, actually, let's just go ahead and get to that right now. So the Federal Reserve, they want to set interest rates based on. 359 00:44:18,000 --> 00:44:33,000 What inflation they observe their past interest rates, as well as what the total output in the market was. 360 00:44:33,000 --> 00:44:43,000 Ideally, they want to know what the output gap was, what expectations were versus what the actual were. 361 00:44:43,000 --> 00:44:48,000 Well, we can forecast with. 362 00:44:48,000 --> 00:45:04,000 The statistical model is a theoretical model. You can apply it to any data set. 363 00:45:04,000 --> 00:45:18,000 You see here we're going to be packing in a lot of variables into each model. Also, we're going to be estimating. 364 00:45:18,000 --> 00:45:27,000 Three equations, so we're doing three equations times. 365 00:45:27,000 --> 00:45:36,000 So you're going to do an AR. 366 00:45:36,000 --> 00:45:38,000 You know. 367 00:45:38,000 --> 00:45:43,000 X. So, you know, so you may do like an AR rate. 368 00:45:43,000 --> 00:45:46,000 So if you wanted to do an AR six. 369 00:45:46,000 --> 00:45:59,000 Where you're going to have six parameters times three equations. So. You know, that's, you know, no less than. 370 00:45:59,000 --> 00:46:09,000 Plus, you're going to have the constants. So, you know, you're looking at like 21 plus degrees of freedom right there. 371 00:46:09,000 --> 00:46:21,000 So long story short is you're going to need a lot of data. If you really want to estimate the high order. 372 00:46:21,000 --> 00:46:24,000 Other aggressive model. 373 00:46:24,000 --> 00:46:35,000 But I'm wondering if you can do this for all the products. 374 00:46:35,000 --> 00:46:42,000 I imagine that'd be a huge model. Have to optimize it a lot, but seems like you could. 375 00:46:42,000 --> 00:46:53,000 What I would do here is essentially if you were doing that, you basically what I would do is you do y t of i. 376 00:46:53,000 --> 00:47:15,000 And then. Where I would be. 377 00:47:15,000 --> 00:47:23,000 At the point where you're doing all the products, I wonder if you're capturing inflation and interest rate. 378 00:47:23,000 --> 00:47:28,000 So. 379 00:47:28,000 --> 00:47:32,000 So essentially. 380 00:47:32,000 --> 00:47:34,000 Like you said you. 381 00:47:34,000 --> 00:47:41,000 Like, so say like i, for example, let's say like i is flower. 382 00:47:41,000 --> 00:47:54,000 Let's say this is an element of like flower or concentrate. 383 00:47:54,000 --> 00:47:58,000 It's getting a little messy. Let's just do this. 384 00:47:58,000 --> 00:48:05,000 So let's just say. 385 00:48:05,000 --> 00:48:13,000 Hi element of. 386 00:48:13,000 --> 00:48:18,000 Okay. 387 00:48:18,000 --> 00:48:22,000 So. 388 00:48:22,000 --> 00:48:32,000 Like I said, it just sort of dilutes the definition of inflation and I'm going to actually actually do a little research. So my expertise is micro economics. 389 00:48:32,000 --> 00:48:44,000 And so I was sort of reaching a little bit for the. So this is what this is sort of your your standard macro economic model. 390 00:48:44,000 --> 00:48:46,000 It would. 391 00:48:46,000 --> 00:48:59,000 Yeah, I'll have to do a little research because I think you could say maybe you could just do output of flower and then just do inflation of flower prices. 392 00:48:59,000 --> 00:49:06,000 However, the way I would just approach this. 393 00:49:06,000 --> 00:49:13,000 Right out of the gate would basically. 394 00:49:13,000 --> 00:49:18,000 You calculate the crude with no CPI. 395 00:49:18,000 --> 00:49:25,000 The consumer price index. 396 00:49:25,000 --> 00:49:48,000 And so that's basically like a weighted average of all your prices. 397 00:49:48,000 --> 00:50:17,000 We. 398 00:50:17,000 --> 00:50:21,000 So this how essentially. 399 00:50:21,000 --> 00:50:32,000 And then you have a weight for each good. And then you have a price for that good. 400 00:50:32,000 --> 00:50:40,000 Here we're doing. 401 00:50:40,000 --> 00:50:47,000 We're seeing. 402 00:50:47,000 --> 00:50:49,000 You're kind of breaking up Josh. 403 00:50:49,000 --> 00:50:51,000 Turn off your camera. 404 00:50:51,000 --> 00:50:52,000 Okay. 405 00:50:52,000 --> 00:50:57,000 The bandwidth works better. 406 00:50:57,000 --> 00:51:01,000 So I'm going to actually put together a bit better. 407 00:51:01,000 --> 00:51:03,000 Say that one more time. 408 00:51:03,000 --> 00:51:06,000 You're breaking up a bit. 409 00:51:06,000 --> 00:51:08,000 Oh. 410 00:51:08,000 --> 00:51:12,000 We have our basket is calculated. 411 00:51:12,000 --> 00:51:23,000 So basically this is where we get our our weights from. 412 00:51:23,000 --> 00:51:26,000 You hear me. 413 00:51:26,000 --> 00:51:31,000 Yeah, so maybe turn off the camera. You come with you come across much clearer. 414 00:51:31,000 --> 00:51:43,000 Sorry. I think it's just a maybe a bandwidth thing. So maybe I can close some of these tabs. 415 00:51:43,000 --> 00:51:44,000 Okay. 416 00:51:44,000 --> 00:51:46,000 Can you can you hear me okay. 417 00:51:46,000 --> 00:51:48,000 Yeah. Yeah. 418 00:51:48,000 --> 00:51:49,000 Okay. 419 00:51:49,000 --> 00:51:52,000 So just to sort of conclude. 420 00:51:52,000 --> 00:51:57,000 With the explanation of this inflation real quick. 421 00:51:57,000 --> 00:52:03,000 And I'll tidy up these notes for next week when we actually calculate this thing. 422 00:52:03,000 --> 00:52:06,000 So essentially. 423 00:52:06,000 --> 00:52:10,000 You have the sales by product type. 424 00:52:10,000 --> 00:52:13,000 And you have the total sales. 425 00:52:13,000 --> 00:52:25,000 So we're we're seeing that okay consumers are spending about 60 percent of their money on flower and 40 percent on concentrate. 426 00:52:25,000 --> 00:52:28,000 So. 427 00:52:28,000 --> 00:52:40,000 You can say basically somebody's basket of goods is going to be 60 percent flower and 40 percent concentrate. 428 00:52:40,000 --> 00:52:48,000 So that's where you get the the weights from. 429 00:52:48,000 --> 00:52:56,000 And then you have the actual prices. So you have the actual price of flower. 430 00:52:56,000 --> 00:53:01,000 He of I and then you have the actual price of. 431 00:53:01,000 --> 00:53:04,000 Concentrate P of I. 432 00:53:04,000 --> 00:53:06,000 So then you basic. 433 00:53:06,000 --> 00:53:08,000 So you then can calculate the CPI. 434 00:53:08,000 --> 00:53:12,000 Which is basically. 435 00:53:12,000 --> 00:53:19,000 So this is the you know consumer. 436 00:53:19,000 --> 00:53:22,000 Price. 437 00:53:22,000 --> 00:53:25,000 Index. 438 00:53:25,000 --> 00:53:28,000 And you can think of it. 439 00:53:28,000 --> 00:53:31,000 You may have heard of index funds. 440 00:53:31,000 --> 00:53:40,000 And what an index fund is is they basically take a weighted average of a hand like a basket of stocks. 441 00:53:40,000 --> 00:53:49,000 And then you can invest in that basket of stocks through the index. 442 00:53:49,000 --> 00:53:57,000 Exactly. So that's like sort of a basket of the top five hundred. 443 00:53:57,000 --> 00:54:02,000 Largest companies essentially. 444 00:54:02,000 --> 00:54:10,000 And maybe weighted. 445 00:54:10,000 --> 00:54:14,000 I need to touch up on some things like that. But. 446 00:54:14,000 --> 00:54:18,000 But exactly. So that's an example of an index. 447 00:54:18,000 --> 00:54:23,000 Into the consumer price index is just an index of prices. 448 00:54:23,000 --> 00:54:26,000 So. 449 00:54:26,000 --> 00:54:34,000 The actual number is. 450 00:54:34,000 --> 00:54:41,000 Like the nominal value is not that important so much as the direction. 451 00:54:41,000 --> 00:54:46,000 So with this like the nominal value is. 452 00:54:46,000 --> 00:54:53,000 Doesn't mean too much what is important is inflation. 453 00:54:53,000 --> 00:55:01,000 And so that would basically be the CPI at time T. 454 00:55:01,000 --> 00:55:07,000 Right. So basically this is the CPI at time T. 455 00:55:07,000 --> 00:55:11,000 So you would basically do the CPI at time T. 456 00:55:11,000 --> 00:55:16,000 Minus the CPI at time T minus one. 457 00:55:16,000 --> 00:55:20,000 Divided by the CPI at T minus one. 458 00:55:20,000 --> 00:55:25,000 And that's how you calculate inflation. 459 00:55:25,000 --> 00:55:31,000 And our hypothesis. 460 00:55:31,000 --> 00:55:35,000 Is that it would be, you know, one to three percent. 461 00:55:35,000 --> 00:55:42,000 And that's my hypothesis for this time frame. 462 00:55:42,000 --> 00:55:45,000 That's my hypothesis. 463 00:55:45,000 --> 00:55:48,000 And next week. 464 00:55:48,000 --> 00:55:53,000 Next week we can actually use. 465 00:55:53,000 --> 00:55:58,000 This data. 466 00:55:58,000 --> 00:56:01,000 I must have saved it in some weird place. 467 00:56:01,000 --> 00:56:05,000 But. 468 00:56:05,000 --> 00:56:09,000 I gathered the data from. 469 00:56:09,000 --> 00:56:12,000 You know the Oregon. 470 00:56:12,000 --> 00:56:14,000 State dashboard. 471 00:56:14,000 --> 00:56:16,000 And so next week. 472 00:56:16,000 --> 00:56:22,000 We can dive in and basically calculate the. 473 00:56:22,000 --> 00:56:25,000 Inflation rate. 474 00:56:25,000 --> 00:56:26,000 Then. 475 00:56:26,000 --> 00:56:28,000 Sorry for this mess. 476 00:56:28,000 --> 00:56:32,000 Then we'll actually estimate the VAR model. 477 00:56:32,000 --> 00:56:37,000 The VAR model using. 478 00:56:37,000 --> 00:56:41,000 Using output sales. 479 00:56:41,000 --> 00:56:43,000 Inflation. 480 00:56:43,000 --> 00:56:46,000 Which is a function of price. 481 00:56:46,000 --> 00:56:52,000 And our third variable will be the federal funds rate. 482 00:56:52,000 --> 00:56:56,000 Which will be our proxy for the interest rate. 483 00:56:56,000 --> 00:57:00,000 And we're going to see how much interest rates. 484 00:57:00,000 --> 00:57:01,000 Affect. 485 00:57:01,000 --> 00:57:02,000 Output. 486 00:57:02,000 --> 00:57:04,000 And prices in the. 487 00:57:04,000 --> 00:57:07,000 In the Oregon cannabis market. 488 00:57:07,000 --> 00:57:12,000 And. 489 00:57:12,000 --> 00:57:15,000 The way we're going to see this through. 490 00:57:15,000 --> 00:57:18,000 Is we're going to essentially measure the output gap. 491 00:57:18,000 --> 00:57:26,000 So. 492 00:57:26,000 --> 00:57:31,000 We're essentially going to use our VAR model. 493 00:57:31,000 --> 00:57:35,000 From earlier and forecast. 494 00:57:35,000 --> 00:57:36,000 Output. 495 00:57:36,000 --> 00:57:38,000 Inflation. 496 00:57:38,000 --> 00:57:39,000 And the interest rate. 497 00:57:39,000 --> 00:57:41,000 Into. 498 00:57:41,000 --> 00:57:43,000 You know. 499 00:57:43,000 --> 00:57:51,000 For the remainder of 2021 and perhaps into 2022. 500 00:57:51,000 --> 00:57:53,000 So we know what we're forecasting. 501 00:57:53,000 --> 00:57:56,000 We know the purpose of forecasting. 502 00:57:56,000 --> 00:58:02,000 And so next week. 503 00:58:02,000 --> 00:58:04,000 We will be. 504 00:58:04,000 --> 00:58:06,000 Finishing up. 505 00:58:06,000 --> 00:58:10,000 And we will forecast. 506 00:58:10,000 --> 00:58:13,000 The interest rate. 507 00:58:13,000 --> 00:58:18,000 Using this VAR model. 508 00:58:18,000 --> 00:58:20,000 So. 509 00:58:20,000 --> 00:58:26,000 Sorry that it was a little. 510 00:58:26,000 --> 00:58:29,000 All over the place today. 511 00:58:29,000 --> 00:58:31,000 It needs to be a little more structured. 512 00:58:31,000 --> 00:58:33,000 So for next week. 513 00:58:33,000 --> 00:58:35,000 I'm going to structure the presentation a little better. 514 00:58:35,000 --> 00:58:37,000 And actually. 515 00:58:37,000 --> 00:58:40,000 I had a feeling the inflation was going to take just a little bit. 516 00:58:40,000 --> 00:58:43,000 Just to kind of get. 517 00:58:43,000 --> 00:58:46,000 You know, just to explain what's going on there. 518 00:58:46,000 --> 00:58:48,000 But. 519 00:58:48,000 --> 00:58:51,000 That's the crux of our analysis. 520 00:58:51,000 --> 00:58:53,000 So next week we can actually. 521 00:58:53,000 --> 00:58:55,000 Calculate the inflation rate. 522 00:58:55,000 --> 00:59:02,000 And forecast it forward for the share on top. 523 00:59:02,000 --> 00:59:07,000 So. 524 00:59:07,000 --> 00:59:11,000 Do you have any quick questions or anything Josh? 525 00:59:11,000 --> 00:59:14,000 Off the top of my head. 526 00:59:14,000 --> 00:59:19,000 Thoughts or Charles? 527 00:59:19,000 --> 00:59:20,000 No. 528 00:59:20,000 --> 00:59:24,000 It was a cool presentation. 529 00:59:24,000 --> 00:59:27,000 Great to see next week when it kind of all comes together. 530 00:59:27,000 --> 00:59:32,000 So I started to put together some Python scripts just to calculate it. 531 00:59:32,000 --> 00:59:33,000 And. 532 00:59:33,000 --> 00:59:36,000 Like I said, there's a little bit more data to. 533 00:59:36,000 --> 00:59:37,000 Assemble. 534 00:59:37,000 --> 00:59:38,000 But. 535 00:59:38,000 --> 00:59:40,000 All the pieces are there. 536 00:59:40,000 --> 00:59:42,000 And. 537 00:59:42,000 --> 00:59:46,000 I don't know if anyone who's done this analysis before. 538 00:59:46,000 --> 00:59:47,000 Or at least I haven't seen it. 539 00:59:47,000 --> 00:59:49,000 If you've seen it. 540 00:59:49,000 --> 00:59:52,000 Send it my way because I'm interested in this. 541 00:59:52,000 --> 00:59:54,000 But. 542 00:59:54,000 --> 00:59:58,000 I'm not sure if you can calculate it or if you're ambitious. 543 00:59:58,000 --> 01:00:01,000 Check out the GitHub. 544 01:00:01,000 --> 01:00:02,000 The analytics. 545 01:00:02,000 --> 01:00:04,000 Anonymous data science. 546 01:00:04,000 --> 01:00:06,000 GitHub repository. 547 01:00:06,000 --> 01:00:07,000 And. 548 01:00:07,000 --> 01:00:10,000 You know, you could get a head start. 549 01:00:10,000 --> 01:00:13,000 But essentially we're going to forecast this. 550 01:00:13,000 --> 01:00:14,000 These series. 551 01:00:14,000 --> 01:00:16,000 And. 552 01:00:16,000 --> 01:00:18,000 We're going to be enlightened. 553 01:00:18,000 --> 01:00:19,000 And we're. 554 01:00:19,000 --> 01:00:20,000 You know, we'll be. 555 01:00:20,000 --> 01:00:21,000 You know. 556 01:00:21,000 --> 01:00:22,000 Some of the. 557 01:00:22,000 --> 01:00:24,000 Some of the most interesting. 558 01:00:24,000 --> 01:00:25,000 And. 559 01:00:25,000 --> 01:00:27,000 Anonymous prices in Oregon. 560 01:00:27,000 --> 01:00:29,000 So it'll be. 561 01:00:29,000 --> 01:00:32,000 Exciting. 562 01:00:32,000 --> 01:00:35,000 All right, team. 563 01:00:35,000 --> 01:00:37,000 Glad to have you aboard, Josh. 564 01:00:37,000 --> 01:00:38,000 I hope you enjoyed it. 565 01:00:38,000 --> 01:00:40,000 You know, let us know your feedback. 566 01:00:40,000 --> 01:00:41,000 What you liked. 567 01:00:41,000 --> 01:00:42,000 What could be improved. 568 01:00:42,000 --> 01:00:44,000 And. 569 01:00:44,000 --> 01:00:48,000 I'm mostly working on sparse neural networks, by the way. 570 01:00:48,000 --> 01:00:49,000 So. 571 01:00:49,000 --> 01:00:51,000 Neural networks. 572 01:00:51,000 --> 01:00:52,000 And. 573 01:00:52,000 --> 01:00:54,000 The new layer in PyTorch. 574 01:00:54,000 --> 01:00:55,000 So. 575 01:00:55,000 --> 01:00:56,000 I'd be good at. 576 01:00:56,000 --> 01:00:57,000 Doing. 577 01:00:57,000 --> 01:01:00,000 I guess sparse versions of huge data sets. 578 01:01:00,000 --> 01:01:04,000 So like that whole market thing I mentioned. 579 01:01:04,000 --> 01:01:08,000 But for now I'm just going to work on this thing. 580 01:01:08,000 --> 01:01:09,000 For the most part. 581 01:01:09,000 --> 01:01:12,000 Because I just want to get a little robot that can move around. 582 01:01:12,000 --> 01:01:13,000 Finally. 583 01:01:13,000 --> 01:01:14,000 We've got. 584 01:01:14,000 --> 01:01:16,000 We've got huge data sets for you. 585 01:01:16,000 --> 01:01:17,000 So you may want to. 586 01:01:17,000 --> 01:01:19,000 Get in touch with Charles. 587 01:01:19,000 --> 01:01:22,000 He's leading the front. 588 01:01:22,000 --> 01:01:23,000 He's spearheading this one. 589 01:01:23,000 --> 01:01:25,000 So. 590 01:01:25,000 --> 01:01:26,000 Definitely reach out. 591 01:01:26,000 --> 01:01:27,000 Get in touch. 592 01:01:27,000 --> 01:01:29,000 And. 593 01:01:29,000 --> 01:01:30,000 That's exciting. 594 01:01:30,000 --> 01:01:32,000 PyTorch is an awesome tool. 595 01:01:32,000 --> 01:01:35,000 So I want to see your contributions on that. 596 01:01:35,000 --> 01:01:36,000 That's awesome. 597 01:01:36,000 --> 01:01:39,000 Yeah, if I can get some. 598 01:01:39,000 --> 01:01:44,000 I don't know about the scale of data where it's hard to fit on my computer. 599 01:01:44,000 --> 01:01:49,000 But if I can get it streaming in some way, then I might be able to work with that. 600 01:01:49,000 --> 01:01:50,000 Yeah. 601 01:01:50,000 --> 01:01:51,000 Are you going. 602 01:01:51,000 --> 01:01:52,000 Are you. 603 01:01:52,000 --> 01:01:55,000 Have you attended any of the GTC conferences this week? 604 01:01:55,000 --> 01:01:57,000 I have not. 605 01:01:57,000 --> 01:01:58,000 Oh yeah. 606 01:01:58,000 --> 01:01:59,000 That has been on my schedule. 607 01:01:59,000 --> 01:02:00,000 GTC. 608 01:02:00,000 --> 01:02:01,000 Yeah. 609 01:02:01,000 --> 01:02:02,000 There. 610 01:02:02,000 --> 01:02:05,000 There have been some good presentations on PyTorch. 611 01:02:05,000 --> 01:02:08,000 There have been some really cool presentations on. 612 01:02:08,000 --> 01:02:11,000 You know, self-supervised learning and robots. 613 01:02:11,000 --> 01:02:12,000 And. 614 01:02:12,000 --> 01:02:13,000 Oh yeah. 615 01:02:13,000 --> 01:02:14,000 It's just, it's amazing. 616 01:02:14,000 --> 01:02:15,000 And it's. 617 01:02:15,000 --> 01:02:16,000 It's here. 618 01:02:16,000 --> 01:02:17,000 So. 619 01:02:17,000 --> 01:02:21,000 And it's going on through Friday still. 620 01:02:21,000 --> 01:02:22,000 So. 621 01:02:22,000 --> 01:02:23,000 That's right. 622 01:02:23,000 --> 01:02:24,000 My whole. 623 01:02:24,000 --> 01:02:25,000 Like all my. 624 01:02:25,000 --> 01:02:28,000 What presentations this week. 625 01:02:28,000 --> 01:02:31,000 I'll probably add some of those to my schedule then. 626 01:02:31,000 --> 01:02:33,000 Oh yeah. 627 01:02:33,000 --> 01:02:35,000 Tuesday, Wednesday, Thursday. 628 01:02:35,000 --> 01:02:37,000 Yep. 629 01:02:37,000 --> 01:02:39,000 Deep learning for autonomous vehicles. 630 01:02:39,000 --> 01:02:41,000 Accelerating CUDA. 631 01:02:41,000 --> 01:02:43,000 I don't want to use. 632 01:02:43,000 --> 01:02:44,000 CUDA necessarily. 633 01:02:44,000 --> 01:02:48,000 Cause I got a ton of AMD GPS. 634 01:02:48,000 --> 01:02:49,000 So. 635 01:02:49,000 --> 01:02:50,000 Yeah. 636 01:02:50,000 --> 01:02:53,000 Oh. 637 01:02:53,000 --> 01:02:56,000 Are they in PCs or Mac? 638 01:02:56,000 --> 01:02:58,000 PCs. 639 01:02:58,000 --> 01:03:00,000 But there's a Vulkan. 640 01:03:00,000 --> 01:03:04,000 Compute repository that I'm trying to work with, but. 641 01:03:04,000 --> 01:03:06,000 That's got to catch up a bit. 642 01:03:06,000 --> 01:03:08,000 Yeah. 643 01:03:08,000 --> 01:03:10,000 Yeah. 644 01:03:10,000 --> 01:03:11,000 So. 645 01:03:11,000 --> 01:03:13,000 TensorFlow is going to start. 646 01:03:13,000 --> 01:03:14,000 And. 647 01:03:14,000 --> 01:03:17,000 It's going to start working on the Mac pretty soon. 648 01:03:17,000 --> 01:03:21,000 Like the next version that comes out, but yeah, I don't know about like. 649 01:03:21,000 --> 01:03:23,000 AMD stuff. 650 01:03:23,000 --> 01:03:24,000 Yeah. 651 01:03:24,000 --> 01:03:28,000 Other than working with metal before Vulkan. 652 01:03:28,000 --> 01:03:30,000 Yeah. 653 01:03:30,000 --> 01:03:31,000 That's. 654 01:03:31,000 --> 01:03:32,000 Surprising. 655 01:03:32,000 --> 01:03:35,000 That's been like a two year push. 656 01:03:35,000 --> 01:03:39,000 Doesn't seem like a good move to me, to be honest. 657 01:03:39,000 --> 01:03:43,000 But I think AMD should be able to support all GPUs and. 658 01:03:43,000 --> 01:03:46,000 Metal is just one very specific sector. 659 01:03:46,000 --> 01:03:47,000 Yeah. 660 01:03:47,000 --> 01:03:48,000 I don't know. 661 01:03:48,000 --> 01:03:52,000 I think Apple pushed for it and. 662 01:03:52,000 --> 01:03:53,000 Yeah. 663 01:03:53,000 --> 01:03:56,000 There was like Plaid ML, but I've never gotten that to work. 664 01:03:56,000 --> 01:04:00,000 I'll just try to get Vulkan working and then. 665 01:04:00,000 --> 01:04:05,000 If that works well enough, the TensorFlow stuff would be obsolete because. 666 01:04:05,000 --> 01:04:08,000 This would just work on so many more systems. 667 01:04:08,000 --> 01:04:09,000 Yeah. 668 01:04:09,000 --> 01:04:10,000 So. 669 01:04:10,000 --> 01:04:13,000 Question, Josh. 670 01:04:13,000 --> 01:04:18,000 Are you specifically trying to apply machine learning to the chems industry or are you 671 01:04:18,000 --> 01:04:21,000 just spearheading machine learning? 672 01:04:21,000 --> 01:04:23,000 As is. 673 01:04:23,000 --> 01:04:26,000 Right now just spearheading machine learning as is. 674 01:04:26,000 --> 01:04:32,000 And honestly, I'm kind of just going to all the different meetups I can find on AI to 675 01:04:32,000 --> 01:04:35,000 see what different people are working on. 676 01:04:35,000 --> 01:04:38,000 So it's so interesting. 677 01:04:38,000 --> 01:04:43,000 So my background was just in statistics and economics and. 678 01:04:43,000 --> 01:04:47,000 It seems. 679 01:04:47,000 --> 01:04:51,000 I get the principle behind machine learning like, you know. 680 01:04:51,000 --> 01:04:55,000 I think of it more as just framing the right problem. 681 01:04:55,000 --> 01:05:01,000 But I'm familiar with a lot of the tools because a lot of the tools used are essentially 682 01:05:01,000 --> 01:05:07,000 statistics models, so I know the models and how did. 683 01:05:07,000 --> 01:05:11,000 You know, estimate the models and use them. 684 01:05:11,000 --> 01:05:17,000 But actually framing the machine learning problem is something of an art, I think. 685 01:05:17,000 --> 01:05:20,000 Yeah, it's wrangling, right? 686 01:05:20,000 --> 01:05:27,000 I think that I've looked at even biology and that sort of framing the problem thing is 687 01:05:27,000 --> 01:05:28,000 still a thing there. 688 01:05:28,000 --> 01:05:34,000 But like you have four year transforms in the ear, you have retinal ganglion cells that 689 01:05:34,000 --> 01:05:42,000 do crazy stuff in the eye and like as no matter how advanced you go, there seems to be data 690 01:05:42,000 --> 01:05:44,000 wrangling just everywhere. 691 01:05:44,000 --> 01:05:52,000 I mean, I need to get you in touch with. 692 01:05:52,000 --> 01:05:57,000 With a couple people. 693 01:05:57,000 --> 01:05:59,000 There's something that comes to mind who's working on them. 694 01:05:59,000 --> 01:06:04,000 You know, a machine learning project, so may have to get you in touch with them. 695 01:06:04,000 --> 01:06:06,000 And then. 696 01:06:06,000 --> 01:06:12,000 If you're interested in analytics is specifically trying to help laboratories and people with 697 01:06:12,000 --> 01:06:15,000 do cannabis analytics. 698 01:06:15,000 --> 01:06:21,000 It's open source, so if you wanted to find a way to plug in some. 699 01:06:21,000 --> 01:06:24,000 She learning uses. 700 01:06:24,000 --> 01:06:29,000 You're welcome to you're welcome to use that as your host. 701 01:06:29,000 --> 01:06:31,000 That'd be cool. 702 01:06:31,000 --> 01:06:37,000 Yeah, the main thing I want to know right now should just be able to do a bit more data. 703 01:06:37,000 --> 01:06:40,000 It's not really the data wrangling part, but. 704 01:06:40,000 --> 01:06:44,000 Yeah, it'd be interesting. 705 01:06:44,000 --> 01:06:46,000 Yeah. 706 01:06:46,000 --> 01:06:51,000 You may have to shoot me whatever you work on in PyTorch because like I said, that's a 707 01:06:51,000 --> 01:06:54,000 tool that I've heard of and. 708 01:06:54,000 --> 01:07:00,000 I wanted to add it to my toolbox, but it's not in my toolbox yet, so always have it to 709 01:07:00,000 --> 01:07:01,000 learn. 710 01:07:01,000 --> 01:07:04,000 Pretty close to Texas. 711 01:07:04,000 --> 01:07:05,000 Same thing with TensorFlow. 712 01:07:05,000 --> 01:07:12,000 I would almost like almost got it in my toolbox, but it's still not in my toolbox yet, so I 713 01:07:12,000 --> 01:07:14,000 would like to add them both. 714 01:07:14,000 --> 01:07:16,000 For now. 715 01:07:16,000 --> 01:07:22,000 OK. 716 01:07:22,000 --> 01:07:31,000 But so perhaps when we one of you and Charles and Josh, you're welcome to present at the 717 01:07:31,000 --> 01:07:33,000 meetup if you want about. 718 01:07:33,000 --> 01:07:38,000 Hence your flow or PyTorch or you know some sort of machine learning. 719 01:07:38,000 --> 01:07:45,000 Application or tools you're welcome to because like I said, I'm actually eager to learn about 720 01:07:45,000 --> 01:07:46,000 those topics. 721 01:07:46,000 --> 01:07:49,000 Certainly when I get the data. 722 01:07:49,000 --> 01:07:53,000 I think GASP will be a huge help to everybody. 723 01:07:53,000 --> 01:07:59,000 But you know, it's just a lot of experimenting and. 724 01:07:59,000 --> 01:08:03,000 And you know, and finally, and I haven't figured it right. 725 01:08:03,000 --> 01:08:10,000 Well, why don't you take some notes Charles and then when you get it up and running, why 726 01:08:10,000 --> 01:08:18,000 don't you let me know and you can host a meetup and then you could show us about to ask and 727 01:08:18,000 --> 01:08:24,000 you could save us all a bunch of time by telling us all the tips and tricks for getting your 728 01:08:24,000 --> 01:08:26,000 configuration just right. 729 01:08:26,000 --> 01:08:27,000 Yeah, yeah. 730 01:08:27,000 --> 01:08:31,000 I'm also curious about GASP versus Pandas. 731 01:08:31,000 --> 01:08:33,000 GASP versus Pandas? 732 01:08:33,000 --> 01:08:36,000 Yeah, is it a different data scale or? 733 01:08:36,000 --> 01:08:40,000 So GASP is basically. 734 01:08:40,000 --> 01:08:42,000 It works. 735 01:08:42,000 --> 01:08:47,000 It's basically a drop in replacement for Pandas. 736 01:08:47,000 --> 01:08:51,000 Instead of like reading everything in at once. 737 01:08:51,000 --> 01:08:53,000 It delays the computation. 738 01:08:53,000 --> 01:08:57,000 And so like when you read a file, it's like instantaneous. 739 01:08:57,000 --> 01:09:03,000 But then when you like, you know, if you want to look at the head of the file, it will or the data frame, 740 01:09:03,000 --> 01:09:06,000 it will, it'll take time, it'll compute it. 741 01:09:06,000 --> 01:09:09,000 But it also works in parallel. 742 01:09:09,000 --> 01:09:10,000 It'll do it off. 743 01:09:10,000 --> 01:09:15,000 It'll do a lot of work across multiple threads where Pandas is single thread. 744 01:09:15,000 --> 01:09:18,000 So it's optimized Pandas. 745 01:09:18,000 --> 01:09:20,000 Yes. 746 01:09:20,000 --> 01:09:25,000 It's like almost like the right tool for the right job. 747 01:09:25,000 --> 01:09:32,000 So it's I think that's it sounds to me like the perfect tool for basically the very first 748 01:09:32,000 --> 01:09:36,000 meetups we started to try to get into this Washington state data. 749 01:09:36,000 --> 01:09:42,000 So that's sort of the sounds like the tool for that job. 750 01:09:42,000 --> 01:09:43,000 Yeah. 751 01:09:43,000 --> 01:09:46,000 And then if you have multiple machines, it's tribute. 752 01:09:46,000 --> 01:09:49,000 They'll use the memory on time. 753 01:09:49,000 --> 01:09:51,000 It will be good. 754 01:09:51,000 --> 01:09:55,000 It will use the memory across multiple machines to distribute. 755 01:09:55,000 --> 01:09:58,000 So. 756 01:09:58,000 --> 01:10:02,000 Could potentially use it in production to do some interesting things. 757 01:10:02,000 --> 01:10:05,000 Yeah. 758 01:10:05,000 --> 01:10:10,000 But you can use the same code on a laptop as you can across a cluster. 759 01:10:10,000 --> 01:10:14,000 So it just automatically scales for you. 760 01:10:14,000 --> 01:10:22,000 Or at least that's what they claim. 761 01:10:22,000 --> 01:10:30,000 Well, I think I'm going to head on out myself. 762 01:10:30,000 --> 01:10:36,000 But it's been awesome talking with you guys about data science. 763 01:10:36,000 --> 01:10:43,000 It's always it's awesome to find like minded people to talk about data, machine learning, 764 01:10:43,000 --> 01:10:45,000 what have you. 765 01:10:45,000 --> 01:10:47,000 So it's fun. 766 01:10:47,000 --> 01:10:48,000 Yeah. 767 01:10:48,000 --> 01:10:52,000 It's been a great meetup. 768 01:10:52,000 --> 01:10:53,000 All right. 769 01:10:53,000 --> 01:10:54,000 All right, guys. 770 01:10:54,000 --> 01:10:56,000 I'm going to go ahead and head on out. 771 01:10:56,000 --> 01:11:00,000 I've got a full agenda today of things to accomplish. 772 01:11:00,000 --> 01:11:04,000 But for next week, we can talk more machine learning. 773 01:11:04,000 --> 01:11:10,000 I'll have a tidier little demonstration of how to do inflation. 774 01:11:10,000 --> 01:11:14,000 And we can take it from the top. 775 01:11:14,000 --> 01:11:15,000 OK, great. 776 01:11:15,000 --> 01:11:16,000 See you next week. 777 01:11:16,000 --> 01:11:19,000 See you. 778 01:11:19,000 --> 01:11:20,000 Bye now. 779 01:11:20,000 --> 01:11:21,000 Bye. 780 01:11:21,000 --> 01:11:47,000 Bye.