YouTube transcript (hCkf_UsASlk) — Transcript

Full transcript

1. 0:01Good afternoon and welcome to our second  interview with a professional in the sector,
2. 0:04for the Ethics and Technology students at ECE.
3. 0:07I'm Keith Sarver, a teacher at ECE.
4. 0:10Hello Daniel, and welcome to you as well.
5. 0:12Hello, thank you! Very glad to be here.
6. 0:14Shall we get started?
7. 0:16Yep.
8. 0:17Could you please introduce yourself and your company, and tell us what you do?
9. 0:21Sure. My name is Daniel Atherton-Moore and I'm one of the Solutions Engineers
10. 0:24at Cogniac,
11. 0:25so I provide a lot of training, a lot of site surveys, and support for customers,
12. 0:30and Cogniac is a vision inspection company utilizing advanced Artificial Intelligence,
13. 0:37It reads kind of like cognac, but it's actually a combination of cognition –
14. 0:42thinking – and ENIAC, the first sort of industrial programmable computer as we think of them today.
15. 0:49Very interesting.
16. 0:50Well, could you tell us has there been any recent ...egislation… has there been any recent legislation that has affected your  work?
17. 0:58So, I feel like GDPR has affected everyone in particular, so a lot of the sort of cutting-edge privacy legislations are coming out of
18. 1:06Europe, not the United States,
19. 1:08but since we have a presence in Europe, that does affect the way we operate.
20. 1:14So, depending on the interpretation of identifying features it can get,
21. 1:19honestly, very difficult to do even basic things, like,
22. 1:22let’s say I want to look at a picture of somebody
23. 1:25I don’t know, they're- they’re assembling some kind of McGuffin.
24. 1:28Well, what if in that picture picture their hand has a tattoo or a ring? Could we identify them
25. 1:34from that? Even though we have no desire to we're not trying to, identify
26. 1:37any people we're just trying to look at the thing they're assembling.
27. 1:40and that was a– a concern that was brought up.
28. 1:42So that caused you lots of extra work?  In some situations…
29. 1:46And it meant that we had to uh at least pull back from  certain spaces while we figured out how to handle those concerns.
30. 1:52And then what happens if someone complains or…
31. 1:55Does that bring up legal issues? Or, do you just have to redo the work, or…?
32. 2:00I mean there could certainly be legal processes but it really depends like we're not storing,
33. 2:04and it’s not as if we’re tracking in information on people who are doing these tasks, so even you said… “Ah well, this-
34. 2:10this hand has an identifying mark!”
35. 2:11Well, that, okay, but we have no idea who that person is!
36. 2:14or if they will, let’s blur it out.
37. 2:16They have lots of people that wear rings.
38. 2:17Like there's workarounds for these concerns and I think that, in general,
39. 2:21trying to preserve people's privacy is important.
40. 2:24Of course, especially more and more.
41. 2:28So, how does Cogniac address CSR in its activities?
42. 2:33So, in terms of Corporate Responsibility, I think it's just what comes down to
43. 2:39automation, in general, across the entire history of automation which is, oh well, are you just
44. 2:44replacing people’s jobs, or you pushing out– people out of work?
45. 2:49And I think what we do is a little bit different because we're working in
46. 2:52inspection spaces that,
47. 2:54right now in manufacturing, and in a lot of um,
48. 2:58like rail and logistics,
49. 3:01they do not have enough people.
50. 3:03They do not have enough employees.
51. 3:05So, in fact, what typically happens is we will take someone from a task that is
52. 3:09this is the a traditional term robotics: “Dull, dirty, or  Dangerous,”
53. 3:13and be able to reposition them to a task where they're
54. 3:15getting more skills, they're doing more value added activities. Things like that,
55. 3:20which honestly was part of the vision for the very first robots.
56. 3:24Right, we wanted to replace people in the most
57. 3:27lethal jobs and I'm I do mean that term lethal very literally
58. 3:32thousands of people used to die every year in manufacturing  spaces.
59. 3:37And so AI, while maybe less dramatic,
60. 3:40were taking people out of the really mind-numbing tasks of staring at
61. 3:44parts as they move past a conveyor
62. 3:46for eight hours a day because humans just are not built for that sort of task.
63. 3:50For sure.
64. 3:51This expression “dull, dirty, and dangerous” that’s the first time I've heard that. it's interesting. Is that a sort of jargon in in the sector or…?
65. 3:57Yeah, it’s– it”s very well often used in in robotics, because…
66. 4:02as the first,
67. 4:04standard, industrial robot – the unimate –
68. 4:07was created that was really its task.
69. 4:09Not simply to, well, make more things, but like make
70. 4:12production safer and better for the people doing production.
71. 4:18Okay.
72. 4:21How has ESG evolved recently, and have you adapted your approach?
73. 4:26So the discussion of social responsibility is…
74. 4:31I think one that everyone should spend a little bit of time on
75. 4:33There’s been a lot of concerns about, say, the power
76. 4:37usage of AI, but again, there’s lots of different kinds of AI out there in the world.
77. 4:42So, if you’re doing…
78. 4:44mass computations to try
79. 4:47build new large language learning models you’re scouring the entire Internet,
80. 4:51you know that does take quite a bit of power and water.
81. 4:55Kind of,
82. 4:56if you have any friends who are Bitcoin miners, like, that takes a tremendous amount of power, and
83. 5:00the output – is depending who you talk to – a little nebulous
84. 5:05Often times,
85. 5:06what my company does is we’re doing much smaller deployments
86. 5:09so we’re installing just like a computer
87. 5:11to do a task among– amongst a bunch of other computers, right, so…
88. 5:16by really embodying Edge Computing –
89. 5:20Edge being a phys– a fancy way of saying not in the cloud –
90. 5:23we keep our, sort of, footprint low.
91. 5:28Good for you, it's not so easy to do.
92. 5:34Could you tell us if we can learn anything from the Luddites,
93. 5:38in relation to AI?
94. 5:40Sure! So, the Luddites
95. 5:43have this reputation of being people that just hated technology. This is a group
96. 5:48about 200 years ago in England.
97. 5:51They were famous for destroying machines and even burning down factories.
98. 5:57And so there’s this idea – at least in, in, in a lot of U.S. English  that if you hate technology you're a Luddite.
99. 6:02If you don’t know how to use your phone properly: you’re  a Luddite! If you,
100. 6:05you know, don’t keep up with trends, you’re a Luddite.
101. 6:09But traditionally, the Luddites were actually technologically savvy. These were machine operators.
102. 6:14These were skilled experts in using industrial looms, in using
103. 6:20these foundational technology products that formed the Industrial Revolution.
104. 6:25The true complaint of the Luddite was : hold on a minute!
105. 6:30You’re installing these looms
106. 6:32and then you’re gonna put
107. 6:34an apprentice on that loom and then just never…
108. 6:37replace that Apprentice; you’re never going to give that Apprentice a promotion; they’re never going to become a Journeyman?
109. 6:43People were using these looms not to create good products,
110. 6:45not to create good work,
111. 6:47not to create good jobs,
112. 6:49but to drive people into poverty,
113. 6:52in order to shore up their own bottom line. And I think…
114. 6:56that’s just a technology thing! That’s going to be
115. 7:00something that’s true of any technology and the way we handle it doesn’t have to do with technological solutions, you can’t
116. 7:06automate and innovate your way out of
117. 7:08this problem, because it’s a human problem.
118. 7:11The way humans handle technology,
119. 7:13the way humans are allowed to handle technology is incredibly important.
120. 7:17So, yes,
121. 7:17It’s still a problem today, for sure, yeah?
122. 7:20Luddites destroyed machines because those machines were being used to drive people into poverty,
123. 7:25and they weren't saying:
124. 7:27“No machines!”
125. 7:28They were saying you have to provide good jobs.
126. 7:32There’s this idea that radical groups across the history have said: “We don’t wanna work!”
127. 7:37when in fact, typically it's the opposite.
128. 7:40whether you look at Luddites, whether you look at  other social and label organizations throughout time,
129. 7:44usually, what they demand is uh mandatory
130. 7:47jobs that everyone who wants a job will have a job.
131. 7:51Right, they're not saying:
132. 7:52“We don’t want to work, we don’t want to do these things!”
133. 7:54They’re saying: “We do want to work. We want to be paid appropriately for our work.
134. 7:58We want to produce good product.
135. 8:00We want our work to make the world and our families a better place.
136. 8:05So, when we think about people and  the way they approach technology,
137. 8:10we do need to ask these questions about social responsibility.
138. 8:12Why is this being implemented?
139. 8:14Is it truly being implemented to
140. 8:17lay off a bunch of people? Is it to…
141. 8:20reposition someone in a factory so they can move from a– a
142. 8:23cruddy job into a better job.
143. 8:25Right? That’s a great example of
144. 8:27using technology to improve the life of people on the assembly line.
145. 8:31and also improving the output of that assembly line.
146. 8:35Like when you talk about corporate responsibility as well,
147. 8:37and going back to the prior question
148. 8:39if because we work in a quality space, very many times,
149. 8:42if we can prevent a bad product
150. 8:46from making it into a car, well then we can prevent that car from being torn down and and scrapped, right?
151. 8:52What’s the total carbon debt of an entire car? Every car that we prevent from being scrapped,
152. 8:57we save a huge amount of carbon on.
153. 9:01As you can see, I’m– I’m a little bit passionate on these topics because you can use technology to make
154. 9:06the world a more environmentally responsible place, if that’s a thing
155. 9:11you care about, if that’s a thing you can make
156. 9:13other people care about. Again, it’s a human issue.
157. 9:16“Technology will not save,” us as the saying goes, and I do believe that despite working in a very high-tech industry.
158. 9:23so, do you think there's been a sort of resurgence of  Luddites?
159. 9:27Or they've just never gone away?
160. 9:29I would say…
161. 9:31it's never gone away in the sense that
162. 9:32people are always on the lookout
163. 9:35for abusive practices in industry and in labor.
164. 9:39Right? Now…
165. 9:43it's, so– so basically when you hear someone being called a Luddite, when you see–
166. 9:46… see, hear someone  called
167. 9:48anti-technology, what I would say is …
168. 9:51Why are they trying to gloss over this issue?
169. 9:55Because is that person really anti- technology, or are they anti-how-that-technology-is-being-used?
170. 10:02Are they anti-technology or are they
171. 10:06pro artists’ rights in the case of, say generative art AI, like Stable  Diffusion.
172. 10:14Luddite is a convenient term, but it can also be a thought terminating phrase.
173. 10:21“Oh, they're a Luddite! I can stop critically examining this issue.”
174. 10:24That’s not really the … you think it is.
175. 10:28Yeah, it’s a good distinction to make, actually.
176. 10:30Okay. What– what key challenges do you see emerging in  the coming years…
177. 10:36With regards to AI in particular?
178. 10:38Yeah.
179. 10:39I think part of the issue is that there's simply so much
180. 10:42change going on, it's going to be very difficult for companies and
181. 10:48clients alike
182. 10:50to sort of decide what's the best choice here.
183. 10:54And there’s a lot of muddiness in these terms, so when I say,
184. 10:56“Artificial Intelligence
185. 10:57what does that actually mean?
186. 10:59People’ve been refusing– using AI to refer to various kinds of machine learning as far back as like the 90s,
187. 11:06and they have this idea of, oh, “Deep learning! We’re doing
188. 11:08cool stuff where we're letting the the system train itself.
189. 11:12And, neural networks! And you know…
190. 11:15different types of neural networks! Like,
191. 11:17convolutional neural networks!”
192. 11:19So, anytime you say someone talk about AI,
193. 11:22we’re going to have to really get people up to speed on…
194. 11:25asking important critical questions, like
195. 11:28You say AI, what does that actually mean?
196. 11:30What– what does that mean in this context?
197. 11:33Does that mean you're using a 30-year old technology that works
198. 11:35pretty well, or does it mean you're using something that came out of Academia
199. 11:39last year, and is really cool but not very practical?
200. 11:42Or something that came out of Academia last year and has a lot of practical applications, like …
201. 11:48and getting
202. 11:49end users up to speed like that
203. 11:52is a process and– and part of what I do in sales and training is explaining these things
204. 11:56so that people are not getting ripped off
205. 11:58and they're getting exactly what they need in order to do
206. 12:01what they want to accomplish
207. 12:03In fact, there is a great paper,
208. 12:06“Hidden Technical Debt in Machine Learning Systems” and you can look up this paper for yourself,
209. 12:10but it's like
210. 12:12people think that machine learning is just machine learning code
211. 12:14but in this paper, there’s a wonderful picture where like
212. 12:17there’s this tiny little box that says ML code
213. 12:20and everything around that box is image intake,
214. 12:23data handling, deployment, Edge infrastructure,
215. 12:27human training… right?
216. 12:29So, if you think you're getting this whole pile of awesome capabilities,
217. 12:34make sure you're not paying for this teeny little box that's cool,
218. 12:38but useless on its own.
219. 12:42I think I might have to have our students investigate that a little bit more.
220. 12:45I mean, part of something we want to cover in class.
221. 12:48Thanks for that tip.
222. 12:50Well, we have a couple of extra minutes, maybe.
223. 12:53Could you maybe comment on
224. 12:55greenwashing? Is greenwashing a problem in your sector, do you think?
225. 12:59So, because I’m in a different space of AI,
226. 13:03it is I think not so much of a problem, right?
227. 13:05Because I work in spaces like
228. 13:08rail car inspection and wheel inspection.
229. 13:14We’re not using tremendous amounts of water and electricity to do these things.
230. 13:20And, of course, if we are able to prevent train derailments by
231. 13:24looking at trains as they move past an inspection station, say,
232. 13:28oops, there’s a crack in that wheel. You might be about to derail
233. 13:30Well, every time we stop a train derailment we have
234. 13:34helped the environment
235. 13:36in some tiny, teeny way.
236. 13:39But for other people, I'm sure there might be a lot of
237. 13:42desire to sort of hide or obfuscate
238. 13:46what they're doing um but, I– I don't wanna –
239. 13:49I can't speak specifically to any one– one thing, but I would say
240. 13:53greenwashing in general is a concern for any company, right?
241. 13:56Like, someone says: “Oh, I’m gonna plant a tree!” Well, that's  great,
242. 14:00and then what are you making sure that tree actually stays planted for 30 years?
243. 14:04Are you– are you then going to bury that tree in a mine to–
244. 14:08to capture the carbon in the mine like what are you doing,
245. 14:11once you planted that tree?
246. 14:13Or are they just planting a tree and then digging it up an hour  later to plant the same tree again for someone else?
247. 14:19Obviously, that's an exaggeration but, be aware of what you're– what you're getting.
248. 14:23That brings me to one final question, as well, how how do you balance ethics in with your work?
249. 14:30I think, for me personally, it's just always about
250. 14:35focusing on those core ideals of
251. 14:39using technology to make the world a better place.
252. 14:42But technology is just a tool.
253. 14:47If you spend your time building tools to hurt people,
254. 14:51that's what you're gonna get.
255. 14:52If you spend your time building tools to help people, that's what you're gonna get
256. 14:55I've said this before, I'll say it again:
257. 14:57it's the human element. What are you what are you sending your  human efforts towards?
258. 15:03What are you putting your political efforts towards, in
259. 15:08keeping your outcomes ethical,
260. 15:15because it's very easy for people to–
261. 15:18to obfuscate what they're actually doing…
262. 15:21and people in the past have used technology
263. 15:26in terrible ways. We talked about the Luddites, but
264. 15:28like, that’s not ancient history. You could look up the history of, say, the Dodge radical union movement,
265. 15:33and the really atrocious working conditions in U.S. plants,
266. 15:37even, you know, back to the ’60s, ’70s, ’80s, ’90s.
267. 15:42Those weren't necessary, it wasn't necessary to have places plants that were that dangerous.
268. 15:46And it  was the work of people who said: “Hey,
269. 15:49manufacturing should be good for its employees as well as the companies,”
270. 15:52that really made that change.
271. 15:54And not only making money, on the backs of others.
272. 15:57Well, you– everyone should make money, right?
273. 15:58Well yeah.
274. 16:00Isn't that– isn't that the promise of– of capitalism?
275. 16:02Of course it is, but we should do it ethically as well, yes?
276. 16:07Okay, well, that's unfortunately all the time we have today.
277. 16:11I'm sure we could talk much longer,
278. 16:13especially about the Luddites,
279. 16:15So, I'd like to thank you Daniel for your time, and answering our questions,
280. 16:18and we hope to see you again soon!
281. 16:21All right, thank you very much! Thanks, Keith.