YouTube2Text

Managing AI - CompTIA A+ 220-1202 - 4.10 — Transcript

by Professor Messer · 1,775 words · 266 segments · language en · Watch on YouTube

Full transcript

  1. 0:02AI is an abbreviation for artificial
  2. 0:04intelligence and the goal of AI is to
  3. 0:07meet or exceed human intelligence in the
  4. 0:10form of technology. This means that our
  5. 0:13computers would be able to learn, infer,
  6. 0:16and reason based on input that we
  7. 0:18provide. This is obviously not a new
  8. 0:21idea. We've been working on concepts of
  9. 0:23AI for a number of decades. And of
  10. 0:25course, you see AI mentioned in numerous
  11. 0:27science fiction stories. But now, of
  12. 0:30course, AI has become much more
  13. 0:32prevalent. We have new tools that allow
  14. 0:33us to create new content such as text,
  15. 0:36audio, and video by simply asking a
  16. 0:39computer to create it.
  17. 0:42And very quickly, we have seen AI become
  18. 0:45integrated into most applications that
  19. 0:47we use every day. For example, if you
  20. 0:49perform a search on a popular search
  21. 0:51engine, you may find not only your
  22. 0:53search results, but also a summary of
  23. 0:56those results that was created using
  24. 0:58artificial intelligence. The goal is to
  25. 1:00take information that normally you would
  26. 1:02need to find across multiple sites and
  27. 1:05present all of that information to you
  28. 1:06as one single AI summary. We've also
  29. 1:10seen AI in our email clients and on our
  30. 1:12mobile devices. Instead of having to
  31. 1:15read through a long email, you can read
  32. 1:17an AI summary that explains the
  33. 1:19information that was contained in that
  34. 1:21message. And as you've probably seen in
  35. 1:23graphics editors, you can use generative
  36. 1:25AI to fill in or remove information from
  37. 1:28an image by simply highlighting it on
  38. 1:30the screen and asking the AI to remove
  39. 1:33it. We've also seen artificial
  40. 1:35intelligence engines that allow you to
  41. 1:37input a description of what you'd like
  42. 1:39an image to be and the AI will create
  43. 1:42that image on the screen.
  44. 1:45We've seen a number of different uses of
  45. 1:47AI that we could put into the category
  46. 1:49of appropriate AI use. For example,
  47. 1:53there may be extremely large data
  48. 1:55repositories that we would like AI to be
  49. 1:57able to sift through and provide us with
  50. 2:00a summary of that information. For
  51. 2:02example, there's AI that can look
  52. 2:03through terabytes and terabytes of log
  53. 2:06files in an organization and find
  54. 2:08potential security issues that may be
  55. 2:10hidden in all of these different logs.
  56. 2:13This would certainly be timeconuming and
  57. 2:15difficult for a human to do, which makes
  58. 2:17it perfect for AI. We've also started
  59. 2:20integrating AI into our scripting and
  60. 2:22our automation so that AI can make
  61. 2:24decisions on what to do when a certain
  62. 2:26situation occurs. Being able to do this
  63. 2:29without human intervention means that we
  64. 2:31can spend our time doing something else.
  65. 2:34We're also seeing AI integrated more
  66. 2:36into health care. For example, AI can
  67. 2:38take the output from an MRI scanner and
  68. 2:41evaluate that to see if there may be
  69. 2:43something that would be concerning. We
  70. 2:45might also have AI go through healthcare
  71. 2:47records and determine if anyone is
  72. 2:49receiving multiple drugs that might have
  73. 2:51negative interactions. This means the
  74. 2:54health care professionals can work with
  75. 2:55us at a human level and have AI take
  76. 2:58care of everything at a digital level.
  77. 3:01And if you're watching this video on
  78. 3:02YouTube, you may be getting realtime
  79. 3:04language translation thanks to the AI
  80. 3:06that's built into our new communications
  81. 3:08apps.
  82. 3:10But of course, there are downsides to
  83. 3:12any technology and AI has been used for
  84. 3:15things that may not be as appropriate.
  85. 3:17For example, we've seen AI used for
  86. 3:19fraud. AI can impersonate a real person
  87. 3:22and even create video of that person,
  88. 3:25something we refer to as a deep fake.
  89. 3:28We've also seen application code that's
  90. 3:30created with AI without any knowledge of
  91. 3:32what the application is actually doing.
  92. 3:34Or perhaps someone is using AI to create
  93. 3:37an image and taking credit for that
  94. 3:39image creation. And we've seen AI used
  95. 3:42in schools and colleges to plagiarize
  96. 3:44works. Students are using AI to
  97. 3:46summarize existing works, but they're
  98. 3:48not citing AI in that summary.
  99. 3:52One of the challenges we have when
  100. 3:53working with AI is that it can sometimes
  101. 3:55have a bias. The AI only knows the
  102. 3:58information that we're feeding into it
  103. 4:00and it has a predefined set of
  104. 4:02algorithms on how to manage that data.
  105. 4:04And sometimes AI can make the wrong
  106. 4:07conclusions based on this combination of
  107. 4:09data and algorithms. For example, we may
  108. 4:12be feeding AI a very large data set
  109. 4:15associated with health care in the
  110. 4:16United States, but inside of that data
  111. 4:19set, we may have under represented
  112. 4:21statistics for a particular ethnic group
  113. 4:23or particular gender group. In this
  114. 4:25case, AI could potentially create
  115. 4:27conclusions that are biased one way or
  116. 4:30the other. We're obviously not trying to
  117. 4:32create bias as part of this AI analysis.
  118. 4:35So we have to make sure that we create
  119. 4:37an AI algorithm that doesn't build bias
  120. 4:40when we have these large data sets. A
  121. 4:42good example of this is what Amazon
  122. 4:44found when they took 10 years of
  123. 4:46submitted résumés and put them into an
  124. 4:49AI engine. This grouping of résumés had
  125. 4:52a number of terms such as executed and
  126. 4:55captured as part of the résé text. The
  127. 4:57artificial intelligence engine tended to
  128. 5:00pick résumés that had those terms as
  129. 5:02part of the resume itself. What we found
  130. 5:05is that those terms were more often used
  131. 5:08when it was a male submitting a resume
  132. 5:10versus a female and therefore the AI
  133. 5:13created a bias towards male résumés.
  134. 5:16Obviously, Amazon chose not to use this
  135. 5:19AI algorithm for their resume research.
  136. 5:23There's also occasions where AI can
  137. 5:25completely misrepresent the data that it
  138. 5:27was provided. You can think of this as a
  139. 5:30confidently incorrect AI where it really
  140. 5:33knows the answer but the answer is
  141. 5:35completely wrong. For example, let's say
  142. 5:37that we've told the AI that any picture
  143. 5:39containing a snout, a tail, and four
  144. 5:42legs is a dog. Then we provide the AI
  145. 5:45with a picture of a cloud and the cloud
  146. 5:47appears to have a nose, a tail, and four
  147. 5:50legs. Therefore, the cloud is obviously
  148. 5:52a real dog. This is obviously a
  149. 5:55simplified way to describe an AI
  150. 5:57hallucination, but this is what tends to
  151. 5:59occur when we have this type of data
  152. 6:01misinterpretation. These are also not
  153. 6:04very unusual to find, and you may have
  154. 6:06seen an AI hallucination when performing
  155. 6:08your own tests with an AI engine. A
  156. 6:11practical example of an AI hallucination
  157. 6:14was when Google and Microsoft announced
  158. 6:16their new AI technologies. Google
  159. 6:18announced Google Bard. This is now
  160. 6:20Google Gemini. and Microsoft introduced
  161. 6:23Microsoft Bing Chat. Both of these
  162. 6:25engines provided a summary of some
  163. 6:27articles that had completely wrong
  164. 6:29information inside of them.
  165. 6:31Unfortunately, this hallucination was in
  166. 6:34the announcement that Google made
  167. 6:36regarding their new AI engine. This AI
  168. 6:39engine took an input that said, "What
  169. 6:41new discoveries from the James Web Space
  170. 6:43Telescope can I tell my 9-year-old
  171. 6:45about?" And then it gave a number of
  172. 6:47bullets that you could use to help
  173. 6:48describe the James Webb Space Telescope
  174. 6:51and its findings. Unfortunately, one of
  175. 6:53these findings was that the James Webb
  176. 6:55Space Telescope took the very first
  177. 6:57picture of a planet outside of our own
  178. 7:00solar system. Bruce Macintosh replied to
  179. 7:03this and said, "Speaking as someone who
  180. 7:04imaged an exoplanet 14 years before the
  181. 7:07JW Space Telescope was launched, it
  182. 7:10feels like you should find a better
  183. 7:12example." This points out that even the
  184. 7:14best AI engines can suffer from an AI
  185. 7:17hallucination.
  186. 7:20Behind the scenes, artificial
  187. 7:21intelligence is creating a conclusion
  188. 7:23using a model that is provided. And
  189. 7:26sometimes it makes the wrong conclusion
  190. 7:28based on the model that it was given.
  191. 7:30And researchers will often perform tests
  192. 7:32of an AI engine to see how accurate it
  193. 7:35might be. They will get predictions from
  194. 7:37the AI and then they will compare those
  195. 7:39predictions to known test data. One of
  196. 7:42these accuracy tests was done by
  197. 7:44originality.ai. They did an AI
  198. 7:46fact-checking accuracy study in August
  199. 7:48of 2024, and you can read through the
  200. 7:51study yourself at the URL that you see
  201. 7:53here. They evaluated their own AI and a
  202. 7:56number of other AIs that were currently
  203. 7:58available. The originality.ai
  204. 8:00engine got a 72.3%
  205. 8:03accuracy. GPT4 got a 64.9%
  206. 8:07accuracy rating and GPT3.5 got a 58.6.
  207. 8:11six accuracy rating. As you can see, our
  208. 8:14current iterations of AI are far from
  209. 8:16perfect and in some cases fall well
  210. 8:19under what most people think AI is able
  211. 8:21to do.
  212. 8:23There's been a lot of work made on both
  213. 8:26public AIs and private AIs. A public AI
  214. 8:30would be one that is publicly available
  215. 8:31on the internet. Things like ChatGpt or
  216. 8:34Google Gemini are good examples of a
  217. 8:37public AI. But some companies are
  218. 8:39creating their own AI engines and these
  219. 8:41would be engines that are private to
  220. 8:43that organization. These would commonly
  221. 8:45have proprietary company data and the
  222. 8:48organization has complete control over
  223. 8:50the modeling used for that artificial
  224. 8:52intelligence engine. The decision on
  225. 8:55whether you use a public AI or a private
  226. 8:57AI may be based on data security.
  227. 9:00Whenever you put information into an AI
  228. 9:02engine, there is a possibility that that
  229. 9:04data could be retrieved. And if that
  230. 9:06data is sensitive or contains private
  231. 9:09information, you could be putting that
  232. 9:11information available for anyone on the
  233. 9:13internet to see. We've also seen cases
  234. 9:15where an AI was able to provide
  235. 9:17passwords, encryption keys, certificate
  236. 9:19details, and other sensitive
  237. 9:21information. Of course, if you're
  238. 9:23running a private AI, only people within
  239. 9:26that organization have access to that
  240. 9:28proprietary data, and therefore, it
  241. 9:30limits the scope of any type of breach.
  242. 9:32There's also quite a big difference
  243. 9:34between a public AI and a private AI
  244. 9:37when it comes to how much data they're
  245. 9:38able to evaluate. Obviously, a private
  246. 9:41AI is only going to have data from a
  247. 9:43single entity, but a public AI is
  248. 9:46collecting data from everybody who's on
  249. 9:48the internet. In those cases, it may be
  250. 9:50that your public AI is more accurate
  251. 9:52than the private AI because the public
  252. 9:54AI has access to so much more data. But
  253. 9:57it is that large amount of data that
  254. 9:59does create concerns about privacy. Some
  255. 10:02of that information that's being stored
  256. 10:04by a public AI is information about you.
  257. 10:07The AI knows where you live. It might
  258. 10:09even know your previous addresses, has
  259. 10:11information about your habits and things
  260. 10:13that you do on the internet, and it may
  261. 10:15have details about any memberships and
  262. 10:17groups that you belong to. Let's say
  263. 10:19you're applying for a job. What type of
  264. 10:21summary would the AI create of you and
  265. 10:24what would be contained within your
  266. 10:26profile?

About this transcript

This page contains the full transcript of Managing AI - CompTIA A+ 220-1202 - 4.10 by Professor Messer, generated from the public captions YouTube serves with the video. The transcript has 1,775 words across 266 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.