1 0:00:00 –> 0:00:03 Hi everyone, this is Lattice Mage.

2 0:00:03 –> 0:00:11 This week marks the final section of my lecture of AI Alchemy Workshop.

3 0:00:11 –> 0:00:18 This lesson is about to teach people how to make their self of the how to create AI workflow,

4 0:00:18 –> 0:00:26 to make AI agent, to create what they want, and spill into nodes, about nodes.

5 0:00:26 –> 0:00:34 So one node, transparent their message to another, and they will pass the node to another next.

6 0:00:34 –> 0:00:42 So this workflow will more like the factory, we will make each data, we will push to another stage,

7 0:00:42 –> 0:00:46 and finally get the target output.

8 0:00:46 –> 0:00:55 So since we already covered the main topic in previous section, the last class is not a real class,

9 0:00:55 –> 0:00:56 it’s more like case study.

10 0:00:56 –> 0:01:01 So I focusing on SEO, that is search engine optimization.

11 0:01:01 –> 0:01:08 So we will apply several concepts discussed earlier, such as the most important thing is that

12 0:01:08 –> 0:01:17 first, we need to thinking about that, my question is we are solving the blue ocean problem or red ocean problem.

13 0:01:17 –> 0:01:25 Because if we are in red ocean market, we need to heavily to increase our ability,

14 0:01:25 –> 0:01:33 and we need very powerful AI, and we need expensive hardware to achieve high quality.

15 0:01:33 –> 0:01:44 But however, if we are solving the blue ocean problem, so we just need such as D score or C score to get better output,

16 0:01:44 –> 0:01:47 that will benefit our output.

17 0:01:47 –> 0:01:54 So the first thing is that is the SEO domain, we need to apply our domain knowledge

18 0:01:54 –> 0:01:57 to verify what we are going to do.

19 0:01:57 –> 0:02:08 And second, we need to use the SEO domain knowledge to apply into the node segmentation to build into workflow,

20 0:02:08 –> 0:02:13 and make our article writing into several steps.

21 0:02:13 –> 0:02:20 And for each step, we will give the agent AI to give the special part,

22 0:02:20 –> 0:02:23 what the current stage the domain knowledge need to reach.

23 0:02:23 –> 0:02:27 To write a better content.

24 0:02:27 –> 0:02:35 For example, we will speed into the first stage is about we need to collecting data on the website,

25 0:02:35 –> 0:02:41 or go to library to search many information, knowledge, anyway.

26 0:02:41 –> 0:02:46 So to collect as our own, more like this is our weapons.

27 0:02:46 –> 0:02:52 And next, we are going to set our target audience TA,

28 0:02:52 –> 0:03:00 to make sure that our article is going to help people to find our articles to solve their problem.

29 0:03:00 –> 0:03:06 And third, our writing need to doing search engine optimization

30 0:03:06 –> 0:03:14 to make the article able to help people and help search engine to give us higher rank.

31 0:03:14 –> 0:03:21 And finally, we will need to write some community articles such as Facebook,

32 0:03:22 –> 0:03:32 Twitter, Threads, or IG to make our website more viewable on the website,

33 0:03:32 –> 0:03:34 have more flux.

34 0:03:34 –> 0:03:40 Okay, so after setting up the theoretical framework,

35 0:03:40 –> 0:03:43 we are trying to implementation this concept.

36 0:03:43 –> 0:03:50 Currently, I’m trying to understand how MetaGPT and chat dev can be modified.

37 0:03:50 –> 0:03:53 Both are for programming frameworks,

38 0:03:53 –> 0:03:57 but I believe it’s more easy to convert into writing frameworks,

39 0:03:57 –> 0:04:00 because we don’t need to debug.

40 0:04:00 –> 0:04:04 So the coding framework will more easy to convert into writing.

41 0:04:04 –> 0:04:06 They just need to writing.

42 0:04:06 –> 0:04:13 And maybe we will create another AI and set him as our target audience

43 0:04:13 –> 0:04:16 to read this article and give us score.

44 0:04:16 –> 0:04:18 Is this higher enough or it’s low quality?

45 0:04:18 –> 0:04:19 We need to improve that.

46 0:04:20 –> 0:04:23 So I’m trying to reading the source code.

47 0:04:23 –> 0:04:26 For example, try to understand chat dev.

48 0:04:26 –> 0:04:31 And now I found that chat dev is based on the KMO framework.

49 0:04:31 –> 0:04:34 So I think it’s more easy to run on Linux.

50 0:04:34 –> 0:04:38 And I’m researching how to run it on Docker,

51 0:04:38 –> 0:04:45 because in our engineering, we don’t like to make our environment complex.

52 0:04:45 –> 0:04:50 So we usually use Docker to make environment isolate.

53 0:04:50 –> 0:04:51 OK.

54 0:04:51 –> 0:04:55 And as we said, we have two solutions.

55 0:04:55 –> 0:04:57 Another solution is MetaGPT.

56 0:04:57 –> 0:05:01 And the MetaGPT maybe need to install NPM

57 0:05:01 –> 0:05:07 because they may installation the mermaid to draw diagrams or not.

58 0:05:07 –> 0:05:11 And chat dev seems not provide this feature.

59 0:05:11 –> 0:05:19 And the most important thing is that if there’s a possible AI able to agent repo,

60 0:05:19 –> 0:05:22 if able to call Ollama API,

61 0:05:22 –> 0:05:26 that will be better because we can use Lama3 in our model.

62 0:05:26 –> 0:05:29 But currently, if we want speed,

63 0:05:29 –> 0:05:39 it’s better to use GPT-3 or GPT-4 to get our such as the MVP product.

64 0:05:39 –> 0:05:44 So it’s not feasible to use MVCard or AMD Card.

65 0:05:44 –> 0:05:46 But in the future,

66 0:05:46 –> 0:05:49 because we always want to calculate on our local

67 0:05:49 –> 0:05:50 privacy issue,

68 0:05:50 –> 0:05:55 so we still need LLM to run our PC.

69 0:05:55 –> 0:06:00 So the one solution is that we can buy Nvidia GPU Card,

70 0:06:00 –> 0:06:03 but that it’s incredibly expensive.

71 0:06:03 –> 0:06:07 So another alternative, it’s the Apple computer.

72 0:06:07 –> 0:06:13 Apple has the M2 chips that the price is not so expensive,

73 0:06:13 –> 0:06:14 but still high.

74 0:06:14 –> 0:06:16 So last year,

75 0:06:16 –> 0:06:18 I guess the solution will maybe the M2 chips.

76 0:06:18 –> 0:06:21 Because AMD have the APU,

77 0:06:21 –> 0:06:27 that the CPU GPU will share the CPU memory as their GPU memory.

78 0:06:27 –> 0:06:35 So we just need to extend the CPU memory to such as 128 or even more,

79 0:06:35 –> 0:06:41 such as 196, 92 kind of the CPU RAM.

80 0:06:41 –> 0:06:44 Then we will get that GPU RAM.

81 0:06:44 –> 0:06:47 And we can put that such as the Lama3,

82 0:06:47 –> 0:06:49 on our computer.

83 0:06:49 –> 0:06:52 So I think this is the possible solution.

84 0:06:52 –> 0:06:56 And another important thing is about the file reading,

85 0:06:56 –> 0:06:59 because we are going to writing SEO files.

86 0:06:59 –> 0:07:02 So as we talked before,

87 0:07:02 –> 0:07:06 our first stage is that we need to collection our knowledge base.

88 0:07:06 –> 0:07:09 And based on this knowledge library,

89 0:07:09 –> 0:07:13 our personal library to writing to articles.

90 0:07:13 –> 0:07:16 So of course we need to read the files.

91 0:07:16 –> 0:07:20 So the RAG system is very important.

92 0:07:20 –> 0:07:25 So that is we need our AI agent able to read the files,

93 0:07:25 –> 0:07:28 and even able to index where the files.

94 0:07:28 –> 0:07:32 Okay, so this is kind of the flow.

95 0:07:32 –> 0:07:37 And for SEO optimization problem,

96 0:07:37 –> 0:07:41 we also need a high performance of our website.

97 0:07:41 –> 0:07:46 Because Google search will score your website,

98 0:07:46 –> 0:07:49 by your performance and your security.

99 0:07:49 –> 0:07:53 Such as you will always need to use HTTPS.

100 0:07:53 –> 0:07:55 And your image,

101 0:07:55 –> 0:07:59 it’s worse if you use JPEG or PNG.

102 0:07:59 –> 0:08:03 That it’s all the format for the website.

103 0:08:03 –> 0:08:05 The better way is use WebP.

104 0:08:05 –> 0:08:09 WebP is the best for the browsers.

105 0:08:09 –> 0:08:13 For years ago, such as Mac OS or Apple series,

106 0:08:13 –> 0:08:15 not able to read the WebP.

107 0:08:15 –> 0:08:17 But recent years,

108 0:08:17 –> 0:08:20 they are all able to accept WebT now.

109 0:08:20 –> 0:08:23 So if you have any image,

110 0:08:23 –> 0:08:25 just convert into WebP.

111 0:08:25 –> 0:08:29 And WebP will use the compressed algorithm

112 0:08:29 –> 0:08:32 to make the image faster and streamable.

113 0:08:32 –> 0:08:35 Okay, and your website.

114 0:08:35 –> 0:08:39 Your website always need to design for mobile first.

115 0:08:39 –> 0:08:42 Because the most users are on the mobile.

116 0:08:42 –> 0:08:44 And Google also will,

117 0:08:44 –> 0:08:47 they will qualify your website

118 0:08:47 –> 0:08:51 by how friendly you target on the mobile.

119 0:08:51 –> 0:08:52 And the next,

120 0:08:52 –> 0:08:55 the most important thing is about the links issue.

121 0:08:55 –> 0:08:56 The links,

122 0:08:56 –> 0:08:59 it’s better that other page have your website link.

123 0:08:59 –> 0:09:01 But this is harder.

124 0:09:01 –> 0:09:04 So another solution is that your website

125 0:09:04 –> 0:09:10 need to link your personal pages to link internal.

126 0:09:10 –> 0:09:13 And such as if you have community,

127 0:09:13 –> 0:09:15 or other platform,

128 0:09:15 –> 0:09:18 it’s better to cross link,

129 0:09:18 –> 0:09:22 make your articles cross reference to others,

130 0:09:22 –> 0:09:25 such as videos, articles,

131 0:09:25 –> 0:09:27 and maybe audio podcast.

132 0:09:27 –> 0:09:31 And link them to make like a net or a network.

133 0:09:31 –> 0:09:37 To make your resource are point to each other.

134 0:09:37 –> 0:09:40 That will make a better, higher score.

135 0:09:40 –> 0:09:42 And that is why

136 0:09:42 –> 0:09:45 we are in the previous,

137 0:09:45 –> 0:09:49 we talking about the workflow of the SEO articles.

138 0:09:49 –> 0:09:53 The final stage is that better to writing Facebook,

139 0:09:53 –> 0:09:54 Twitter, Threads, IG,

140 0:09:54 –> 0:09:57 this kind of community platform.

141 0:09:57 –> 0:10:02 And we need to understand that the SEO articles,

142 0:10:02 –> 0:10:05 it’s for long term.

143 0:10:05 –> 0:10:08 And it will solve the key problem.

144 0:10:08 –> 0:10:11 And for the community platform,

145 0:10:11 –> 0:10:13 it’s not solving the problem.

146 0:10:13 –> 0:10:16 It’s just follow up some hot news.

147 0:10:16 –> 0:10:21 And you can just quick to get the flow of the people

148 0:10:21 –> 0:10:26 viewing your small threads or tweets.

149 0:10:26 –> 0:10:27 Yes.

150 0:10:27 –> 0:10:28 Okay.

151 0:10:28 –> 0:10:31 So this is kind of the SEO part.

152 0:10:31 –> 0:10:34 How we try to optimization with the AI.

153 0:10:34 –> 0:10:36 And another part,

154 0:10:36 –> 0:10:38 it’s about the,

155 0:10:38 –> 0:10:40 I’m reading about the game design book.

156 0:10:40 –> 0:10:42 This week.

157 0:10:42 –> 0:10:44 So it’s more about the writing.

158 0:10:44 –> 0:10:45 And the book,

159 0:10:45 –> 0:10:48 we will have the club to share our view.

160 0:10:48 –> 0:10:50 And for me,

161 0:10:50 –> 0:10:51 this week,

162 0:10:51 –> 0:10:54 the most important thing in the book is about

163 0:10:54 –> 0:10:56 your size of the,

164 0:10:56 –> 0:10:59 your length of your story

165 0:10:59 –> 0:11:03 must corresponds to the depth of your story.

166 0:11:03 –> 0:11:04 For example,

167 0:11:04 –> 0:11:06 if your story have a,

168 0:11:06 –> 0:11:09 such as your movie is very long.

169 0:11:09 –> 0:11:10 For example,

170 0:11:10 –> 0:11:11 two hours.

171 0:11:11 –> 0:11:16 So your story need to have a complex dilemma,

172 0:11:16 –> 0:11:19 complex contradiction to make them,

173 0:11:19 –> 0:11:25 the story will have a complex progress

174 0:11:25 –> 0:11:29 and make people can understand more deeper about a story.

175 0:11:29 –> 0:11:32 So in other words,

176 0:11:32 –> 0:11:35 if you have a long story,

177 0:11:35 –> 0:11:38 but you were on your dilemma,

178 0:11:38 –> 0:11:39 it’s just,

179 0:11:39 –> 0:11:41 the shadow not deeply enough.

180 0:11:41 –> 0:11:43 People will feel that boring.

181 0:11:43 –> 0:11:45 So in other side,

182 0:11:45 –> 0:11:48 if your movie time is very short,

183 0:11:48 –> 0:11:52 such as minutes or just 10 minutes,

184 0:11:52 –> 0:11:56 you should not write a complex story in that,

185 0:11:56 –> 0:11:58 because that will make people confused.

186 0:11:58 –> 0:12:02 And they cannot understand because you don’t have enough time to come,

187 0:12:02 –> 0:12:06 convey what you are really deep going to think.

188 0:12:06 –> 0:12:07 Okay.

189 0:12:07 –> 0:12:08 And for the next,

190 0:12:08 –> 0:12:09 the problem.

191 0:12:09 –> 0:12:10 Yeah.

192 0:12:10 –> 0:12:11 So for game design,

193 0:12:11 –> 0:12:14 I’m the writer of the KK Academy.

194 0:12:14 –> 0:12:20 We are going to game fight our drawing club to make our drawing society.

195 0:12:20 –> 0:12:23 People will feel like playing a game.

196 0:12:23 –> 0:12:24 For example,

197 0:12:24 –> 0:12:26 we will apply D&D story,

198 0:12:26 –> 0:12:27 D&D framework,

199 0:12:27 –> 0:12:30 D&D world on our society.

200 0:12:30 –> 0:12:33 And we will spill the story.

201 0:12:33 –> 0:12:35 And we will make a game.

202 0:12:35 –> 0:12:36 And we will make a game.

203 0:12:36 –> 0:12:40 And we will spill into three major part.

204 0:12:40 –> 0:12:41 For example,

205 0:12:41 –> 0:12:42 it’s color.

206 0:12:42 –> 0:12:45 Color is one of the major drawing subject.

207 0:12:45 –> 0:12:48 And another part is composition.

208 0:12:48 –> 0:12:52 And composition is another also important drawing part.

209 0:12:52 –> 0:12:55 And another one is perspective.

210 0:12:55 –> 0:13:04 So we will have these three major subjects to domain our drawing field.

211 0:13:04 –> 0:13:05 So,

212 0:13:05 –> 0:13:09 we will convert that into different skill set.

213 0:13:09 –> 0:13:10 Like,

214 0:13:10 –> 0:13:11 for example,

215 0:13:11 –> 0:13:12 Diablo,

216 0:13:12 –> 0:13:13 your career,

217 0:13:13 –> 0:13:18 your role will have three different system of your skills.

218 0:13:18 –> 0:13:19 So,

219 0:13:19 –> 0:13:22 we will create color skills,

220 0:13:22 –> 0:13:24 composition skills,

221 0:13:24 –> 0:13:31 and perspective skills to make people able to use them to fighting with the monsters.

222 0:13:31 –> 0:13:32 So,

223 0:13:32 –> 0:13:34 we are going to merge that into our TRPG.

224 0:13:35 –> 0:13:37 And further,

225 0:13:37 –> 0:13:40 because we have the Discord,

226 0:13:40 –> 0:13:42 so the Discord society,

227 0:13:42 –> 0:13:43 the robot,

228 0:13:43 –> 0:13:47 it can calculate how much money you have.

229 0:13:47 –> 0:13:50 And you can spend money to buy something.

230 0:13:50 –> 0:13:51 So,

231 0:13:51 –> 0:13:54 after we creating this TRPG game,

232 0:13:54 –> 0:13:58 we will let people able to use Discord robot,

233 0:13:58 –> 0:14:02 the inner robot Shudo money,

234 0:14:02 –> 0:14:05 that to buy the equipment,

235 0:14:05 –> 0:14:07 equipment in that game.

236 0:14:07 –> 0:14:08 So,

237 0:14:08 –> 0:14:12 they need to join like the part-time job.

238 0:14:12 –> 0:14:13 If they join longer,

239 0:14:13 –> 0:14:17 they will get more money from the robot.

240 0:14:17 –> 0:14:18 So,

241 0:14:18 –> 0:14:22 it will make it like the closure to play the game and join.

242 0:14:22 –> 0:14:23 So,

243 0:14:23 –> 0:14:25 we think this will be funny.

244 0:14:25 –> 0:14:26 And also,

245 0:14:26 –> 0:14:30 we will also make that this like a academy school.

246 0:14:30 –> 0:14:31 So,

247 0:14:31 –> 0:14:35 people will like the student to learning the magic.

248 0:14:35 –> 0:14:36 So,

249 0:14:36 –> 0:14:39 they will be able to play the game in that academy.

250 0:14:39 –> 0:14:41 And people will use these skills,

251 0:14:41 –> 0:14:42 use these spells,

252 0:14:42 –> 0:14:46 use the casting the spells to fighting with the monsters.

253 0:14:46 –> 0:14:47 So,

254 0:14:47 –> 0:14:51 they need to keep drawing to make their character growth up.

255 0:14:51 –> 0:14:55 And fighting to the monsters.

256 0:14:55 –> 0:14:56 Okay.

257 0:14:56 –> 0:14:57 Yeah.

258 0:14:57 –> 0:15:00 And the final part is that,

259 0:15:00 –> 0:15:03 although I quit my job last year,

260 0:15:03 –> 0:15:04 but now,

261 0:15:04 –> 0:15:05 I start up,

262 0:15:05 –> 0:15:08 it’s going to take fewer times.

263 0:15:08 –> 0:15:09 It’s okay.

264 0:15:09 –> 0:15:10 So,

265 0:15:10 –> 0:15:17 I’m going to find a normal job such as full-time coding job about algorithm or I’m not sure,

266 0:15:17 –> 0:15:23 maybe GPU related or cloud-based related jobs.

267 0:15:23 –> 0:15:24 Yeah.

268 0:15:24 –> 0:15:25 To make our,

269 0:15:25 –> 0:15:26 make my money better.

270 0:15:26 –> 0:15:29 Because in the last month,

271 0:15:29 –> 0:15:31 I’m doing algo trading.

272 0:15:31 –> 0:15:34 And after we finish develop the algo trading,

273 0:15:34 –> 0:15:37 I’m going to start to do some trading strategies.

274 0:15:37 –> 0:15:39 The key problem is that,

275 0:15:39 –> 0:15:46 how much base money you have will reflect how much your profit.

276 0:15:46 –> 0:15:47 For example,

277 0:15:47 –> 0:15:49 if you have only few,

278 0:15:49 –> 0:15:50 for example,

279 0:15:50 –> 0:15:53 just 10,000 money,

280 0:15:53 –> 0:15:56 and you have a huge profit ratio,

281 0:15:56 –> 0:15:59 such as 200,

282 0:15:59 –> 0:16:00 2000,

283 0:16:00 –> 0:16:01 200,

284 0:16:01 –> 0:16:03 200% of the profit.

285 0:16:03 –> 0:16:09 Your 10,000 money will become 3000 money.

286 0:16:09 –> 0:16:10 It’s not much.

287 0:16:10 –> 0:16:14 But if you have 100,000 money,

288 0:16:14 –> 0:16:19 you just get only one fifth of the profit ratio.

289 0:16:19 –> 0:16:22 You still get a large amount of the reward.

290 0:16:22 –> 0:16:23 So,

291 0:16:23 –> 0:16:32 the most important thing is that you need to accumulate your base money to make it to enlarge the efficiency.

292 0:16:32 –> 0:16:35 Of the strategy of trading.

293 0:16:35 –> 0:16:36 So,

294 0:16:36 –> 0:16:38 that is why I’m finding a job.

295 0:16:38 –> 0:16:40 And also,

296 0:16:40 –> 0:16:42 another thing is that,

297 0:16:42 –> 0:16:44 what you can do,

298 0:16:44 –> 0:16:50 it’s based on the how much level you know about the world or skills.

299 0:16:50 –> 0:16:51 For example,

300 0:16:51 –> 0:16:53 for the gold chess,

301 0:16:53 –> 0:16:56 it’s not about how long time you calculation.

302 0:16:56 –> 0:16:59 You need to move the next step.

303 0:16:59 –> 0:17:00 The important thing is that,

304 0:17:00 –> 0:17:12 how much you understand the basic ability to understand the shape of the complexity changing of the chess moving.

305 0:17:12 –> 0:17:14 So,

306 0:17:14 –> 0:17:20 if you don’t have the ability to understand that the complex chess moving,

307 0:17:20 –> 0:17:22 actually,

308 0:17:22 –> 0:17:24 if you still spend a lot of time,

309 0:17:24 –> 0:17:28 you still get a fewer outcome.

310 0:17:28 –> 0:17:29 So,

311 0:17:29 –> 0:17:33 this is also similar to the investing part.

312 0:17:33 –> 0:17:40 That you need to understand more to get fast but higher quality.

313 0:17:40 –> 0:17:45 That is why I think my startup is almost close to that.

314 0:17:45 –> 0:17:49 I can use fewer time but still get higher quality.

315 0:17:49 –> 0:17:50 So,

316 0:17:50 –> 0:17:52 I think this is a very important thing.

317 0:17:52 –> 0:17:53 Yeah.

318 0:17:53 –> 0:17:54 Okay.

319 0:17:54 –> 0:17:55 So,

320 0:17:55 –> 0:17:56 this is what we’ve done this week.

321 0:17:56 –> 0:17:57 Thank you,

322 0:17:57 –> 0:17:58 everyone.