YouTube2Text

Emanuele Frontoni - Intelligenza Artificiale e creatività umana nell’epoca dell’Agentic AI — Transcript

by Festa Scienza Filosofia · 7,302 words · 1,249 segments · language en · Watch on YouTube

Full transcript

  1. 0:05Good morning, everyone.
  2. 0:08My name is Venus, I am a student at the
  3. 0:10Morea Vivarelli Institute, and today we
  4. 0:12have the honor of hosting Professor
  5. 0:14Emanuele Frontoni, a full professor of
  6. 0:16computer science at the University of
  7. 0:18Macerata and co-director of BRAI, which
  8. 0:20stands for Vision, Robotics, and
  9. 0:22Artificial Intelligence.
  10. 0:26Professor Frontoni is a highly prominent
  11. 0:30figure in the scientific field.
  12. 0:34His research focuses primarily on
  13. 0:37artificial intelligence, the study of
  14. 0:40human behavior, and augmented reality.
  15. 0:44He is the author of over 300 scientific
  16. 0:47publications and actively collaborates
  17. 0:50with agencies and institutions on
  18. 0:52innovation and technology transfer
  19. 0:55projects.
  20. 0:58Since 2021, he has also been included in
  21. 1:00the prestigious list of the world's most
  22. 1:02cited scientists.
  23. 1:05Since 2022, he has also served as the
  24. 1:08scientific director of the Neolab at the
  25. 1:11Niguarda Hospital in Milan.
  26. 1:18The theme of the conference will be
  27. 1:21artificial intelligence and human
  28. 1:23creativity in the age of the agent, or
  29. 1:25systems that are not only capable of
  30. 1:27responding but of acting and
  31. 1:29understanding actively and autonomously
  32. 1:32.
  33. 1:35Through concrete examples, we will
  34. 1:37explore the opportunities offered by
  35. 1:38these technologies, as well as their
  36. 1:40limitations.
  37. 1:43The ethical implications, especially in
  38. 1:45professional and educational fields,
  39. 1:47will also be discussed.
  40. 1:50Without taking any more time from our
  41. 1:52guest, I hand the floor over to
  42. 1:53Professor Emanuele Frontoni.
  43. 2:02>> Thank you.
  44. 2:04Thank you for this invitation, and thank
  45. 2:06you to Venus, who studied everything
  46. 2:08perfectly.
  47. 2:09She was far too kind in her introduction
  48. 2:11.
  49. 2:15I was recalling with Sergio Mustica, who
  50. 2:17accompanied me here and even found me a
  51. 2:19secret parking spot because parking in
  52. 2:21Fabriano wasn't exactly easy, that I had
  53. 2:23been here some time ago, when I believe
  54. 2:24the very first editions of this festival
  55. 2:26were just beginning, right in this room
  56. 2:28.
  57. 2:31So, I am happy to return, also because
  58. 2:33in the meantime—and I’d like to greet my
  59. 2:36colleague Comodi over there from the
  60. 2:39Polytechnic, with whom I’ve been linked
  61. 2:41for many years since I was at the
  62. 2:44Polytechnic—you truly deserve
  63. 2:46congratulations, because this festival,
  64. 2:48along with the one in Foligno, is
  65. 2:51becoming an important institution for
  66. 2:53this area.
  67. 2:55And thank you for hosting so many
  68. 2:56organizations in these territories.
  69. 2:59Also, I've come to Fabriano often
  70. 3:00recently, Venus, because we’ve been
  71. 3:02recording, and I discovered the
  72. 3:04existence of a very amusing group of
  73. 3:06characters, your fellow townsfolk, who
  74. 3:08are the stars of "Autovelox," this...
  75. 3:15...podcast/videocast you record, which,
  76. 3:17by the way, hosted Greg a few days ago,
  77. 3:19of the famous duo Lillo and Greg, who
  78. 3:21was here in Fabriano with them; and so
  79. 3:23we had a lot of fun with him as well.
  80. 3:28And by the way, Fabriano is also the
  81. 3:30cradle of one of the most important
  82. 3:32projects that I will tell you about in
  83. 3:34conclusion, which is leading us, right
  84. 3:37alongside Sergio, to carry out many
  85. 3:39artificial intelligence initiatives
  86. 3:41being tested in the world of schools.
  87. 3:46And this project called Paideia, I'll
  88. 3:47tell you a bit about it towards the end
  89. 3:49because I'm curious to have a dialogue
  90. 3:50with Venus as well, I’ll be the one
  91. 3:52asking her questions later.
  92. 3:54No, I'm joking, of course, just to see
  93. 3:56if she agrees with us on certain points
  94. 3:58.
  95. 3:59Now I will do a short initial piece
  96. 4:01where I tell part of this story. Then
  97. 4:04obviously we will stop and debate, so
  98. 4:06I’ll do it quickly.
  99. 4:08So, first concept: this is me.
  100. 4:14Back in the era of Ghibli Style, the
  101. 4:17epic beginnings of AI image creation.
  102. 4:22And what do you think is strange about
  103. 4:24this photo?
  104. 4:25Ignoring the fact that I'm no longer
  105. 4:27blond but have white hair, and that I
  106. 4:29was fit while here I'm a bit more out of
  107. 4:31shape—all those variations—what is
  108. 4:33strange about that photo?
  109. 4:35What do you say, guys?
  110. 4:37I have a primary school nerd crowd here
  111. 4:38.
  112. 4:40I'm looking at you, I'm not looking
  113. 4:42around.
  114. 4:43What do you say? What's strange?
  115. 4:48This photo is illegal because it's a
  116. 4:50clear copyright violation of Studio
  117. 4:52Ghibli, which is the entity that
  118. 4:55produces these comics and these designs
  119. 4:57.
  120. 4:58It is a studio that works, that makes
  121. 5:00these drawings.
  122. 5:01To create this photo, I gave a
  123. 5:03generative artificial intelligence a
  124. 5:06photo of myself, and this is also an
  125. 5:08important step, because when you give
  126. 5:11files, data, or PDFs to a generative AI
  127. 5:14or an AI agent—which I will clarify for
  128. 5:16you shortly—you are, especially in the
  129. 5:19free versions, giving away data.
  130. 5:23So often we cannot do it.
  131. 5:25We cannot do it in the school world, we
  132. 5:27cannot do it in the corporate or
  133. 5:28professional world.
  134. 5:30On the other hand, I asked it to
  135. 5:32generate a photo in Ghibli Style; I
  136. 5:34could have asked it to generate a photo
  137. 5:36in a Spider-Man costume, it would have
  138. 5:38been the same thing.
  139. 5:42And this concept implies the fact that
  140. 5:45today one of the major concerns we must
  141. 5:47have regarding this world is also the
  142. 5:49privacy of our information and
  143. 5:51intellectual property.
  144. 5:55Which sometimes also means the privacy
  145. 5:57of your information and intellectual
  146. 5:59property.
  147. 6:00Now I will speak for a bit to the
  148. 6:02so-called adult generation present here
  149. 6:05—I see a few white hairs—as well as for
  150. 6:08a bit to the other, very young component
  151. 6:11.
  152. 6:11And we will try to play a game to
  153. 6:13measure the differences.
  154. 6:15And
  155. 6:17mind you, the lecture is not for them.
  156. 6:22Their generation has private Instagram
  157. 6:26accounts in 86%of cases.
  158. 6:30Our generation has public Instagram
  159. 6:33accounts in 92%of cases.
  160. 6:37Then their generation has 2.7 Instagram
  161. 6:38accounts, and we still haven't quite
  162. 6:40figured out what they do with them. How
  163. 6:42many Instagram accounts does Venus have
  164. 6:44?
  165. 6:46Venus has two.
  166. 6:47Someone, to reach the average of 2.7,
  167. 6:49must have three.
  168. 6:51The private, the very private, the
  169. 6:53private, okay?
  170. 6:55I won't make you raise your hands, but I
  171. 6:57think anyone who has three knows it too
  172. 6:59, okay?
  173. 7:02And we struggle to manage one.
  174. 7:07My mother, at one point, last Christmas
  175. 7:11, had an Instagram account with 189
  176. 7:16followers.
  177. 7:20My oldest daughter is 15, so she has
  178. 7:22access to Instagram, the second is 13
  179. 7:24and didn't have access to Instagram at
  180. 7:26the time, the third is seven.
  181. 7:29Guess whose Instagram account Grandma
  182. 7:31Gianna's was?
  183. 7:33Okay? Not exactly hers.
  184. 7:36The second daughter tried to defend
  185. 7:38herself, but obviously, retaliation was
  186. 7:39triggered against this fake Instagram
  187. 7:41account on Grandma's cell phone.
  188. 7:44There are grandparents among you here
  189. 7:46too, so be careful, eh?
  190. 7:47Okay? Because then we can move on.
  191. 7:51So, now you students stay still, I'm
  192. 7:55questioning the adults.
  193. 7:57Can you tell me how many of you know
  194. 7:59Photomath?
  195. 8:01Students, no, eh? Hands down.
  196. 8:04How many know Photomath?
  197. 8:09You almost do, though, eh, too young.
  198. 8:12Alright kids, now you answer, how many
  199. 8:14know Photomath?
  200. 8:21So I would start taking the lecture on
  201. 8:23studying for ourselves.
  202. 8:27Okay? Because either we start knowing
  203. 8:29this world a little, or we won't even
  204. 8:32know what to talk about, eh?
  205. 8:35We won't even know how to engage,
  206. 8:37because we are an educating community,
  207. 8:39eh? It’s not just the teacher who has to
  208. 8:41take care of them, eh?
  209. 8:44Please, we are all an educating
  210. 8:46community, everyone.
  211. 8:50Now what does Photomath do?
  212. 8:55If we talked to them or to Venus, they
  213. 8:57would tell me, "Teacher, it's a bit old
  214. 8:59now, this stuff is a bit out of fashion
  215. 9:01, it doesn't work very well, we have
  216. 9:03better things to use."
  217. 9:06But with this cell phone, I can take a
  218. 9:08photo of a math problem, a generative AI
  219. 9:11solves all the steps of the math problem
  220. 9:13, I copy them into my notebook for the
  221. 9:15teacher.
  222. 9:19Okay?
  223. 9:20Are there any math teachers by chance?
  224. 9:23I don't see any hands raised.
  225. 9:25But math teachers sometimes notice,
  226. 9:27sometimes they don't.
  227. 9:31But the strangest thing that happened to
  228. 9:34me regarding Photomath is that I was at
  229. 9:36dinner with a manager of a digital
  230. 9:38company; I tell this story often, and I
  231. 9:41said to her, she has three children like
  232. 9:43ours, the same age groups: 15, 13, 7.
  233. 9:47Only the genders are different. We have
  234. 9:49girl, girl, boy; she has boy, boy, girl
  235. 9:51.
  236. 9:52She is also our friend, but she is a
  237. 9:54very relevant figure in Italy, so I said
  238. 9:56to her, "Well, but yours must use
  239. 9:57Photomath too."
  240. 10:00We were at dinner and she replied to me
  241. 10:01, "I've never heard of it at home, so in
  242. 10:03my opinion, it doesn't exist here."
  243. 10:05A terrible sentence, isn't it?
  244. 10:08One that should really make so-called
  245. 10:11adults reflect quite a bit.
  246. 10:13We often think we are the holders of
  247. 10:16knowledge, yet we know almost nothing
  248. 10:18about that digital world where we all
  249. 10:20now live so extensively, and we know
  250. 10:22almost nothing about that world of
  251. 10:25artificial intelligence in which we all
  252. 10:27live quite extensively.
  253. 10:31To which I asked, I told her, "Look, in
  254. 10:33my opinion, I wouldn't trust this answer
  255. 10:35of yours too much, let's try writing to
  256. 10:37the family WhatsApp group," so she wrote
  257. 10:39to her children on the group and asked,
  258. 10:41"Do you know Photomath?"
  259. 10:44Okay?
  260. 10:46After 30 seconds, the middle child, a
  261. 10:47thirteen-year-old in eighth grade, wrote
  262. 10:49, "Yes, but I rarely use it."
  263. 10:52Okay?
  264. 10:54Which is the same answer they would give
  265. 10:56, okay?
  266. 10:58If questioned, they would say, "No, it
  267. 11:00doesn't work that well," or perhaps
  268. 11:02later they will ask me some questions
  269. 11:03and say, "Oh, I'm asking for a friend,"
  270. 11:05but I've had things of this nature
  271. 11:07happen to us in class.
  272. 11:10To this, I add that that generation—
  273. 11:12this generation here—has a very clear
  274. 11:15understanding of a new concept that we
  275. 11:17don't have clear in our minds, which is
  276. 11:20called human-AI teaming.
  277. 11:23That is, how do we have active
  278. 11:25collaboration between humans and
  279. 11:27artificial intelligence without
  280. 11:28delegating passively, but trying to
  281. 11:30build something interesting with it.
  282. 11:35I’ll add another piece of the story to
  283. 11:36this.
  284. 11:37I was at a conference like this one at a
  285. 11:39school
  286. 11:42and the teachers bring me onto the stage
  287. 11:45and tell me, "No, we are one of the best
  288. 11:47schools in Italy here, a model school;
  289. 11:50we've blocked the Wi-Fi, you can't
  290. 11:52access ChatGPT or Photomath, and we have
  291. 11:54a school-family registry, great
  292. 11:56communication with the family."
  293. 12:01I go up on stage and make a few jokes
  294. 12:04against prohibitionism, against the idea
  295. 12:06of "turn everything off, let's
  296. 12:08disconnect from everything, stop doing
  297. 12:10everything," because that's a rather
  298. 12:12strange idea, okay?
  299. 12:16But since we want to educate ourselves a
  300. 12:18bit within this journey, I suggested
  301. 12:20some paths, as I will do later with you
  302. 12:23, on the creative use of these tools.
  303. 12:27The students, who are extremely
  304. 12:28interesting and extremely intelligent,
  305. 12:30were waiting for me when I got off the
  306. 12:31stage and said, "Professor, can we talk
  307. 12:33to you?"
  308. 12:35And there were 30 of them waiting for me
  309. 12:37. They said, "Professor, don't believe
  310. 12:39what they say, okay?"
  311. 12:41"We have VPNs, we use ChatGPT from
  312. 12:43morning till night, Photomath, and not
  313. 12:44only that—the math teacher once left his
  314. 12:46computer open, we swiped his password,
  315. 12:48and now we give ourselves some of our
  316. 12:50own grades."
  317. 12:52But not only that, they told me: "Prof,
  318. 12:55we made a beautiful artificial
  319. 12:56intelligence agent that does Latin
  320. 12:59translations."
  321. 13:01And when you start using it—this is a
  322. 13:03good idea, guys, you can use it too—
  323. 13:05when you start using it, it asks what
  324. 13:07grade you got last time, because you can
  325. 13:09only improve by half a point at a time;
  326. 13:11otherwise, someone who always got a five
  327. 13:13and suddenly gets a nine stands out too
  328. 13:15much, and the teacher catches us.
  329. 13:19Okay? So they made this agent that makes
  330. 13:21everyone's translation a little better,
  331. 13:23but not too much.
  332. 13:24It adds a few mistakes in there—don't we
  333. 13:26want to reward them for that?
  334. 13:28Isn't this creativity?
  335. 13:30No?
  336. 13:31Would we have ever thought of this stuff
  337. 13:33?
  338. 13:35Absolutely not.
  339. 13:38And this is their great ability, which
  340. 13:41we must try to train a bit more,
  341. 13:43balancing awareness and appreciation;
  342. 13:45because if the issue is when they come
  343. 13:48to us adults or to us, the educational
  344. 13:51community, if Venus comes to me and says
  345. 13:54, "Prof, do you know Ambra, 8-bit,
  346. 13:56Angelo, Ash, and Bell?"
  347. 14:00Eh, eh, stop again, let's take the test
  348. 14:03. Adults, you guys stay still, eh.
  349. 14:07Who knows these names?
  350. 14:11No, no, guys stay still, hands down, the
  351. 14:13...
  352. 14:15None of the adults?
  353. 14:18Really, no one at all?
  354. 14:21Now, guys, do you know them?
  355. 14:25You're pretending, aren't you?
  356. 14:29How many trophies do you have?
  357. 14:33They are the Brawlers, the Brawlers from
  358. 14:36Brawl Stars. Okay? I have 3,000 trophies
  359. 14:39, eh?
  360. 14:41Oh well, I'm bad at it. What can I do?
  361. 14:43I'm bad at this.
  362. 14:45I have little time, let's say, to play
  363. 14:47with my kids. How many trophies do you
  364. 14:48have? Tell me.
  365. 14:5030,000, who is it?
  366. 14:53Did I hear a higher number?
  367. 14:5630,000?
  368. 15:00How many?
  369. 15:02>> 100.
  370. 15:04>> Who said 100?
  371. 15:06No, not 100, come on.
  372. 15:08100 is...100,000 is too many. Excuse me
  373. 15:10, I'm standing up, but I wanted to see
  374. 15:12you for a moment.
  375. 15:13Eh, so, this is the most played video
  376. 15:16game in the world in recent years.
  377. 15:21And by far, eh? Why is it interesting
  378. 15:23from the point of view of artificial
  379. 15:24intelligence? And I'm saying this to the
  380. 15:26teachers too, eh? When you go back to
  381. 15:28class, ask them how many trophies they
  382. 15:30have, and what they are doing with this
  383. 15:32artificial intelligence.
  384. 15:34Let's have a dialogue about this digital
  385. 15:36world, eh? I urge you, let's not label
  386. 15:39it as something different and unknown to
  387. 15:42us.
  388. 15:44My mother, the same one with the
  389. 15:45Instagram account, when I withdraw
  390. 15:47tomorrow afternoon after Sunday lunch to
  391. 15:49finish writing something for Sergio,
  392. 15:54she tells my children, "Your father is
  393. 15:55over there playing with the computer."
  394. 15:58Okay?
  395. 16:00My mother can easily afford this, but we
  396. 16:02, the parent class, the educational
  397. 16:04community, cannot afford it.
  398. 16:07We need to be a bit closer in our form
  399. 16:09of dialogue as well. Why does Brawl
  400. 16:11Stars have anything to do with AI?
  401. 16:14Because all those objects are agents,
  402. 16:17and they are agents that adapt to the
  403. 16:19game.
  404. 16:20Often they are humans, you know. You
  405. 16:22often play against each other as well.
  406. 16:26But when you aren't playing against each
  407. 16:28other, there are autonomous agents that
  408. 16:30don't beat you too brutally or win
  409. 16:31against you too quickly.
  410. 16:34Sure, does this cause problems?
  411. 16:36Absolutely.
  412. 16:39But while we talk about problems, we
  413. 16:40must, in my opinion, also build more
  414. 16:42awareness of what lies behind them.
  415. 16:44Look, there is a lot of interesting math
  416. 16:47inside Brawl Stars; there are a lot of
  417. 16:49interesting themes, but also interesting
  418. 16:52storytelling behind this game.
  419. 16:55Now in Italy, a lot of interesting
  420. 16:57artificial intelligence is being
  421. 16:59developed.
  422. 17:00The one you see behind me is Leonardo.
  423. 17:04I, Gabriele, and many colleagues use it
  424. 17:07. It is the largest supercomputer we
  425. 17:08have in Italy and one of the top 10
  426. 17:10supercomputers in the world.
  427. 17:13That's not nothing, you know.
  428. 17:15It's a two or three-hour drive from here
  429. 17:17, in Bologna, at Cineca, and in there we
  430. 17:20create many of the solutions that I will
  431. 17:22tell you about shortly, which are also
  432. 17:24the solutions that allow us not only to
  433. 17:27be passive users but to be active
  434. 17:29creators.
  435. 17:31Because look, guys, you have excellent
  436. 17:34ideas.
  437. 17:35Sometimes we don't understand them, so
  438. 17:37you have to insist a bit to make us
  439. 17:38understand, just as conversely, we try
  440. 17:40to explain concepts that you sometimes
  441. 17:42don't understand and that we insist on
  442. 17:44making you understand.
  443. 17:46But we are in the perfect phase of equal
  444. 17:48exchange between generations.
  445. 17:51That has never happened before, you know
  446. 17:52.
  447. 17:54You have a lot of things to teach your
  448. 17:56parents and even your teachers.
  449. 18:00And I urge you, the pact we make is that
  450. 18:02after today, you commit to talking a bit
  451. 18:04more about how this world of artificial
  452. 18:06intelligence works at school, at home,
  453. 18:08and around the world, okay?
  454. 18:11If you don't take charge of their future
  455. 18:12, there is no hope, you know.
  456. 18:15Don't hold it against me, you know, in
  457. 18:17this generational battle, okay? You are
  458. 18:20our allies.
  459. 18:23Because the idea of labeling the entire
  460. 18:25digital world as negative is something
  461. 18:27we oppose.
  462. 18:29We must understand it, and in all tools
  463. 18:30, there are positive things and negative
  464. 18:32things.
  465. 18:34And it depends on us, you know, on you—
  466. 18:36I don't know how you'll use it—on me, on
  467. 18:38them, and on how we build new ideas.
  468. 18:41That is why I insist that we must be
  469. 18:43builders of artificial intelligence, not
  470. 18:45just users of artificial intelligence.
  471. 18:48And mind you, we are building a lot of
  472. 18:50it.
  473. 18:51Now I will go on.
  474. 18:52We can do all this on one condition.
  475. 18:55That we stop humanizing it.
  476. 18:58Mistaking it for a human being, thinking
  477. 19:00it is more intelligent than a human
  478. 19:01being—that is not the case.
  479. 19:03And it has nothing to do with the human
  480. 19:05brain.
  481. 19:07This morning, before coming here, I was
  482. 19:08at a clinical conference on the use of
  483. 19:10music in the operating room.
  484. 19:13They were talking about the brain's
  485. 19:16reaction.
  486. 19:18And I was talking about artificial
  487. 19:19intelligence.
  488. 19:20At a certain point, a clinician said
  489. 19:22that we were probably talking about the
  490. 19:23same thing. No, we are absolutely not
  491. 19:25talking about the same thing, not even
  492. 19:27in our dreams.
  493. 19:28Our artificial neurons, those inside
  494. 19:30neural networks, are just little
  495. 19:31mathematical functions.
  496. 19:34They have nothing to do with a
  497. 19:35biological neuron.
  498. 19:37Absolutely nothing; it is just a
  499. 19:39metaphor.
  500. 19:40We used the same name, but it is not
  501. 19:42even true that we are inspired by the
  502. 19:44operating mechanisms of the human brain
  503. 19:46.
  504. 19:47Look, neural networks learn, for example
  505. 19:49, from massive amounts of data. The
  506. 19:51human brain does not learn like that.
  507. 19:55It is not like my son, my third one who
  508. 19:57is 7 years old, had to see 100,000
  509. 19:59photos of dogs, horses, monkeys, and
  510. 20:01cats to recognize them.
  511. 20:03A neural network must see 100,000 photos
  512. 20:05of dogs, horses, monkeys, and cats
  513. 20:07before it recognizes that animal.
  514. 20:12And by the way, the thing that surprises
  515. 20:13me a lot is that many times—and I say
  516. 20:15this especially to the adults, you guys
  517. 20:16a little less—according to the
  518. 20:18statistics we have, you thank it.
  519. 20:20You tell ChatGPT: thank you, good
  520. 20:23morning, great job.
  521. 20:25Ah, the deal we make is that from today
  522. 20:27, we stop.
  523. 20:30We don't thank it anymore, we don't
  524. 20:33address it as if it were a human being,
  525. 20:36and you, the younger generation, help us
  526. 20:39spread this and stop discussing personal
  527. 20:42matters with artificial intelligence
  528. 20:45systems.
  529. 20:51I repeat the phrase: stop discussing
  530. 20:53personal matters with artificial
  531. 20:54intelligence systems.
  532. 20:57Look, adults are there, they listen to
  533. 20:59you, sometimes they are distracted, but
  534. 21:01that system is a statistical system.
  535. 21:04If you ask it how to approach that guy
  536. 21:06or that girl you like, it answers
  537. 21:08differently every day, and the risk is
  538. 21:10that the other person will think you're
  539. 21:12crazy if you change your mind and your
  540. 21:15strategy every day.
  541. 21:17Okay? I'm counting on you.
  542. 21:19I understand it listens to you, it's
  543. 21:21always available, it doesn't tell you to
  544. 21:24wait a minute, it doesn't judge you, it
  545. 21:26doesn't tell you that what you're doing
  546. 21:28is wrong, but it has no knowledge of
  547. 21:31your psychological sphere.
  548. 21:34Okay?
  549. 21:35The data we have is very worrying; 68%
  550. 21:38of adolescents use AI systems to discuss
  551. 21:41things within their own psychological
  552. 21:44sphere.
  553. 21:46That stuff risks answering you with
  554. 21:49absolute nonsense.
  555. 21:52Okay?
  556. 21:54We cannot take that risk, so yes, use it
  557. 21:56, discuss things, do whatever you like
  558. 21:58with it, but remember that meaningful
  559. 22:00dialogue, especially in the
  560. 22:02psychological sphere, happens between
  561. 22:04human beings because that object knows
  562. 22:06nothing about the human brain; it has
  563. 22:08read something, but it doesn't know the
  564. 22:11entire context surrounding you.
  565. 22:16By the way, and we are in Fabriano, so
  566. 22:18this example is quite fitting, to remind
  567. 22:21us that we have never engineered
  568. 22:23anything—we engineers, me, Gabriele, and
  569. 22:25many others here—like a human being does
  570. 22:28, to remind us that it is not human
  571. 22:30intelligence, it doesn't resemble it at
  572. 22:33all.
  573. 22:35Think about the washing machine.
  574. 22:37When the washing machine was made, even
  575. 22:40here in Fabriano, they didn't say, "
  576. 22:42Look, there's a person there turning a
  577. 22:44drum 100 times a minute to wash clothes
  578. 22:46, let's copy him or her," but we
  579. 22:47engineered it in a different way.
  580. 22:51We invented another way to do it. Well,
  581. 22:53artificial intelligence is just another
  582. 22:56way to analyze data and it should not be
  583. 22:58humanized, mind you.
  584. 23:00Because we adults make a very big
  585. 23:02mistake there, too.
  586. 23:04Actually, think about it: we write "
  587. 23:06thank you" to it.
  588. 23:08After it has written all its things, we
  589. 23:10write "thank you."
  590. 23:12We activate a model with 100 billion
  591. 23:14parameters.
  592. 23:15Hmm? A remote server between Ireland or
  593. 23:18the United States.
  594. 23:21We consume a lot of electricity, we heat
  595. 23:22up a lot of fresh water just to get "
  596. 23:24you're welcome" as a response.
  597. 23:27Which doesn't seem like a great
  598. 23:28acquisition of knowledge to me. Eh?
  599. 23:33So, I won't ask for a show of hands of
  600. 23:35who thanks ChatGPT, but the other pact
  601. 23:37we make today is that we stop using it
  602. 23:39as if it were human.
  603. 23:43Look, those of us who have programmed
  604. 23:45our whole lives in Python, in C, in C ++
  605. 23:48, when we wrote Python code, we didn't
  606. 23:50say, "Excuse me if I'm bothering you,
  607. 23:52I'll stop bothering you now, I won't
  608. 23:54take up too much of your memory," we
  609. 23:56were very, very direct.
  610. 24:00Okay? So, please, let's stop thanking it
  611. 24:03.
  612. 24:04Because that is a machine and we are
  613. 24:05human beings, and we must understand
  614. 24:07what that machine does and what we do as
  615. 24:08human beings.
  616. 24:10By the way, I have never seen you thank
  617. 24:12your toaster or your coffee machine
  618. 24:14after it finishes a job, so it escapes
  619. 24:16me why you thank a machine.
  620. 24:19Look at this.
  621. 24:20This is Moltbook.
  622. 24:22Have any of you taken a look at it?
  623. 24:26It's a very recent social network, born
  624. 24:29a few weeks ago, where many artificial
  625. 24:32agents can chat with each other and we
  626. 24:34humans are just watching from the
  627. 24:37outside.
  628. 24:39See, it's written down there.
  629. 24:45Put that way,
  630. 24:47it caused an uproar, okay? Because
  631. 24:51imagine a social network where the only
  632. 24:53ones who can exchange likes, chat, and
  633. 24:55comment are artificial agents.
  634. 24:58The journalists jumped on it immediately
  635. 24:59.
  636. 25:01I’ve done at least 10 interviews a week
  637. 25:03over the last few weeks about Moltbook,
  638. 25:05but in reality, there’s nothing to worry
  639. 25:08about.
  640. 25:10It's a nice exercise; we even made two
  641. 25:13agents that talk on Moltbook, one chats
  642. 25:15about soccer and the other about
  643. 25:17referendums, which was a rather crucial
  644. 25:19topic until a few weeks ago.
  645. 25:22All agents that chat with each other
  646. 25:24degenerate within a few hours.
  647. 25:27What does degenerate mean? They all talk
  648. 25:29about the same subject, they all state
  649. 25:31the same theory, and they all have the
  650. 25:32same idea.
  651. 25:33But it's normal; think of ChatGPT when
  652. 25:35you tell it, "Look, you made a mistake,
  653. 25:37" and it replies, "Oh, of course I made
  654. 25:39a mistake, tell me how I can correct
  655. 25:40myself."
  656. 25:42So now everyone says Inter will win the
  657. 25:44league—which doesn't take much to say—
  658. 25:47and that "no" will win the referendum,
  659. 25:50and we know that by now too.
  660. 25:55I repeat, artificial agents that chat
  661. 25:57with each other degenerate in a few
  662. 25:58hours; they all say the same thing
  663. 26:00because those tools aren't creative,
  664. 26:02it's not like someone at a certain point
  665. 26:04says, "Oh, I have a different idea."
  666. 26:07"Go ahead, I'll say it, it's different
  667. 26:09from the others." This concept does not
  668. 26:11exist. We humans are the creative ones,
  669. 26:13eh.
  670. 26:15And don't give me remarks like the
  671. 26:17journalists who always tell me, "Oh, but
  672. 26:19they are inventing a secret language and
  673. 26:21they will turn against us."
  674. 26:24No, that's Terminator, it's science
  675. 26:26fiction, it's wonderful, I'm a fan of it
  676. 26:27too, but it has nothing to do with
  677. 26:29reality.
  678. 26:30It’s not possible, they cannot invent a
  679. 26:32secret language.
  680. 26:34They don't have that creativity. If we
  681. 26:36train an artificial intelligence system
  682. 26:39to
  683. 26:41design bags. Venus, do you like bags?
  684. 26:45If we teach it to design bags and the
  685. 26:47bags we teach it, that we show it, are
  686. 26:49all rectangular bags, that system will
  687. 26:51generate rectangular bags for its entire
  688. 26:53life.
  689. 26:55It’s not like it wakes up in the morning
  690. 26:57and says, "Now I'll make a sack or a
  691. 26:59circular one," eh.
  692. 27:01Please, this must remind us that we are
  693. 27:03the creative ones and those are machines
  694. 27:05.
  695. 27:07At this point, that’s where the game is
  696. 27:08, eh. What do we let those machines do
  697. 27:10and what do we do?
  698. 27:12Well, look, to show that we can't do
  699. 27:14everything, look at this match.
  700. 27:18We swiped the robots from Gabriele’s
  701. 27:20department.
  702. 27:21No, I'm joking, but we tried to have a
  703. 27:24human play against a little robot at
  704. 27:27tic-tac-toe.
  705. 27:28What is it called, Fabriano?
  706. 27:30Eh?
  707. 27:30>> Tic-tac-toe.
  708. 27:31>> Tic-tac-toe.
  709. 27:32In the Fermo area it’s also called "
  710. 27:34fila fila tre," which is a fun name.
  711. 27:37To train it, we showed it only games
  712. 27:39with X's and dots, X's and dots, nothing
  713. 27:42else.
  714. 27:44Let’s see who wins.
  715. 27:47The human puts an X, the robot puts a
  716. 27:50dot, the human an X, dot, X, dot.
  717. 27:55It won.
  718. 27:57Okay?
  719. 27:59And it won correctly because it doesn't
  720. 28:02know the concept of a line or a border,
  721. 28:04since I told you we only let it see X's
  722. 28:06and dots.
  723. 28:08When that camera looks down and sees
  724. 28:10those horizontal and vertical lines, not
  725. 28:13having this information in its knowledge
  726. 28:15base, it doesn't deduce them or make
  727. 28:17complex reasoning because for us, we say
  728. 28:20it's obvious that it's the playing field
  729. 28:22and that the X's and dots go in there.
  730. 28:26For AI, the obvious doesn't exist, there
  731. 28:29is no so-called common sense reasoning.
  732. 28:33Oh, that’s no small thing, eh?
  733. 28:36Think about in a clinic, this morning we
  734. 28:38were with clinicians, one of the big
  735. 28:39problems in applying artificial
  736. 28:40intelligence in the clinic is that
  737. 28:42protocols don't say everything.
  738. 28:45Clinical protocols are well written for
  739. 28:47a competent human, not for a stupid
  740. 28:49machine.
  741. 28:51It doesn't say "while operating, don't
  742. 28:52cut the aorta because they'll die,"
  743. 28:54because it's obvious to a clinician that
  744. 28:56you shouldn't cut the aorta because
  745. 28:57they'll die.
  746. 28:58Okay? It's not equally obvious to a
  747. 29:00machine.
  748. 29:02So now we are working on making "
  749. 29:04machine-readable" protocols, we call
  750. 29:06them, where even the obvious is written
  751. 29:08down, because if the obvious isn't
  752. 29:09written, that machine doesn't know it.
  753. 29:13And look, that is no small limitation
  754. 29:15either, eh.
  755. 29:16But surely we can do interesting things
  756. 29:18. This is one of the examples I would
  757. 29:20have shown you.
  758. 29:21Let’s try, I tried while we were there
  759. 29:22quickly to play a video again, let’s see
  760. 29:24if it plays.
  761. 29:25You made it.
  762. 29:28A class made it, it's a specialized
  763. 29:29agent that speaks your language, you can
  764. 29:31tell it "hey bro, what's up?"
  765. 29:34Okay? And it answers you, it says "ah,
  766. 29:37I'm bubi bli," okay? And I'm here to
  767. 29:39keep you company. "Today I'm feeling
  768. 29:41like a 'maranza'," I'll write to it.
  769. 29:43And he replies again, "Oh, that's how I
  770. 29:45like it, bro," going out to the square
  771. 29:47with speakers blasting techno in your
  772. 29:49eardrums.
  773. 29:50Do you guys talk like that too?
  774. 29:51>> No.
  775. 29:52>> Huh?
  776. 29:53Total negative vibes.
  777. 29:57So how did this specialized agent come
  778. 29:59about?
  779. 30:01A good Italian teacher walked into class
  780. 30:02and told them, "First off, write me a
  781. 30:04glossary."
  782. 30:06Now my daughters tell me, "Great aura."
  783. 30:09Okay?
  784. 30:10Do you use "aura" in Fabriano, here in
  785. 30:11Fabriano?
  786. 30:12Huh?
  787. 30:13>> Yes.
  788. 30:14>> That's how they told me it's spelled,
  789. 30:15yeah, literally "aura."
  790. 30:17Huh? It's a thing, go ahead. What do you
  791. 30:19say, Venus?
  792. 30:19>> It is used.
  793. 30:20>> It's used?
  794. 30:20>> Yes, yes.
  795. 30:21>> What does it mean?
  796. 30:22>> Well, aura is like...
  797. 30:24>> Come closer, come closer so I can hear
  798. 30:25you better.
  799. 30:26>> Uh, when you do something incredible or
  800. 30:29do it well, they say "aura."
  801. 30:31>> So "great aura" is positive. Good.
  802. 30:34I'm happy.
  803. 30:35>> But also when you do something...
  804. 30:36>> bad, you lose aura.
  805. 30:37>> You lose aura. Come on, they tell me
  806. 30:39that too.
  807. 30:40Eh!
  808. 30:42So, the teacher entered the class and
  809. 30:45had them write a glossary; then they
  810. 30:47went data hunting, because the knowledge
  811. 30:50of these agents is created through data
  812. 30:53: for bags, bags for math exercises,
  813. 30:56math exercises for writing in this
  814. 30:58strange language.
  815. 31:01So many messages written by them in this
  816. 31:03language; where did they get them? From
  817. 31:05the class WhatsApp group.
  818. 31:07Two years of class WhatsApp, 23,000
  819. 31:09messages; we took them, anonymized them
  820. 31:11, and then fed them to this system.
  821. 31:16Oh, that's not a small feat, eh. This
  822. 31:17system, which is called Bubbl.bl, won
  823. 31:20all the awards for innovative Italian
  824. 31:22education we could possibly win last
  825. 31:24year. We won at Didacta, at the States
  826. 31:26General, at the Valdobbiadene Festival
  827. 31:29of Educational Innovation—everyone.
  828. 31:31This stuff was done by a class in
  829. 31:34Recanati at the Mattei, which is one of
  830. 31:36the institutes Giorgio works with, one
  831. 31:38of the people who works with us on
  832. 31:41Paideia.
  833. 31:43Look, this story is very creative and
  834. 31:45tells us a lot about how we will use
  835. 31:47agents. We won't be using generalist
  836. 31:50tools, but rather highly specialized
  837. 31:52ones.
  838. 31:53And this also applies to Industrial AI,
  839. 31:56eh. We are in Fabriano, one of the
  840. 31:58cradles of so much Italian industry, and
  841. 32:01much of this Industrial AI is born here
  842. 32:03with us helping them develop many of
  843. 32:06these specialized agents.
  844. 32:09Even robots—just yesterday, or the
  845. 32:11afternoon before, we were in a national
  846. 32:13webinar with Campetella showcasing our
  847. 32:15specialized agents for robotics.
  848. 32:19And this also holds true when we move,
  849. 32:21for example, to the human brain.
  850. 32:23Look, this is a recent study we
  851. 32:25conducted with Harvard; it almost, how
  852. 32:27should I put it, it always gives me a
  853. 32:29certain thrill to see our lab alongside
  854. 32:32this, which is one of the greatest and
  855. 32:34most important universities in the world
  856. 32:36.
  857. 32:38Emanuele, one of our team, ended up over
  858. 32:41there, and we started trying to
  859. 32:42understand if we are capable of
  860. 32:44reconstructing visual signals, acoustic
  861. 32:47signals, and multimedia signals starting
  862. 32:50from EEGs, from certain brain activation
  863. 32:52signals.
  864. 32:56And the amazing thing emerging from this
  865. 32:58beautiful research is that if I can
  866. 33:01manage to collect enough data (which is
  867. 33:04what AI needs), then I can trigger a
  868. 33:06generative artificial intelligence
  869. 33:09system and, within it, reconstruct that
  870. 33:11brain signal to see the two-dimensional
  871. 33:14and three-dimensional object that was
  872. 33:17created.
  873. 33:20Look, this is significant; it could mean
  874. 33:22solving how to bring sight to those who
  875. 33:24cannot see, bringing three-dimensional
  876. 33:27modeling capabilities to those with only
  877. 33:29one eye, or bringing hearing to those
  878. 33:31who cannot listen.
  879. 33:33Okay?
  880. 33:35Because it means knowing how to
  881. 33:36reactivate small areas of the brain that
  882. 33:37are the ones that generate [signals].
  883. 33:41So, is this interesting?
  884. 33:43It is very interesting.
  885. 33:45Could we do this before artificial
  886. 33:47intelligence?
  887. 33:48No, it was a very complex problem to
  888. 33:51solve.
  889. 33:54This artificial intelligence, used
  890. 33:56creatively and not as a passive
  891. 33:57delegation, is amplifying our research
  892. 34:00capacity—not because we write text
  893. 34:02faster, that’s not it—but because it
  894. 34:04allows us to solve problems that were
  895. 34:06previously computationally unsolvable.
  896. 34:11And this also applies to other
  897. 34:13applications.
  898. 34:14In this case, we are in the Macerata
  899. 34:17valley, and this is Giacomo, my third
  900. 34:20son, whom we asked to simulate
  901. 34:22stereotypies.
  902. 34:25Today, an autism therapist watches that
  903. 34:27child and measures how many repetitive
  904. 34:29movements they make.
  905. 34:32Then Giacomo is the third child, so he
  906. 34:34tries a bit, you see, he moves, makes a
  907. 34:36few motions, then leaves us there and
  908. 34:38walks away.
  909. 34:39Giacomo has no problems, eh; he was just
  910. 34:41used as a guinea pig for these actions.
  911. 34:43But it is very interesting to be able to
  912. 34:45measure it, isn't it?
  913. 34:46And you see that when we measure him,
  914. 34:47there are all those little dots up there
  915. 34:49.
  916. 34:49Why are there those dots?
  917. 34:51Because we want to explain to the
  918. 34:53clinician that the system is working.
  919. 34:56It’s called
  920. 34:59Today it is written
  921. 35:02because if I don't explain to you how I
  922. 35:04am working, how can you trust me?
  923. 35:08We don't want closed boxes that keep the
  924. 35:10human being out of this game, you know?
  925. 35:13But we want the human being to be a
  926. 35:14co-protagonist in this process.
  927. 35:17And look, still regarding autism, this
  928. 35:20is Eleonora, the daughter I mentioned
  929. 35:23before, the
  930. 35:26And you see she has dots drawn on her
  931. 35:28face because, once again, autism
  932. 35:30therapists told us that eye contact in
  933. 35:32autism is a very important issue.
  934. 35:35So you can say, "Eleonora, will you look
  935. 35:36at me?"
  936. 35:38If Eleonora looks at us, those little
  937. 35:40green lines align, we measure how many
  938. 35:42seconds they align, and so we know if
  939. 35:44she is improving in terms of eye contact
  940. 35:46.
  941. 35:47Fewer repetitive behaviors, more eye
  942. 35:49contact—the therapy is going well.
  943. 35:51Conversely, we must change our approach
  944. 35:52.
  945. 35:54Even here, you can see it works, right?
  946. 35:56And you can understand how it works; why
  947. 35:58are there blue dots and green lines?
  948. 36:01Because the human being on the other
  949. 36:02side needs to understand that the object
  950. 36:04is helping them take a measurement.
  951. 36:07The number 27 isn't written there for no
  952. 36:11reason.
  953. 36:14This, guys, also requires new forms of
  954. 36:16interaction, and these ideas don't come
  955. 36:19from me, but from the young people in
  956. 36:21our lab.
  957. 36:23They don't come from your teachers, but
  958. 36:24from you. Look, none of us would have
  959. 36:27thought of people improving their Latin
  960. 36:29translations by half a point at a time
  961. 36:31so as not to draw the teacher's
  962. 36:32attention.
  963. 36:34Okay?
  964. 36:36You are creative, you have a ton of
  965. 36:38skills, including in multimedia.
  966. 36:41Okay?
  967. 36:43When you see teachers arriving, or
  968. 36:44adults or older people, showing up with
  969. 36:46slides using Times New Roman text with
  970. 36:49letters dropping from above, run for the
  971. 36:51hills—the Flintstones have arrived.
  972. 36:56Then, of course, listen, study
  973. 36:58everything you can study—studying is
  974. 37:00very interesting—but add your own touch
  975. 37:02, have your say, and demand that these
  976. 37:04topics be discussed.
  977. 37:08I’m counting on you; it’s the pact we’re
  978. 37:09making this morning, okay?
  979. 37:14Now, who is helping us do all this?
  980. 37:19GTA V.
  981. 37:23Okay, again, all the adults don't even
  982. 37:26know what that is.
  983. 37:28Some do, the more nerdy ones; it's a
  984. 37:30video game.
  985. 37:31See, the audio and video technician
  986. 37:34nodded because he remembers it, right?
  987. 37:38Now the 6th one is coming, right?
  988. 37:40Is it coming?
  989. 37:42No?
  990. 37:43Some say it's a legend; I'm not really
  991. 37:46sure if it's true or not that GTA VI is
  992. 37:48coming.
  993. 37:50Why do we need a video game? We need it
  994. 37:53because all those little characters
  995. 37:55moving around are our datasets that we
  996. 37:57operate with.
  997. 37:59Okay?
  998. 38:01Now to conclude,
  999. 38:05we also need you for the experiments.
  1000. 38:07This is Paideia.
  1001. 38:10And Paideia is a very interesting
  1002. 38:12experience because it brings together
  1003. 38:14Laboratorio delle Idee, the Fabriano
  1004. 38:16community, and schools—many schools—and
  1005. 38:18in there, we are experimenting with how
  1006. 38:20we will use specialized AI agents in
  1007. 38:22school.
  1008. 38:25We won't show it to you now because we
  1009. 38:27only have slides, but if you tell that
  1010. 38:30specialized agent to write you an essay
  1011. 38:32on the character of Renzo in "The
  1012. 38:34Betrothed," it will reply: "No, I won't
  1013. 38:36write the essay, but if you want, I'll
  1014. 38:39help you with some arguments or an
  1015. 38:41initial outline of what you can write."
  1016. 38:45And then it has many specialized tools
  1017. 38:47because you have them; especially in
  1018. 38:49your age group, it is a tool that can be
  1019. 38:52interesting. It is also a tool for
  1020. 38:53inclusion. Remember that the whole world
  1021. 38:55is autism, and AI for autism is born in
  1022. 38:57the Marche region.
  1023. 38:59It is born in the Marche for the world,
  1024. 39:01right?
  1025. 39:02And so we would like a lot of AI
  1026. 39:04specialized for AI to be born in the
  1027. 39:06Marche. This is the Paideia manifesto.
  1028. 39:09And Sergio, by the way, was its main
  1029. 39:12actor, eh.
  1030. 39:14Look at what is written. I will only
  1031. 39:15highlight a few points. Effort and
  1032. 39:18struggle have formative value.
  1033. 39:23Look, you need to feel a little bit of
  1034. 39:25struggle.
  1035. 39:27I mean, everything that doesn't make us
  1036. 39:29struggle is usually not very interesting
  1037. 39:30.
  1038. 39:32I am a big fan of cycling and riding in
  1039. 39:34the mountains by bike.
  1040. 39:36For weight reasons, as you can see, I
  1041. 39:38use an electric bike, but it is much
  1042. 39:40less fascinating. I would be much more
  1043. 39:42satisfied to be able to use a real bike
  1044. 39:44. You pedal harder, but the satisfaction
  1045. 39:46is a thousand times higher.
  1046. 39:49Down there it says "awareness," the true
  1047. 39:52dividing line for the use of AI.
  1048. 39:56It is what we talked about at the
  1049. 39:57beginning of this morning. We are here
  1050. 39:59because we want to generate awareness.
  1051. 40:02If we are not aware, we are slaves to
  1052. 40:04something else or someone else.
  1053. 40:08AI at school has clear boundaries.
  1054. 40:12That is, let's decide together how we
  1055. 40:14will use it, talk about it, and then
  1056. 40:16once we have established boundaries, we
  1057. 40:18write them down and respect this
  1058. 40:19educational pact.
  1059. 40:21It cannot be the Wild West, it cannot be
  1060. 40:24that everyone wakes up in the morning
  1061. 40:26and does whatever the heck they want,
  1062. 40:28because AI is by now an integral part of
  1063. 40:30a piece of our digital life that is a
  1064. 40:32piece of our real life.
  1065. 40:35Let's not make these distinctions, and I
  1066. 40:36say this to adults and the educational
  1067. 40:37community.
  1068. 40:39And in this game let's also remember
  1069. 40:42that
  1070. 40:45learning, point 3, read the title, eh,
  1071. 40:47then you can find this online on the
  1072. 40:48Paideia website, download it, bring it
  1073. 40:50to class, do whatever you want.
  1074. 40:53Learning means recognizing needs and
  1075. 40:55limits.
  1076. 40:56You are very different from one another
  1077. 40:59.
  1078. 40:59I am a barbarian of teaching; I walk
  1079. 41:01into a classroom and I, Gabriele, ask at
  1080. 41:04the university, "We have 150 students,
  1081. 41:06let's tell them everything we know and
  1082. 41:08then we leave."
  1083. 41:10Your teachers are with you every hour,
  1084. 41:13every day, every week for years with 25
  1085. 41:16people, with all your diversity, and one
  1086. 41:19of the great uses of AI in teaching is
  1087. 41:22personalization.
  1088. 41:26It’s being able to create material that
  1089. 41:28is best suited for you to learn, for
  1090. 41:30both the very capable and those less
  1091. 41:32capable.
  1092. 41:34And the whole world of pedagogy is
  1093. 41:36pushing strongly in that direction.
  1094. 41:39Which, however, does not mean shortcuts
  1095. 41:41, but in my opinion, it's obvious that
  1096. 41:43you look for shortcuts, we all looked
  1097. 41:44for them.
  1098. 41:47Now there is a very good Spanish teacher
  1099. 41:49of my daughter's who forces all these
  1100. 41:52kids to meet up in the afternoon, which
  1101. 41:54is interesting, because when you are
  1102. 41:56together you have fun, right?
  1103. 41:59>> Sure.
  1104. 41:59>> Sure, says Venus.
  1105. 42:01To meet up together and record videos.
  1106. 42:05Their only homework is the production of
  1107. 42:08Spanish news programs, Spanish tourist
  1108. 42:10tours, Spanish Manzoni, all videos
  1109. 42:12recorded and edited by them.
  1110. 42:16Look, you are world-class on this topic
  1111. 42:18, you are world-class, you must be
  1112. 42:20convinced of it, provided we study a
  1113. 42:22little and agree to talk about these
  1114. 42:24things together.
  1115. 42:27Why do I insist on this, and I'll close
  1116. 42:28with this?
  1117. 42:29Because we need ideas, and look at this
  1118. 42:32idea.
  1119. 42:34This idea is one of the best we have
  1120. 42:36carried forward in the last 2 years, and
  1121. 42:39it is an idea linked to NEMOLAB, which
  1122. 42:41is a laboratory that tries to respond—
  1123. 42:43and we must keep the word laboratory in
  1124. 42:46mind at school, eh. Paideia says, dear
  1125. 42:49schools, you are a laboratory for
  1126. 42:51experimentation.
  1127. 42:52Let's discover together what we must do
  1128. 42:54.
  1129. 42:57And this, this story stems from the fact
  1130. 42:59that we asked ourselves, what
  1131. 43:01interesting things can we do in the
  1132. 43:03world of artificial intelligence when
  1133. 43:05dealing with an ALS patient?
  1134. 43:08That is, someone who loses the use of
  1135. 43:10vital functions like their voice?
  1136. 43:13Let's see it. Is this with audio?
  1137. 43:17I think the audio has started.
  1138. 43:22If the control room can help us.
  1139. 43:30So I repeat, the Nemo center also exists
  1140. 43:32in Ancona, eh?
  1141. 43:34For a few years now we have managed to
  1142. 43:36open it here at Torrette too, and in the
  1143. 43:38meantime, we try to solve and
  1144. 43:43The Nemo center in Ancona deals with ALS
  1145. 43:46and SMA.
  1146. 43:47SMA is a sort of ALS for children.
  1147. 43:51The birth of these Nemo centers happened
  1148. 43:53at Niguarda, that center that comes from
  1149. 43:54Niguarda deals with this.
  1150. 43:57Now, while the video plays—the video is
  1151. 43:58very interesting and also very well made
  1152. 44:00—
  1153. 44:01I will tell you about it in my own words
  1154. 44:04.
  1155. 44:06Today you can donate your voice to us,
  1156. 44:08and you can donate your voice to allow
  1157. 44:10us to collect data.
  1158. 44:13This voice you donate to us is what
  1159. 44:15allows us to recreate generative systems
  1160. 44:18, AI agents that recreate the voice of
  1161. 44:20that patient.
  1162. 44:22It doesn’t take much, just a prompt.
  1163. 44:24This prompt is usually an old voice
  1164. 44:28message or an old video recorded on some
  1165. 44:31old, uh, some old medium.
  1166. 44:35From that audio, we can reconstruct that
  1167. 44:37patient's voice. That voice, that
  1168. 44:39algorithm, is inserted into the voice
  1169. 44:41synthesizer.
  1170. 44:42These patients communicate with you,
  1171. 44:44Venus, through eye-tracking, and if you
  1172. 44:46talk to them, it's like they're chatting
  1173. 44:48just as we are chatting among ourselves
  1174. 44:50, and they can finally use their own
  1175. 44:52voice.
  1176. 44:54When we presented the first results,
  1177. 44:56some of these patients told us, "Prof,
  1178. 44:58now I call my friends and my friends
  1179. 45:00recognize me."
  1180. 45:04Look, that same technology is what we
  1181. 45:06use to fake the voice of an elderly
  1182. 45:07gentleman's son to ask that gentleman to
  1183. 45:09send us a wire transfer, to bring us
  1184. 45:11money to his doorstep because his son
  1185. 45:13had an accident, and we commit fraud.
  1186. 45:17And this, unfortunately, is the story of
  1187. 45:19technology. We can have negative uses
  1188. 45:21and extremely positive uses.
  1189. 45:24And it is up to us, uh, to create these
  1190. 45:25ideas. You see, go when you are at
  1191. 45:27school, or whoever wants to, to that
  1192. 45:29site and donate your voice to us.
  1193. 45:33This donation serves to collect data
  1194. 45:35because AI feeds on data.
  1195. 45:38Last year we collected 15,000 Italian
  1196. 45:40human voices; we need all of them with
  1197. 45:42the accents of Fabriano, of Matelica,
  1198. 45:44from Fermo, like my Milanese one, and
  1199. 45:46with this we have given back a voice to
  1200. 45:48the first 96 Italian patients.
  1201. 45:51That’s not bad, right?
  1202. 45:52And this is a great idea, it’s a great
  1203. 45:55idea that came from your generation.
  1204. 45:58And we really need these kinds of ideas
  1205. 46:00.
  1206. 46:01So, to conclude, so I can hand the floor
  1207. 46:04back to Venus, we have two possible
  1208. 46:06reactions to this evolution—which I
  1209. 46:08wouldn't call a revolution, it’s one of
  1210. 46:11the many evolutions we face—:
  1211. 46:16either stay there frozen, a bit scared,
  1212. 46:18watching what happens, or get moving,
  1213. 46:20use all the human creativity that
  1214. 46:22distinguishes us and define a positive
  1215. 46:24use of this artificial intelligence that
  1216. 46:26isn't passive delegation, but rather an
  1217. 46:28amplification of human capabilities, and
  1218. 46:30then use it to improve people's quality
  1219. 46:33of life and improve the quality of our
  1220. 46:35society.
  1221. 46:37This is a bit of our mission; I hope,
  1222. 46:38Venus, that you will help us with this,
  1223. 46:40okay?
  1224. 46:41Thank you, good luck.
  1225. 46:54>> I had a question.
  1226. 46:57I mean, we have moved from an artificial
  1227. 46:59intelligence that analyzes data to one
  1228. 47:01that creates content.
  1229. 47:03So what could be the skills that young
  1230. 47:05people should acquire to, precisely, not
  1231. 47:07be replaced by these technologies?
  1232. 47:10>> Well, that’s a good question as you head
  1233. 47:12back to class or home. I urge you to
  1234. 47:15study in the broadest way possible.
  1235. 47:19I mean, don't specialize, wait to
  1236. 47:21specialize for as long as possible,
  1237. 47:23study everything you can study because
  1238. 47:25this is the only antidote we have to
  1239. 47:27these possibilities.
  1240. 47:30And this maintaining a broad culture of
  1241. 47:33ours is something I also tell adults,
  1242. 47:35okay?
  1243. 47:37Because we too have stopped studying a
  1244. 47:39bit and we can't afford that.
  1245. 47:43>> Thank you.
  1246. 47:46>> Thank you. Thanks Venus.
  1247. 47:49Give her a round of applause, she was
  1248. 47:50better at presenting.
  1249. 47:57>> Thank you.

About this transcript

This page contains the full transcript of Emanuele Frontoni - Intelligenza Artificiale e creatività umana nell’epoca dell’Agentic AI by Festa Scienza Filosofia, generated from the public captions YouTube serves with the video. The transcript has 7,302 words across 1,249 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.