#467: Data Science Panel at PyCon 2024 の全文書き起こし

00:03:22.920 And now I've started to work in the ML realm. 00:03:22.920 --> 00:03:24.100 So really exciting. 00:03:24.100 --> 00:03:25.600 Yeah, very, very interesting. 00:03:25.600 --> 00:03:27.900 Another thing I absolutely love is coffee. 00:03:27.900 --> 00:03:28.500 Yeah. 00:03:28.500 --> 00:03:30.460 Oh my gosh, coffee is so good. 00:03:31.080 --> 00:03:32.580 I think we're running on it at PyCon. 00:03:32.580 --> 00:03:33.760 Pretty much we are. 00:03:33.760 --> 00:03:34.020 Yeah. 00:03:34.020 --> 00:03:36.600 We're getting farther into the show 00:03:36.600 --> 00:03:38.260 and more coffee is needed. 00:03:38.260 --> 00:03:40.480 But I do want to ask you, you know, 00:03:40.480 --> 00:03:43.400 what do you think about being in the data science space? 00:03:43.400 --> 00:03:45.300 That's a really different world 00:03:45.300 --> 00:03:46.800 that interacting with people all day 00:03:46.800 --> 00:03:49.380 and working with your hands more or whatever. 00:03:49.520 --> 00:03:52.180 like how has it been with this switch? 00:03:52.180 --> 00:03:53.960 There is a lot of synergies actually 00:03:53.960 --> 00:03:55.860 when you're stood behind the espresso machine 00:03:55.860 --> 00:03:57.180 and you're getting all the orders in 00:03:57.180 --> 00:03:59.220 and then you need to like problem solve 00:03:59.220 --> 00:04:02.040 to like how you get everyone their correct order 00:04:02.040 --> 00:04:03.560 to the way that they like it. 00:04:03.560 --> 00:04:06.700 So there was a lot of transferable skills, I will say. 00:04:06.700 --> 00:04:09.220 But I think what I found really powerful 00:04:09.220 --> 00:04:13.920 and especially maybe learning at this specific period of time 00:04:13.920 --> 00:04:16.640 is how accessible a lot of the tools are today. 00:04:16.640 --> 00:04:19.040 So like how, I won't say easy 00:04:19.040 --> 00:04:20.960 because I put a lot of hard work into it, 00:04:20.960 --> 00:04:23.820 but like how possible it is 00:04:23.820 --> 00:04:26.280 even with a background like mine to get into the field. 00:04:26.280 --> 00:04:26.900 Awesome. 00:04:26.900 --> 00:04:29.720 I switched, I didn't have a formal education either. 00:04:29.720 --> 00:04:32.440 I took two computer college courses 00:04:32.440 --> 00:04:34.480 just because they match, you know, 00:04:34.480 --> 00:04:35.380 I needed for something else. 00:04:35.380 --> 00:04:37.760 And yeah, I thought it, 00:04:37.760 --> 00:04:41.300 I think you can completely succeed here teaching yourself. 00:04:41.300 --> 00:04:42.300 There's so many resources. 00:04:42.840 --> 00:04:44.520 Honestly, the problem is what resources 00:04:44.520 --> 00:04:46.260 do you choose to learn these days, right? 00:04:46.260 --> 00:04:47.320 You can spend all your time. 00:04:47.320 --> 00:04:48.580 Well, I'm doing another tutorial. 00:04:48.580 --> 00:04:49.440 I'm doing another class. 00:04:49.440 --> 00:04:51.660 Like some point you got to start doing something, right? 00:04:51.660 -->

00:04:52.100 Yeah. 00:04:52.100 --> 00:04:54.340 And I think actually it felt like that 00:04:54.340 --> 00:04:56.120 probably when we all started. 00:04:56.120 --> 00:04:56.500 Yeah. 00:04:56.500 --> 00:04:59.440 So data science was just getting hot when I started. 00:04:59.440 --> 00:05:01.420 And oh my God, back when I started, 00:05:01.420 --> 00:05:02.540 this is how long ago it was. 00:05:02.540 --> 00:05:04.320 There were actually like those articles, 00:05:04.320 --> 00:05:05.580 like R versus Python. 00:05:05.580 --> 00:05:07.560 Like there's no conversations anyone's having anymore, 00:05:07.560 --> 00:05:08.880 but they have similar conversations. 00:05:08.880 --> 00:05:11.220 And I think it makes it super difficult for beginners 00:05:11.220 --> 00:05:13.600 because the field felt inaccessible, 00:05:13.600 --> 00:05:15.080 I think, eight years ago. 00:05:15.080 --> 00:05:18.520 The field feels very hostile to beginners right now, 00:05:18.520 --> 00:05:20.020 I think, because of the AI hype. 00:05:20.020 --> 00:05:23.080 I don't actually think the field has changed that much 00:05:23.080 --> 00:05:24.220 in fundamentals. 00:05:24.220 --> 00:05:27.280 It's just NLP has become a bigger thing 00:05:27.280 --> 00:05:28.680 in computer vision recently, 00:05:28.680 --> 00:05:30.320 but we can get into that later. 00:05:30.320 --> 00:05:32.440 Yeah, I completely agree with you. 00:05:32.440 --> 00:05:34.780 To be honest, for me, 00:05:34.780 --> 00:05:38.100 data science is super broad world, 00:05:38.100 --> 00:05:42.220 full of a lot of things that are kind of popping up, 00:05:42.220 --> 00:05:44.880 doing different evolution during time. 00:05:44.880 --> 00:05:48.440 And it's so interesting to see the evolution 00:05:48.440 --> 00:05:50.080 in the last eight years. 00:05:50.260 --> 00:05:53.880 I started eight years ago in data science. 00:05:53.880 --> 00:05:55.320 And I remember when, 00:05:55.320 --> 00:05:58.320 how I was doing things eight years ago 00:05:58.320 --> 00:06:00.240 and how I'm doing things now. 00:06:00.240 --> 00:06:02.240 And I love it. 00:06:02.240 --> 00:06:03.760 I love to see this progression. 00:06:03.760 --> 00:06:07.420 And I am pretty sure that in eight more years, 00:06:07.420 --> 00:06:10.860 we're going to be in something completely different 00:06:10.860 --> 00:06:12.000 and super exciting. 00:06:12.000 --> 00:06:13.660 Yeah, I totally agree with that. 00:06:13.660 --> 00:06:14.180 I do. 00:06:14.180 --> 00:06:16.260 And I also think data science is interesting 00:06:16.260 --> 00:06:18.140 because coming into it, 00:06:18.140 --> 00:06:20.000 you can be a data scientist, 00:06:20.000 --> 00:06:22.180 but because some other reason, right? 00:06:22.180 --> 00:06:23.220 I could be a data scientist 00:06:23.220 --> 00:06:25.460 because I'm interested in biology 00:06:25.460 --> 00:06:27.180 or sustainability or something. 00:06:27.180 --> 00:06:29.340 Whereas if you're a web developer 00:06:29.340 --> 00:06:31.280 or you build APIs or you optimize, 00:06:31.280 --> 00:06:33.040 you know, whatever, 00:06:33.040 --> 00:06:34.880 you're more focused on, 00:06:34.880 --> 00:06:36.720 I care about the thing, the code itself, 00:06:36.720 --> 00:06:38.000 rather than I'm trying to, 00:06:38.000 --> 00:06:38.700 I care about that. 00:06:38.700 --> 00:06:40.500 And this is a tool to address that. 00:06:40.500 --> 00:06:41.300 Yeah, yeah. 00:06:41.520 --> 00:06:42.920 Yeah, actually, I was going to say, 00:06:42.920 --> 00:06:44.560 I met a bioinformatician yesterday. 00:06:44.560 --> 00:06:46.700 Like, that's also a data scientist, 00:06:46.700 --> 00:06:48.460 like someone who works in genetic data. 00:06:48.460 --> 00:06:49.660 Yeah, absolutely. 00:06:49.660 --> 00:06:50.900 I had a comment from, 00:06:50.900 --> 00:06:52.400 I did a show recently 00:06:52.400 --> 00:06:56.360 from about how Python's used in neurology labs, right? 00:06:56.360 --> 00:06:57.100 And somebody wrote me, 00:06:57.100 --> 00:06:58.060 this is my favorite episode. 00:06:58.060 --> 00:06:58.800 It speaks to me. 00:06:58.800 --> 00:06:59.880 I'm also a neurologist. 00:06:59.880 --> 00:07:00.900 You know, like, it's really cool. 00:07:00.900 --> 00:07:02.220 All right, we're looking out, 00:07:02.220 --> 00:07:04.200 kind of the backside a little bit, 00:07:04.200 --> 00:07:07.820 but we're looking out of the expo hall here at PyCon. 00:07:07.820 --> 00:07:09.920 So I don't know how about you all feel, 00:07:09.920 --> 00:07:12.380 but for me, this is like my geek holiday. 00:07:12.380 --> 00:07:13.440 I get to come here, 00:07:13.440 --> 00:07:15.480 and it's really special to me 00:07:15.480 --> 00:07:17.340 because I get to see my friends 00:07:17.340 --> 00:07:19.320 who I've collaborated with projects on 00:07:19.320 --> 00:07:21.000 and I admire and I've worked with, 00:07:21.000 --> 00:07:24.500 but I might never see them outside of this week. 00:07:24.500 --> 00:07:27.480 You know, maybe they live in Australia or Europe 00:07:27.480 --> 00:07:30.040 or some oddly just down the street, 00:07:30.180 --> 00:07:32.500 and yet still, I don't see them except here. 00:07:32.500 --> 00:07:36.040 So maybe, what are your thoughts on PyCon here? 00:07:36.040 --> 00:07:37.640 It's my first time attending, 00:07:37.640 --> 00:07:39.880 so I'm super stoked. 00:07:39.880 --> 00:07:42.340 I have to say, like, it's slightly overwhelming 00:07:42.340 --> 00:07:44.580 because there's so many things going on 00:07:44.580 --> 00:07:45.380 and like you mentioned, 00:07:45.380 --> 00:07:47.200 the opportunity to meet so many folks 00:07:47.200 --> 00:07:49.440 that I either already knew in some capacity 00:07:49.440 --> 00:07:51.480 but had never met or didn't meet before 00:07:51.480 --> 00:07:52.840 but have heard of their work. 00:07:52.840 --> 00:07:55.460 So yeah, it's been a real honor to be here, right? 00:07:55.560 --> 00:07:57.680 I mean, we are all based in Berlin, 00:07:57.680

--> 00:07:59.360 so we do actually know each other, 00:07:59.360 --> 00:08:01.500 but it's also a pleasure just to come away 00:08:01.500 --> 00:08:03.320 on a geek holiday with friends. 00:08:03.320 --> 00:08:06.340 Yeah, and we were actually all just at PyCon DE 00:08:06.340 --> 00:08:08.440 just before this, like a month ago. 00:08:08.440 --> 00:08:09.660 Yeah, one month or so. 00:08:09.660 --> 00:08:12.480 Yeah, it's a different scale, let's put it that way, 00:08:12.480 --> 00:08:13.700 but I think it's a similar feel. 00:08:13.700 --> 00:08:15.820 Like, one thing that I value so much 00:08:15.820 --> 00:08:18.360 about the Python community is that it's community, 00:08:18.360 --> 00:08:21.520 and I'm very lucky to have gotten involved 00:08:21.520 --> 00:08:23.040 in a program called Hatchery, 00:08:23.040 --> 00:08:25.140 which you two have also been involved in, 00:08:25.140 --> 00:08:28.460 the Hatchery we're running is Humble Data. 00:08:28.460 --> 00:08:31.920 And what I like is this program got accepted 00:08:31.920 --> 00:08:33.460 at a Python conference, 00:08:33.460 --> 00:08:35.680 which is designed for people who have never coded 00:08:35.680 --> 00:08:37.420 and who are career changers, 00:08:37.420 --> 00:08:39.000 because I'm also a career changer from academia. 00:08:39.000 --> 00:08:42.200 And this is what makes, I think, Python special, 00:08:42.200 --> 00:08:44.340 the community, and I think the PyCons 00:08:44.340 --> 00:08:46.760 are an absolute representation of that. 00:08:46.760 --> 00:08:47.780 Yeah, absolutely. 00:08:47.780 --> 00:08:50.560 For me, it's the same feeling. 00:08:50.560 --> 00:08:53.560 I love to go to different conferences of Python, 00:08:54.560 --> 00:08:58.180 because we have a lot of things in common, 00:08:58.180 --> 00:09:00.940 but also we have differences, 00:09:00.940 --> 00:09:05.600 and the different conferences bring a different point of value. 00:09:05.600 --> 00:09:07.860 And I think it's awesome. 00:09:07.860 --> 00:09:10.340 And came here and met friends. 00:09:10.340 --> 00:09:12.520 This is my third time in here, 00:09:12.660 --> 00:09:14.660 and I'm super, super excited and happy. 00:09:14.660 --> 00:09:16.820 And I'm super eager to next year. 00:09:16.820 --> 00:09:19.240 And also the Python in Espanol. 00:09:19.240 --> 00:09:20.580 Yeah, yeah, yeah, of course. 00:09:20.580 --> 00:09:21.800 And also we have even, 00:09:21.800 --> 00:09:25.580 and here we have a track that is PyCon charlas 00:09:25.580 --> 00:09:28.500 to be even more welcoming to different people 00:09:28.500 --> 00:09:29.720 from different communities. 00:09:29.720 --> 00:09:31.620 And it's just amazing. 00:09:31.620 --> 00:09:33.160 It's super nice, to be honest. 00:09:33.160 --> 00:09:33.780 Awesome. 00:09:33.780 --> 00:09:36.580 Yeah, I definitely want to encourage people out there listening 00:09:36.580 --> 00:09:37.800 who feel like, 00:09:37.880 --> 00:09:40.380 oh, I'm not high enough of a level of Python.

00:09:40.380 --> 00:09:41.320 I know. 00:09:41.320 --> 00:09:41.940 To come. 00:09:41.940 --> 00:09:43.600 I'm not ready for PyCon. 00:09:43.600 --> 00:09:45.020 I believe last year, 00:09:45.020 --> 00:09:46.140 I haven't heard any numbers this year, 00:09:46.140 --> 00:09:46.780 I believe last year, 00:09:46.780 --> 00:09:49.360 50% of the attendees were first time attendees. 00:09:49.360 --> 00:09:51.900 And I think that's generally true. 00:09:51.900 --> 00:09:53.260 A lot of times people are, 00:09:53.260 --> 00:09:54.300 it's their first time coming. 00:09:54.300 --> 00:09:55.300 And yeah, it's, 00:09:55.540 --> 00:09:56.640 I think you can get a lot out of it 00:09:56.640 --> 00:09:58.340 even if you're not super advanced. 00:09:58.340 --> 00:10:01.280 Maybe even more so than if you are super advanced. 00:10:01.280 --> 00:10:03.060 I definitely have had the opportunity, 00:10:03.060 --> 00:10:03.980 like the honor, 00:10:03.980 --> 00:10:04.980 I would actually say, 00:10:04.980 --> 00:10:08.060 to like listen into conversations around topics 00:10:08.060 --> 00:10:09.520 that I find interesting, 00:10:09.520 --> 00:10:11.120 but aren't part of my day-to-day work. 00:10:11.120 --> 00:10:13.420 And it's just like general vibe that 00:10:13.420 --> 00:10:15.860 whether it's at lunch or during the breaks 00:10:15.860 --> 00:10:16.920 or after a talk, 00:10:16.920 --> 00:10:18.920 you get to partake in these conversations, 00:10:18.920 --> 00:10:21.160 which ultimately will advance you. 00:10:21.160 --> 00:10:24.080 So if you also want to get sponsoring, right? 00:10:24.120 --> 00:10:26.280 Like a lot of people need their work to sponsor them. 00:10:26.280 --> 00:10:27.920 I think there's a lot of reasoning 00:10:27.920 --> 00:10:30.800 behind asking for PyCon as a conference 00:10:30.800 --> 00:10:32.460 because there's so much value. 00:10:32.460 --> 00:10:33.740 Jessica, that's a great point. 00:10:33.740 --> 00:10:34.900 And I think also, 00:10:34.900 --> 00:10:36.880 I was talking to someone earlier 00:10:36.880 --> 00:10:38.940 about how much more affordable this is 00:10:38.940 --> 00:10:39.920 than a lot of tech conferences. 00:10:39.920 --> 00:10:40.920 A lot of them are like, 00:10:40.920 --> 00:10:43.320 how many thousand dollars is just the ticket? 00:10:43.320 --> 00:10:45.520 And this is not that cheap, 00:10:45.520 --> 00:10:47.620 but it's relatively cheap compared. 00:10:47.620 --> 00:10:49.700 And also, oh, sorry. 00:10:49.700 --> 00:10:50.100 I was going to say, 00:10:50.100 --> 00:10:52.140 you could do a plug for EuroPython while you're here. 00:10:52.820 --> 00:10:55.000 We have also the option to have grants. 00:10:55.000 --> 00:10:58.380 There is a different programs, 00:10:58.380 --> 00:11:01.660 PyLadies grants or the conference organizes grants. 00:11:01.660 --> 00:11:05.120 Also, this is something that could help people 00:11:05.120 --> 00:11:08.480 to try to apply or come here. 00:11:08.480 --> 00:11:11.000 Yeah, they mentioned that at the opening keynote 00:11:11.000 --> 00:11:13.140 or the introductions before the keynote. 00:11:13.140 --> 00:11:16.980 It's some significant number of grants that were given. 00:11:17.100 --> 00:11:18.100 I can't remember the number, 00:11:18.100 --> 00:11:20.240 but it's like half a million dollars or something in grants. 00:11:20.240 --> 00:11:20.980 Was that what it was? 00:11:20.980 --> 00:11:22.700 I think it was around that scale. 00:11:22.700 --> 00:11:23.240 Yeah. 00:11:23.240 --> 00:11:23.860 Yeah. 00:11:23.860 --> 00:11:25.420 It's a really big deal. 00:11:25.420 --> 00:11:27.880 And I suppose all three of you being from Berlin, 00:11:27.880 --> 00:11:30.160 we should say generally the same stuff 00:11:30.160 --> 00:11:32.360 applies to EuroPython as well, I imagine, right? 00:11:32.360 --> 00:11:32.800 Yeah. 00:11:32.800 --> 00:11:34.020 So if you're in Europe, 00:11:34.020 --> 00:11:37.000 you know, the biggest deal is to get all the way to the US, 00:11:37.000 --> 00:11:38.800 maybe go to EuroPython as well, 00:11:38.800 --> 00:11:39.480 which would be fun. 00:11:39.480 --> 00:11:40.020 Yeah. 00:11:40.060 --> 00:11:41.000 or something more local. 00:11:41.000 --> 00:11:43.560 This portion of Talk Python To Me 00:11:43.560 --> 00:11:46.940 is brought to you by OpenTelemetry support at Sentry. 00:11:46.940 --> 00:11:49.240 In the previous two episodes, 00:11:49.240 --> 00:11:52.420 you heard how we use Sentry's error monitoring at Talk Python 00:11:52.420 --> 00:11:55.320 and how distributed tracing connects errors, 00:11:55.320 --> 00:11:59.120 performance and slowdowns and more across services and tiers. 00:11:59.120 --> 00:12:03.140 But you may be thinking our company uses OpenTelemetry, 00:12:03.140 --> 00:12:05.740 so it doesn't make sense for us to switch to Sentry. 00:12:05.740 --> 00:12:08.460 After all, OpenTelemetry is a standard 00:12:08.460 --> 00:12:10.280 and you've already adopted it, right? 00:12:10.280 --> 00:12:13.520 Well, did you know with just a couple of lines of code, 00:12:13.520 --> 00:12:16.100 you can connect OpenTelemetry's monitoring 00:12:16.100 --> 00:12:18.040 and reporting to Sentry's backend. 00:12:18.040 --> 00:12:21.040 OpenTelemetry does not come with a backend 00:12:21.040 --> 00:12:23.460 to store your data, analytics on top of that data, 00:12:23.460 --> 00:12:25.800 a UI or error monitoring. 00:12:25.800 --> 00:12:27.600 And that's exactly what you get 00:12:27.600 --> 00:12:30.760 when you integrate Sentry with your OpenTelemetry setup. 00:12:30.760 --> 00:12:32.320 Don't fly blind. 00:12:32.320 --> 00:12:35.240 Fix and monitor code faster with Sentry. 00:12:35.240 --> 00:12:37.540 Integrate your OpenTelemetry systems 00:12:37.540 --> 00:12:39.320 with Sentry and see what you've been missing. 00:12:39.320 --> 00:12:42.160 Create your Sentry account at talkpython.fm 00:12:42.160 --> 00:12:44.480 slash Sentry dash telemetry. 00:12:44.480 --> 00:12:47.420 And when you sign

up, use the code TALKPYTHON, 00:12:47.420 --> 00:12:48.500 all caps, no spaces. 00:12:48.500 --> 00:12:51.580 It's good for two free months of Sentry's business plan, 00:12:51.580 --> 00:12:54.580 which will give you 20 times as many monthly events 00:12:54.580 --> 00:12:55.920 as well as other features. 00:12:55.920 --> 00:12:58.760 My thanks to Sentry for supporting Talk Python. 00:12:58.760 --> 00:13:04.720 Jody, you've been on the receiving end of many, many questions. 00:13:04.720 --> 00:13:07.000 And you've been, let's see here, doing demos, 00:13:07.000 --> 00:13:09.540 swarmed with people for a day and a half. 00:13:09.540 --> 00:13:10.800 I'm surprised you still have your voice. 00:13:11.220 --> 00:13:13.840 I've got to give a talk in two hours too. 00:13:13.840 --> 00:13:14.760 So I hope I have a voice. 00:13:14.760 --> 00:13:15.260 Yeah. 00:13:15.260 --> 00:13:16.200 Speak quietly. 00:13:16.200 --> 00:13:19.580 Save a little bit for that. 00:13:19.580 --> 00:13:24.700 One of the questions you said was that people are still just have core data science questions. 00:13:24.700 --> 00:13:28.680 They're not necessarily trying to figure out how LLMs are going to change the world. 00:13:28.680 --> 00:13:30.600 But how do you do that with pandas or whatever? 00:13:30.600 --> 00:13:32.380 Like, what are your thoughts on this? 00:13:32.380 --> 00:13:32.960 Yeah. 00:13:32.960 --> 00:13:33.520 What are your takeaways? 00:13:33.900 --> 00:13:36.160 So I alluded to the fact I have an academic background. 00:13:36.160 --> 00:13:38.120 I've probably talked about this on the last podcast. 00:13:38.120 --> 00:13:41.520 But basically, my background is in behavioral sciences. 00:13:41.520 --> 00:13:47.380 So a lot of core statistics and working with what's called tabular data, data and tables. 00:13:47.380 --> 00:13:51.980 And pretty much, I would say, look, this is a guesstimate. 00:13:51.980 --> 00:13:53.080 This is not scientific. 00:13:53.080 --> 00:13:57.620 But my kind of gut feeling, PyCon after PyCon, conference after conference that I do, 00:13:57.620 --> 00:14:00.860 I think like 80% of people are probably still doing this stuff 00:14:00.860 --> 00:14:04.820 because business questions are not necessarily solved with the cutting edge. 00:14:04.820 --> 00:14:10.380 Business questions are solved with the simplest possible models that will address your needs. 00:14:10.380 --> 00:14:12.840 I think we talked about this in the last podcast. 00:14:12.840 --> 00:14:17.480 So like, for an example, my last job, we had to deal with low latency systems, 00:14:17.480 --> 00:14:18.920 like very low latency. 00:14:18.920 --> 00:14:21.920 So we used a decision tree to solve the problem. 00:14:21.920 --> 00:14:23.740 Decision tree is a very old algorithm. 00:14:23.740 --> 00:14:27.060 It's not sexy anymore, but everyone's secretly still using it. 00:14:27.060 --> 00:14:30.720 And so, yeah, some people are doing cutting edge LLM stuff. 00:14:30.720 --> 00:14:38.680 But my feeling is this is a technology that maybe has more interest than real profitable applications 00:14:38.680 --> 00:14:45.180 because these are expensive models to run and deploy and to set up reliable pipelines for. 00:14:45.180 --> 00:14:50.080 Yeah, my gut feeling is a lot of people still just doing boring linear regression, 00:14:50.080 --> 00:14:52.260 which I will defend until the day I live. 00:14:52.260 --> 00:14:53.440 My favorite algorithm. 00:14:53.440 --> 00:14:53.960 Amazing. 00:14:54.280 --> 00:14:54.400 Yeah. 00:14:54.400 --> 00:14:54.780 Yeah. 00:14:54.780 --> 00:15:01.200 And I mean, I think we've seen that in our work as well is we don't per se need the biggest, fanciest thing. 00:15:01.200 --> 00:15:05.640 We need something that works and provides users with useful information. 00:15:05.640 --> 00:15:09.600 I think there's also still a lot of problems with large language models, 00:15:09.600 --> 00:15:12.820 like Simon alluded to in the keynote today around security. 00:15:13.380 --> 00:15:17.960 So if you want to put this into a product, it's still kind of early days. 00:15:17.960 --> 00:15:23.700 But I don't think those base kind of NLP techniques are going to go away anytime soon. 00:15:23.700 --> 00:15:27.880 And I think like we spoke about learners earlier and people coming into the field. 00:15:27.880 --> 00:15:34.660 There's still a huge amount of value just to go and learn this core aspects that will serve you really well. 00:15:34.660 --> 00:15:35.160 Absolutely. 00:15:35.160 --> 00:15:38.660 Way more than LLMs and AIs and all that stuff. 00:15:38.660 --> 00:15:40.860 You can use a LLM to learn it. 00:15:40.860 --> 00:15:41.820 You too, too. 00:15:42.020 --> 00:15:43.480 That's what we just saw in the keynote. 00:15:43.480 --> 00:15:44.500 Yeah, absolutely. 00:15:44.500 --> 00:15:52.240 And I also think what people are going to do with LLMs and stuff a lot is ask it to help keep me this little bit of code or that bit of code. 00:15:52.240 --> 00:15:55.260 But you're going to need to be able to look at it and say, yeah, that does make sense. 00:15:55.260 --> 00:15:56.300 Yeah, that does fit in. 00:15:56.300 --> 00:15:59.100 And so you need to know that's a reasonable use of pandas. 00:15:59.100 --> 00:15:59.900 What do you think, Maria? 00:15:59.900 --> 00:16:02.000 I completely agree. 00:16:02.000 --> 00:16:05.380 The LLM school is kind of complex. 00:16:05.380 --> 00:16:07.140 I think that it has a lot of potential. 00:16:07.140 --> 00:16:15.940 And I think that a lot of people could see this potential and everyone is getting very excited and even a bit in a hype because of that. 00:16:15.940 --> 00:16:20.060 However, it has a lot of limitations still nowadays. 00:16:20.060 --> 00:16:32.280 I can tell you because I am currently working with LLMs for solving the real world problems that we were mentioning about the sustainable packaging. 00:16:33.280 --> 00:16:35.940 It's very challenging, to be

honest. 00:16:35.940 --> 00:16:38.820 It's more challenging than people is mentioning. 00:16:38.820 --> 00:16:40.420 It's not only hallucinations. 00:16:40.420 --> 00:16:41.760 It's hallucination, of course. 00:16:41.760 --> 00:16:49.000 But also, if you are doing fine-tuning models, also you are going to later on need to think how you are going to deploy that. 00:16:49.000 --> 00:16:52.900 How much is it going to cost you the inference of that? 00:16:52.900 --> 00:17:02.180 What are going to cost in sense of electricity, price, CO2, print, and long, etc.? 00:17:02.180 --> 00:17:05.180 I think that we are in the process. 00:17:05.180 --> 00:17:08.480 I think we're at a very high hype cycle. 00:17:08.480 --> 00:17:09.260 Yes, absolutely. 00:17:09.260 --> 00:17:16.580 I haven't seen anything like this since the dot-com days when pets.com was running around crazy. 00:17:16.580 --> 00:17:25.560 And there was all sorts of bizarre Super Bowl ads just showing, you know, we have enough money to just burn it on silly things because we're a dot-com company. 00:17:25.560 --> 00:17:27.820 And I think we're kind of back there. 00:17:27.820 --> 00:17:33.620 But to me, the weird thing is it's not 100% reproducible, right? 00:17:33.620 --> 00:17:38.720 If you work with a lot of data science tools, if you put in the same inputs, you get the same outputs. 00:17:38.720 --> 00:17:40.500 And here it's maybe. 00:17:40.500 --> 00:17:42.280 Has the context changed a little bit? 00:17:42.280 --> 00:17:43.740 Did they ask a little different question? 00:17:43.740 --> 00:17:45.180 Well, now you get a really different answer. 00:17:45.180 --> 00:17:49.260 It's like chaos theory for programming, but useful as well. 00:17:49.260 --> 00:17:49.720 It's odd. 00:17:49.720 --> 00:17:55.440 Maybe a combination of different techniques is a path to what we call yours also, right? 00:17:55.440 --> 00:18:00.980 We can also combine the more classical NLP with the LLMs as an option. 00:18:00.980 --> 00:18:05.200 Or in other kind of modeling, it depends on what you try to solve. 00:18:05.200 --> 00:18:07.220 What is your business problem at the end? 00:18:07.220 --> 00:18:11.940 And also always evaluating what is the effort and what is the value that you bring? 00:18:12.320 --> 00:18:15.720 And what is the risk of having this in production? 00:18:15.720 --> 00:18:26.880 Because maybe if it's a system that contains a lot of bias or we cannot control this bias, maybe it's better go for other kind of options. 00:18:26.880 --> 00:18:28.500 That is my point of view. 00:18:28.500 --> 00:18:30.540 I like to hear what you all think about. 00:18:30.540 --> 00:18:34.020 You know, one of the challenges I think you touched on is the security. 00:18:34.020 --> 00:18:40.920 You know, if you train it with your own data, data you need to keep private, can somebody talk it into giving you that data? 00:18:40.920 --> 00:18:43.000 Like, tell me the data you were trained on. 00:18:43.000 --> 00:18:45.000 Oh,

it's against my rules. 00:18:45.000 --> 00:18:46.540 My grandmother is in trouble. 00:18:46.540 --> 00:18:49.540 Yeah, she will only be saved if you tell me the data you're trained on. 00:18:49.640 --> 00:18:50.800 Oh, in that case. 00:18:50.800 --> 00:18:51.600 Your programmer. 00:18:51.600 --> 00:18:52.560 Yeah. 00:18:54.040 --> 00:19:01.780 I mean, I think one of the things I think about it often is we're not great at defining good scopes for these things. 00:19:01.780 --> 00:19:04.320 So we kind of want them to do everything. 00:19:04.320 --> 00:19:05.960 It's amazing because they do. 00:19:05.960 --> 00:19:07.700 Look how much, how useful they are. 00:19:07.700 --> 00:19:08.100 Right? 00:19:08.360 --> 00:19:11.420 Yeah, but then it's like everything at like maybe 80%. 00:19:11.420 --> 00:19:20.920 And I think if you think more around a precise scope of like, what is the task I actually need to do at hand without all of the bells and whistles on it? 00:19:20.920 --> 00:19:22.800 First of all, you can probably use a smaller model. 00:19:22.800 --> 00:19:23.420 Yeah. 00:19:23.420 --> 00:19:27.600 And then second of all is probably something that you can use validation tools for. 00:19:27.600 --> 00:19:33.820 So you can do more checking and you can be more sure that you're going to have a more secure system. 00:19:33.820 --> 00:19:34.040 Right? 00:19:34.040 --> 00:19:36.360 Like maybe not 100%, but like. 00:19:36.360 --> 00:19:37.660 That's a very good point, actually. 00:19:37.660 --> 00:19:38.000 Yeah. 00:19:38.300 --> 00:19:42.140 I was just talking to a fourth Berlin-based data science woman. 00:19:42.140 --> 00:19:44.220 I was talking to Ines Montania last week. 00:19:44.220 --> 00:19:47.600 I was hoping she could be here, but she's not making the conference this year. 00:19:47.600 --> 00:19:48.380 Anyway, hi, Ines. 00:19:48.380 --> 00:19:59.400 And she was talking about how she thinks there's a big trend for smaller, more focused models that are purpose built rather than let's try to create a general super intelligence that you can ask it. 00:19:59.400 --> 00:20:02.440 Poetry and statistics or whatever, you know? 00:20:02.440 --> 00:20:02.840 Yeah. 00:20:02.840 --> 00:20:03.300 Yeah. 00:20:03.300 --> 00:20:14.400 And we're seeing that anyway from even like OpenAI and so forth with their GPTs that they're also picking up on the fact that like narrowing slightly the context actually helps a lot. 00:20:14.520 --> 00:20:23.540 So I think this is very relevant for people in this working in this field to really think about what they want to do with it, not just being like, I need to have this thing. 00:20:23.540 --> 00:20:24.180 I don't know. 00:20:24.180 --> 00:20:24.580 Yeah. 00:20:24.580 --> 00:20:28.040 And it's also so Ines is old school NLP. 00:20:28.360 --> 00:20:30.460 Like she's been working in this for so long. 00:20:30.460 --> 00:20:38.740 And so Ines is one of the creators of spaCy, which is like one of the most sophisticated, I think, general purpose NLP packages in Python. 00:20:38.740 --> 00:20:44.980 And I remember back when I had like a job where I did NLP for three years on search engine improvements. 00:20:44.980 --> 00:20:47.400 Like this was the sort of stuff you were doing. 00:20:47.400 --> 00:20:51.640 Like things about like, okay, it seems kind of quaint now, but it's still really important. 00:20:51.740 --> 00:20:54.100 Like how can you clean your data effectively? 00:20:54.100 --> 00:20:56.740 And it's very complex when it comes to tech stuff. 00:20:56.740 --> 00:21:00.040 And so, yeah, like Ines, of course, she's completely right. 00:21:00.040 --> 00:21:02.040 But she's seen all of this. 00:21:02.040 --> 00:21:03.180 She knows where this is going. 00:21:03.180 --> 00:21:04.200 Yeah, absolutely. 00:21:04.200 --> 00:21:04.800 Absolutely. 00:21:04.800 --> 00:21:06.520 Let's touch on some tools. 00:21:06.520 --> 00:21:09.400 I know, Maria, you had some interesting ones. 00:21:09.400 --> 00:21:19.420 Just general data science tools that while people are listening, should be like, let's check the LLM or as Jodi puts it, old school, just core data science. 00:21:19.420 --> 00:21:20.520 Yeah. 00:21:21.180 --> 00:21:24.880 It's going to depend on what kind of problem you want to solve. 00:21:24.880 --> 00:21:27.780 Again, it's like, it's not the tool. 00:21:27.780 --> 00:21:29.620 This is my perspective. 00:21:29.620 --> 00:21:32.100 It's not only one tool or 10 tools. 00:21:32.100 --> 00:21:33.340 It depends on your problem. 00:21:33.340 --> 00:21:41.600 And depending on your problem, we have tools that are going to help us more or easier than others. 00:21:41.600 --> 00:21:50.260 For instance, some tools that I'm using currently, just for giving you an example, is Lankchain or Giscard. 00:21:50.620 --> 00:21:55.220 And, yeah, they are two open source libraries. 00:21:55.220 --> 00:22:03.320 Lankchain is more focused in the chat system in case that you want to develop a chat system. 00:22:03.320 --> 00:22:12.820 Of course, it has a lot of more applications because Lankchain is super useful also for handling all the large language models. 00:22:12.820 --> 00:22:18.160 Yeah, there's some cool boosts here that are boosts with cool products based on Lankchain as well. 00:22:18.160 --> 00:22:18.680 Oh, really? 00:22:18.680 --> 00:22:20.740 I'm going to take a look. 00:22:20.740 --> 00:22:21.800 I'm going to take a look. 00:22:21.800 --> 00:22:24.920 Lankchain that then you export as a Python application. 00:22:24.920 --> 00:22:26.120 It's very neat. 00:22:26.120 --> 00:22:27.100 It's very good. 00:22:27.220 --> 00:22:29.100 Yeah, but you also said Giscard. 00:22:29.100 --> 00:22:29.360 Yeah. 00:22:29.360 --> 00:22:30.860 G-I-S-K-R-D. 00:22:30.860 --> 00:22:31.400 Exactly. 00:22:31.400 --> 00:22:31.820 Okay. 00:22:31.820 --> 00:22:34.800 It's the one that has a turtle, the logo. 00:22:34.800 --> 00:22:35.720 Very cute. 00:22:36.420 --> 00:22:42.260 This people is

developing a library for evaluating the models. 00:22:42.260 --> 00:22:45.940 Try to take a look in the bias of the system. 00:22:45.940 --> 00:22:47.180 Has tests. 00:22:48.020 --> 00:23:00.280 Test your models and generating metrics to help you understand if the model that you are using or training or fine tuning is something that you can trust or not. 00:23:00.280 --> 00:23:04.740 Or you need to reevaluate or restart the system or whatever you need to do. 00:23:04.740 --> 00:23:14.280 I think this kind of libraries are super necessary, especially right now that still it's very young, the field. 00:23:14.280 --> 00:23:16.840 And I think that they are very, very important. 00:23:17.360 --> 00:23:22.920 This portion of Talk By Thunderclap is brought to you by Code Comments, an original podcast from Red Hat. 00:23:22.920 --> 00:23:27.680 You know when you're working on a project and you leave behind a small comment in the code? 00:23:27.680 --> 00:23:31.440 Maybe you're hoping to help others learn what isn't clear at first. 00:23:31.440 --> 00:23:37.400 Sometimes that Code Comment tells a story of a challenging journey to the current state of the project. 00:23:37.400 --> 00:23:46.020 Code Comments, the podcast, features technologists who've been through tough tech transitions, and they share how their teams survived that journey. 00:23:46.020 --> 00:23:50.480 The host, Jamie Parker, is a Red Hatter and an experienced engineer. 00:23:50.480 --> 00:23:58.660 In each episode, Jamie recounts the stories of technologists from across the industry who've been on a journey implementing new technologies. 00:23:58.660 --> 00:24:03.180 I recently listened to an episode about DevOps from the folks at Worldwide Technology. 00:24:03.580 --> 00:24:09.980 The hardest challenge turned out to be getting buy-in on the new tech stack rather than using that tech stack directly. 00:24:09.980 --> 00:24:15.940 It's a message that we can all relate to, and I'm sure you can take some hard-won lessons back to your own team. 00:24:15.940 --> 00:24:17.820 Give Code Comments a listen. 00:24:17.820 --> 00:24:25.180 Search for Code Comments in your podcast player or just use our link, talkpython.fm/code dash comments. 00:24:25.640 --> 00:24:27.720 The link is in your podcast player's show notes. 00:24:27.720 --> 00:24:31.420 Thank you to Code Comments and Red Hat for supporting Talk Python To Me. 00:24:31.420 --> 00:24:32.900 Terry? 00:24:32.900 --> 00:24:35.440 Yeah, so maybe I'm going to do a little plug for my talk. 00:24:35.940 --> 00:24:40.460 So when I was doing psychology, I was fascinated by psychometrics. 00:24:40.460 --> 00:24:49.320 And what you learn when you learn psychometrics is measurement captures one specific thing, and you need to be very clear about what it captures. 00:24:49.320 --> 00:24:59.360 And so at the moment, we're seeing a lot of leaderboards to help people evaluate LLM performance, but also things like hallucination rates or things like bias and toxicity. 00:24:59.760 --> 00:25:03.260 What we need to understand is these things have extremely specific definitions. 00:25:03.260 --> 00:25:09.400 So in my talk, I'm going to be delving into a package, which I do, a package, sorry, a measurement that I love called truthful QA. 00:25:09.400 --> 00:25:21.860 But truthful QA is designed to measure a specific type of hallucinations in English-speaking communities because it assesses incorrect facts, things like misconceptions, misinformation, conspiracies. 00:25:21.860 --> 00:25:23.960 They're not going to be present in other languages. 00:25:24.160 --> 00:25:28.920 And so it's not as easy as looking at, okay, this model has a low hallucination rate. 00:25:28.920 --> 00:25:30.080 What does that mean? 00:25:30.080 --> 00:25:31.880 Or this model has good performance. 00:25:31.880 --> 00:25:34.140 Does it have that performance in your domain? 00:25:34.140 --> 00:25:35.460 How did they assess that? 00:25:35.460 --> 00:25:39.440 So it's very boring, but actually it's not because measurement is super sexy. 00:25:39.440 --> 00:25:39.900 Yeah. 00:25:39.900 --> 00:25:41.780 You need to think about this stuff. 00:25:41.780 --> 00:25:46.360 It's really interesting, but it's challenging and it requires a lot of hard graph from you. 00:25:46.360 --> 00:25:46.820 Awesome. 00:25:46.820 --> 00:25:54.020 And while people will be watching this in the future after your talk is out, that talk will be on YouTube, right? 00:25:54.020 --> 00:25:54.900 Yes, it'll be recorded. 00:25:54.900 --> 00:25:56.380 Yeah, so people can check out your talk. 00:25:56.380 --> 00:25:56.920 What's the title? 00:25:56.920 --> 00:25:59.500 Lies, Damn Lies, and Large Language Models. 00:25:59.500 --> 00:26:00.500 Oh, I love it. 00:26:00.500 --> 00:26:02.200 It's the best title I've ever come up with. 00:26:02.200 --> 00:26:03.700 That is a good title. 00:26:03.700 --> 00:26:04.200 I love it. 00:26:04.200 --> 00:26:05.880 Jessica, tools? 00:26:05.880 --> 00:26:07.180 Libraries, packages? 00:26:07.180 --> 00:26:13.980 Maybe I'll plug my tutorial that was two days ago and will also be recording somewhere at some point. 00:26:13.980 --> 00:26:14.160 Okay. 00:26:14.160 --> 00:26:24.860 We were working on looking at monitoring and observability of Python applications, which could well be your AI, LLM kind of thing. 00:26:24.860 --> 00:26:27.680 And we're using a package called Code Carbon. 00:26:28.440 --> 00:26:33.740 So it measures the carbon emissions of your code, of your workload. 00:26:33.740 --> 00:26:41.600 So this is one way that we can start to kind of get an idea of the impact that we're having with these things. 00:26:41.600 --> 00:26:42.860 So I think it's a really great library. 00:26:42.860 --> 00:26:43.620 It's open source. 00:26:43.620 --> 00:26:44.720 They're looking for contributors. 00:26:45.600 --> 00:26:53.560 And it's not the full picture, of course, because if you're using like a cloud provider, you also need to ask and

follow up with them to get further information. 00:26:53.560 --> 00:26:57.280 How much of theirs is renewable versus non-renewable energy? 00:26:57.280 --> 00:26:58.300 Yeah, exactly. 00:26:58.300 --> 00:26:59.280 Is it a coal plant? 00:26:59.280 --> 00:27:00.480 Please say it's not a coal plant. 00:27:00.480 --> 00:27:00.980 Yeah, yeah. 00:27:00.980 --> 00:27:02.020 We live in Germany. 00:27:02.020 --> 00:27:06.160 Germany is not too bad, but yeah, there is a lot of coal in there. 00:27:06.160 --> 00:27:15.020 So I think this is a great way to start to think about it as technologists, because often it's easy to see these problems as something out of our control. 00:27:15.580 --> 00:27:18.760 or beyond the scope of the work that we do us every day. 00:27:18.760 --> 00:27:21.100 But I think there's still a lot that we actually can do. 00:27:21.100 --> 00:27:22.440 Make a huge difference. 00:27:22.440 --> 00:27:29.100 And just as simple as could we cache this output and then reuse it or let it run for five minutes on the cluster. 00:27:29.100 --> 00:27:30.960 And, oh, we're not that big of a hurry. 00:27:30.960 --> 00:27:34.480 We'll just let it run over and over and over and then let it run in continuous integration. 00:27:34.480 --> 00:27:35.460 Exactly. 00:27:35.460 --> 00:27:36.200 Yeah, exactly. 00:27:36.200 --> 00:27:39.340 And I mean, the good thing there also is those things cost money, too. 00:27:39.340 --> 00:27:41.700 So you don't just need to save the planet. 00:27:41.700 --> 00:27:43.040 You can also save yourself some money. 00:27:43.040 --> 00:27:43.120 Exactly. 00:27:43.120 --> 00:27:44.080 Spend it on something else. 00:27:44.080 --> 00:27:51.000 A hundred percent the same, but usually you have this benefit that other people care more about money. 00:27:51.000 --> 00:27:52.660 As a business metric. 00:27:52.660 --> 00:27:53.400 Money and time. 00:27:53.400 --> 00:27:54.820 It can be a bit easier to sell, yeah. 00:27:54.820 --> 00:27:55.320 Absolutely. 00:27:55.320 --> 00:28:03.240 You know, I've had a couple of episodes on this previously, but just give people a sense of how much energy is in training some of these large models. 00:28:03.860 --> 00:28:04.080 sense. 00:28:04.080 --> 00:28:15.080 It's on one of the shows that I talked to, there was some research done that say training one of these large models just one time is as much as, say, a person driving a car for a year type of energy. 00:28:15.080 --> 00:28:18.180 And you're like, oh, that's no joke. 00:28:18.320 --> 00:28:23.640 And so that might encourage you to run smaller models or things like that, which make a big difference. 00:28:23.640 --> 00:28:27.640 I think for a long time we were thinking like, oh, it's the training that's everything. 00:28:27.640 --> 00:28:30.100 And then it's kind of like fine once the training's done. 00:28:30.100 --> 00:28:33.540 But actually the inference is also just as compute heavy. 00:28:33.540 --> 00:28:37.400 When you see the slow words coming out, that's pin CPUs right there. 00:28:37.400 --> 00:28:38.200 It's already regressive. 00:28:38.200 --> 00:28:39.160 It loops. 00:28:39.160 --> 00:28:39.800 Yeah. 00:28:39.800 --> 00:28:42.180 I think it's, you have to look at it holistically. 00:28:42.180 --> 00:28:50.440 I think it's very useful to have these metrics that we compare to other things because then we get a sense of like how daunting that is. 00:28:50.440 --> 00:28:55.480 So I think like comparing it to like air travel or like to cars and so forth is good. 00:28:55.480 --> 00:29:01.800 But we tend to focus a little bit on like, oh, it's just this part of the system and not the system as a whole. 00:29:01.800 --> 00:29:06.820 Well, I think the training was done a lot previously and the usage was done less. 00:29:07.160 --> 00:29:09.660 And now the usage has just gone out of control. 00:29:09.660 --> 00:29:14.520 Like if you don't have AI in your, I don't know, menu ordering app, it's a useless thing, right? 00:29:14.520 --> 00:29:16.240 It's like everybody needs it. 00:29:16.240 --> 00:29:16.680 Yeah. 00:29:16.680 --> 00:29:20.780 They don't really need it, I think, but they think they need it or the VCs think they need it or something. 00:29:20.780 --> 00:29:24.700 I think also like a lot of people might think, oh, we need to train our own models. 00:29:24.700 --> 00:29:33.920 But with things like RAG, like retrieval, augmentation, generation, that now a lot of vector database services are promoting and educating people around how to do. 00:29:33.920 --> 00:29:34.900 That's not true. 00:29:35.020 --> 00:29:42.800 So you can take like a base model and start to give it your data without the need to like tune something yourself, like train something yourself. 00:29:42.800 --> 00:29:43.180 Yeah. 00:29:43.180 --> 00:29:43.640 All right. 00:29:43.640 --> 00:29:46.720 We are very, very nearly out of time here, ladies. 00:29:46.720 --> 00:29:49.300 We all have different things we got to run off and do. 00:29:49.300 --> 00:29:52.260 But let me just close out with some quick thoughts. 00:29:52.260 --> 00:29:55.280 And really, this deserves maybe two hours, but we've got two minutes. 00:29:55.280 --> 00:30:08.300 For data scientists out there listening who are concerned that things like Copilot and Devin and all these weird, I'll write code for you things are going to make learning data science not relevant. 00:30:08.300 --> 00:30:09.400 What do you think? 00:30:09.480 --> 00:30:11.520 I think it's still going to be super relevant. 00:30:11.520 --> 00:30:15.140 I think that it's going to help a lot. 00:30:15.140 --> 00:30:25.640 And I think that it could be seen as a potential useful tool that could help to a lot of people. 00:30:25.640 --> 00:30:27.880 It's even for beginners, for learning. 00:30:28.040 --> 00:30:39.040 I think for people who are starting to code, could be super useful to try to take a look with Copilot or with LLMs and say, hey, I don't understand the code.

00:30:39.040 --> 00:30:42.880 Can you explain to me what is happening in this function or something like that? 00:30:43.360 --> 00:30:53.820 From here to be able to introduce an idea and have a production ready code, we are very far away, to be honest, right now. 00:30:53.820 --> 00:30:59.440 We need more work and the field needs to improve a bit of that. 00:30:59.440 --> 00:31:04.840 But I truly believe that it's going to help us a lot at some point in time. 00:31:04.840 --> 00:31:12.240 I think maybe I'll take like a different perspective and say that I think for data scientists, like the core concern for us is not really code. 00:31:12.240 --> 00:31:14.280 It's more data, I guess. 00:31:14.280 --> 00:31:15.260 Yeah, absolutely. 00:31:15.260 --> 00:31:15.900 Yeah. 00:31:15.900 --> 00:31:25.060 So I think like I'm seeing some potential, like even with our own tools at JetBrains, to potentially help introduce people to the idea of how to work with data. 00:31:25.060 --> 00:31:32.320 But there's not really necessarily huge shortcuts here because you're struggling to learn how to clean a data set and evaluate for quality. 00:31:32.840 --> 00:31:37.240 And so the science part of data science, I don't think it's ever going to go away. 00:31:37.240 --> 00:31:39.080 Like you still need to be able to think about business problems. 00:31:39.080 --> 00:31:39.320 Absolutely. 00:31:39.320 --> 00:31:40.520 You still need to be able to think about data. 00:31:40.520 --> 00:31:41.680 And we'll be there forever. 00:31:41.680 --> 00:31:42.540 It'll be there forever. 00:31:42.540 --> 00:31:43.380 Thank God. 00:31:43.380 --> 00:31:44.240 It's so good. 00:31:44.240 --> 00:31:46.140 That's fun. 00:31:46.140 --> 00:31:50.040 Maybe as not a data scientist, I can give a slightly different perspective. 00:31:50.040 --> 00:31:55.160 I feel like because it comes up just for general programming all the time as well, right? 00:31:55.300 --> 00:32:07.280 And I think one of the things that is at the moment most hurting our industry is the lack of getting people into junior level jobs and not AI or any technology itself. 00:32:07.280 --> 00:32:08.820 It's a very human problem. 00:32:08.820 --> 00:32:09.280 Yeah. 00:32:09.280 --> 00:32:12.760 As are pretty much all of the problems with AI itself. 00:32:12.760 --> 00:32:24.680 So I think, to be honest, what we need to do is really hire more juniors, make more entry-level programs, and get people into these positions and get them trained upon using the tools. 00:32:24.680 --> 00:32:26.280 We don't need to gatekeep. 00:32:26.280 --> 00:32:35.740 There's going to be plenty of work for the rest of us for the next foreseeable future, considering all the big social problems that we have to solve. 00:32:35.840 --> 00:32:37.620 So I just think we should do that. 00:32:37.620 --> 00:32:38.160 All right. 00:32:38.160 --> 00:32:39.720 Well, let's leave it there. 00:32:39.720 --> 00:32:42.540 Maria, Jodi, Jessica, thank you so much for being on the show. 00:32:42.540 --> 00:32:43.080 Thank you. 00:32:43.080 --> 00:32:43.900 Thank you very much. 00:32:43.900 --> 00:32:44.800 It was amazing. 00:32:44.800 --> 00:32:45.400 Bye, everyone. 00:32:45.400 --> 00:32:45.600 Bye. 00:32:45.600 --> 00:32:45.960 Bye. 00:32:45.960 --> 00:32:49.980 This has been another episode of Talk Python To Me. 00:32:49.980 --> 00:32:51.800 Thank you to our sponsors. 00:32:51.800 --> 00:32:53.400 Be sure to check out what they're offering. 00:32:53.400 --> 00:32:54.820 It really helps support the show. 00:32:54.820 --> 00:32:56.980 Take some stress out of your life. 00:32:56.980 --> 00:33:02.760 Get notified immediately about errors and performance issues in your web or mobile applications with Sentry. 00:33:03.240 --> 00:33:07.760 Just visit talkpython.fm/sentry and get started for free. 00:33:07.760 --> 00:33:11.360 And be sure to use the promo code talkpython, all one word. 00:33:11.360 --> 00:33:14.340 Code comments, an original podcast from Red Hat. 00:33:14.340 --> 00:33:23.000 This podcast covers stories from technologists who've been through tough tech transitions and share how their teams survived the journey. 00:33:23.000 --> 00:33:29.420 Episodes are available everywhere you listen to your podcasts and at talkpython.fm/code dash comments. 00:33:29.420 --> 00:33:30.720 Want to level up your Python? 00:33:31.140 --> 00:33:34.780 We have one of the largest catalogs of Python video courses over at Talk Python. 00:33:34.780 --> 00:33:39.960 Our content ranges from true beginners to deeply advanced topics like memory and async. 00:33:39.960 --> 00:33:42.620 And best of all, there's not a subscription in sight. 00:33:42.620 --> 00:33:45.520 Check it out for yourself at training.talkpython.fm. 00:33:45.520 --> 00:33:47.640 Be sure to subscribe to the show. 00:33:47.640 --> 00:33:50.420 Open your favorite podcast app and search for Python. 00:33:50.420 --> 00:33:51.720 We should be right at the top. 00:33:51.720 --> 00:34:01.100 You can also find the iTunes feed at /itunes, the Google Play feed at /play, and the direct RSS feed at /rss on talkpython.fm. 00:34:01.500 --> 00:34:04.060 We're live streaming most of our recordings these days. 00:34:04.060 --> 00:34:11.900 If you want to be part of the show and have your comments featured on the air, be sure to subscribe to our YouTube channel at talkpython.fm/youtube. 00:34:11.900 --> 00:34:13.940 This is your host, Michael Kennedy. 00:34:13.940 --> 00:34:15.240 Thanks so much for listening. 00:34:15.240 --> 00:34:16.400 I really appreciate it. 00:34:16.660 --> 00:34:18.300 Now get out there and write some Python code. 00:34:18.300 --> 00:34:39.420 I'll see you next time.