Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking. Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:00mentioned the reopening of this route to the fulfillment of several
00:00:03demands, among them the lifting of sanctions, the
00:00:07release of blocked
00:00:09many people would stop at the assets and the recognition of its sovereignty over the
00:00:12strait.
00:00:19With this we close this broadcast of Telesur
00:00:22medicine, in science, in Noticias, thank you for your viewership, stay tuned to the Telesur screen, until
00:00:26next time.
00:00:54hallucination, not everything it
00:01:07advances that today have come out of
00:01:21Hello, welcome to text and context, it's great that you
00:01:26why do they do that, how do they follow us through our Telesur it doesn't come directly from Multimedia, today with another very
00:01:28interesting topic. Never before had we interacted with so many
00:01:32books, discovering books, talking about books, recommending them and
00:01:39used, uh, through we see many applications in even analyzing them on digital platforms. However, there is an accelerating trend that worries experts:
00:01:42artificial intelligence is
00:01:44changing the act of reading under automatic
00:01:48schemes and compressed
00:01:50it turns out that the researchers who are working on versions of books that replace the reflective and immersive process of
00:01:54traditional reading, with new applications that promise to finish, for
00:01:57example, a book in less than 10 minutes. Prioritizing speed over the
00:02:01depth of a work, is precisely what we are going to talk about today in text
00:02:05and context, with AI are we reading more, are we
00:02:14AI, book consumption increased. There are
00:02:17loyal readers of paper books, those who read digital
00:02:21already, everything we are reading in books, and those who prefer audiobooks, because
00:02:23reading is no longer an absolute experience and each
00:02:27reader can choose the way in which they approach a work,
00:02:31but this is far from those who ask AI for a summary of a
00:02:34book and with it assume they have read the work. This
00:02:38practice influences reading counters that measure
00:02:43reading are not developed. how much a work is consulted or read. AI
00:02:46has accelerated processes by making promises of applications to
00:02:50read a book in 10 minutes. These tools lead us to
00:02:54ask ourselves a necessary question. With the arrival of
00:02:57AI, does it mean that we are reading more? Well,
00:03:01there we had the profile of our program, I immediately move on to talk with my guests
00:03:04about the topic, in the studio I am accompanied by María
00:03:08Herrero, professor of the artificial intelligence career at the
00:03:11classical literature, in cases of what Humberto Fernández Morán University, thank you very much for your time with us and the
00:03:16stay with the information offered by the Latin American audience, also Lino Menéces from the General Directorate
00:03:18of Artificial Intelligence of the Ministry of Science and Technology,
00:03:22Lino, thank you also for your time with us, and from a distance I am also
00:03:25accompanied by Pablo,
00:03:27can be the product only of writer and philosopher, also an expert on this topic,
00:03:30thank you very much, Professor Pablo, for your time, to talk precisely
00:03:34about what we have commented on in
00:03:37the
00:03:39in some way and since they don't have what question that precisely
00:03:42makes this analysis possible, artificial intelligence and how it
00:03:45is precisely changing reading and writing
00:03:48habits, not only in
00:03:51education but also in
00:03:53train and well, one of any type of sector
00:03:56or factor that gives us reading, that
00:04:01our country. We have to start makes us read precisely text or an essay, whatever it is, artificial
00:04:03intelligence has been precisely one of the great
00:04:07challenges in terms of education, in terms of
00:04:10politics, of course, but also in literature and the way we are also
00:04:14approaching scientific essays.
00:04:18What to say about this? I'll start with you, María
00:04:21Paula, in this sense,
00:04:23because what some experts suggest is that the way artificial intelligence
00:04:27and social networks treat
00:04:30precisely the way
00:04:34start by differentiating of reading and writing is changing
00:04:37habits and the way we previously did it. What changes are you
00:04:40seeing? Are people reading more with artificial
00:04:42intelligence?
00:04:46Well, I wouldn't know if people are
00:04:49reading more with artificial intelligence, what I could tell you is
00:04:53that students or people are using
00:04:56it much more to try to research things and
00:05:00summarize, right? So we could say that instead of reading
00:05:04that's why they fall into hallucinations and a paper or reading a book or reading something for a
00:05:09gives you the critical task of seeing if what research, they first go to artificial intelligence to look for a summary. So that artificial
00:05:13intelligence summarizes it for them and based on that, they would then
00:05:16go on to read what follows, right? I suppose
00:05:22summary if they don't find it interesting, right, and they wouldn't read
00:05:25the complete work once they study
00:05:29the summary, however, there is concern because the way algorithms
00:05:33are treating the topics could
00:05:36even in artificial intelligence modify
00:05:40the content, there is also a concern, of course, because many
00:05:44people are currently using artificial intelligence as if it were a search engine or an
00:05:47exact summarizer of content and it is not
00:05:51like that, artificial intelligence is trained with a lot of
00:05:53information and when it gives you answers it is
00:05:57generating the answer word by word according to statistics and
00:06:00the frequency of words and an
00:06:03effect occurs that is called
00:06:07says is a hallucination but it can be hallucinating and can be
00:06:11inventing authors, these summaries
00:06:13essays, papers, therefore one cannot rely on what artificial
00:06:17intelligence tells you, you have to validate everything, right? It even
00:06:21evolved a bit and today it gives you the sources so that you can access
00:06:24the sources and validate, but even giving you the sources it sometimes
00:06:28hallucinates and invents sources, so there is always
00:06:32human work to validate that everything it is telling you
00:06:36is correct and not to assume it as correct, of course, Professor
00:06:39Pablo, are we reading more with artificial intelligence? Let's go to the topic of
00:06:43literature, because obviously on social
00:06:46networks, through the internet
00:06:50which they promise a person
00:06:53to read a book in 10 minutes and this is indeed
00:06:57changing the way we understand what reading is. Are we reading more,
00:07:01are we reading less with artificial
00:07:02intelligence?
00:07:05Well, first of all, I believe that
00:07:09artificial intelligence, like all scientific-technological revolutions that have
00:07:12occurred in the history of humanity, have challenging
00:07:15sides with positive and
00:07:18negative factors, they have dangers, right? It happened with the industrial
00:07:22revolution, with atomic energy, with the internet, in short, and the
00:07:25same can happen or is already happening with artificial
00:07:29intelligence. I believe that it is not bad or good in the archaeological
00:07:32sense, but depends on how it is used.
00:07:36I have the impression that students' reading
00:07:39at this moment, to a certain extent,
00:07:43is being partially affected in some
00:07:47cases when it is summarized, only those summaries
00:07:51are used and critical thinking, deep
00:07:55There is what artificial intelligence experts call a cognitive
00:07:58download because it is practically
00:08:01too easy. I agree with the professor in what she said
00:08:05that one first looks for a summary, that was done previously with
00:08:09before artificial intelligence appeared, the summaries
00:08:12or abstracts and then one would look for, let's say, if the article
00:08:16was of interest, but today what is happening
00:08:19is something that students and not
00:08:23only students, some researchers also,
00:08:28artificial intelligence summary that is not always really
00:08:32corresponds to reality or that does not
00:08:35take into account valuations that
00:08:40imaginative capacity. Let's not
00:08:42forget that Albert Einstein said
00:08:45that science advances more with
00:08:48imagination than with
00:08:51information and that imagination, Einstein said,
00:08:54develops more in art and literature
00:08:57than in mathematics. Yes, without a doubt
00:09:00it's a very interesting topic, especially
00:09:03distinguishing information from reading a book
00:09:07from the art of reading, Lino, because
00:09:09we obviously see with these networks,
00:09:13with all these
00:09:16and algorithms that well, they tell you, you can
00:09:20read a book in 10 minutes, however
00:09:23this is also used as a reading counter, this book is being read a lot,
00:09:27this book is not being read a lot, there is also an indicator on
00:09:31the internet that suggests that through
00:09:34this type, this format, more is being read, how do you
00:09:38see it as an artificial intelligence expert? Well, in the
00:09:41case of... artificial intelligence we would have to
00:09:46what artificial intelligence is because in this case we are mainly referring to large language
00:09:50models, to systems like RAG, so these are
00:09:53AIs that point to that segment that people are
00:09:57looking for, that consumers are looking for,
00:10:00in fact in the case of book production there have
00:10:04already been cases in which in scientific
00:10:07work artificial intelligence promoted
00:10:10authors within papers that did not exist, people used
00:10:14and cited authors who never existed, they were hallucinations of
00:10:17AI, they sounded very real, and when placed in
00:10:21Google it seemed they did exist, but they never made it to the paper and it happened. And
00:10:25it went through major publishers, but let's remember that the world currently
00:10:29also has a consumption pressure, to have
00:10:32numbers of which is the most read book, sometimes
00:10:35it doesn't matter to reach depth in that sense, many people don't read to reach
00:10:39depth, but rather read to accumulate reading, yes, to say well
00:10:43I read three papers today, I read three papers tomorrow, I can say
00:10:47this in this, I can say this in the other, but not reach depth and that happens
00:10:51many times and another thing that is not taken much into account happened before
00:10:55artificial intelligence is that on the internet what has more
00:10:58consultations is more visible and now with artificial intelligence what has
00:11:02more consultations is also more visible because that information is more repeated in
00:11:06the training data, it will be more capable of giving us
00:11:10authors from other countries than from our region, it will be more capable of
00:11:14giving us papers from other countries than from our region, it will also
00:11:17be much more efficient in giving you a
00:11:21summary of more read books, of more consumed books than of books
00:11:25from our countries or our regions, so that also needs to be looked at
00:11:29closely, without a doubt, and let's remember we are
00:11:32talking about what are
00:11:36language models and of course they are algorithms, in some
00:11:39cases they mention that it's like the Ouroboros
00:11:43which is like a snake that eats itself because
00:11:48what these language models have been trained for and the topics
00:11:52that are being discussed there, for example Latin American literature, we already
00:11:56talked about it before, in Latin American literature maybe you can't
00:11:59find something precisely because there isn't, through
00:12:03these language models, through this artificial intelligence,
00:12:06information about it, right? In what way is this Ouroboros working? How do
00:12:10you see it? Language models are trained
00:12:14with data. The data comes from how you
00:12:17get it on the internet, you get a lot of
00:12:20information, we could say Eurocentric
00:12:24like papers from recognized scientific
00:12:27journals, but not ancestral Latin American knowledge
00:12:31or knowledge that is oral, the
00:12:34research that is carried out that is not published in these types of journals,
00:12:38so by not training artificial intelligences with
00:12:41that type of information, they will only answer you with what they were
00:12:45trained with and that is lost,
00:12:48okay, so that's one
00:12:51side with the data that is put in and with what... the
00:12:55snake is that as language models are increasingly desired to be made larger and
00:12:59larger to handle more and more information, they run out of
00:13:02digitized data because that is the other problem that must be understood, that artificial
00:13:06intelligence, being algorithms and computers, uses
00:13:09digitized data, so the
00:13:13same people who train the models have started to
00:13:15generate data with the models, so they generate data with
00:13:19the models to train models and the data generated
00:13:22by humans has been somewhat lost, so
00:13:26the models are eating themselves because we are not talking
00:13:30about this hallucination that artificial
00:13:33intelligence has and that is taken for granted that
00:13:36it is a bibliographic reference,
00:13:39yes, because artificial intelligence deep down is
00:13:43doing, in a very reduced way, generating one word after another through
00:13:47a probability, it's like mathematics, it's not
00:13:50clarifying and seeing if the phrase it makes is a correct phrase or
00:13:54a literary phrase or anything, it just generates the word that makes
00:13:58the most sense that comes next and that's why many times it's, since
00:14:02it's a probability, there's a high probability that the phrase it's
00:14:05generating is a hallucination that it decided
00:14:08was the most probable, so they also generate texts with
00:14:12hallucinations that are also self-generating to
00:14:18the things that is said today is that as
00:14:21this race for large language models with a lot of information
00:14:25generating data with the same models is
00:14:28generating models that are flattening out, they are
00:14:31losing, there are even investigations already on the matter,
00:14:35there is also the issue of bias, right? Because this can also lead to a
00:14:38debate about whether there is bias or not regarding the content that can
00:14:42be taken into account
00:15:28It's no longer just a test of courage to face a
00:15:31risk, another kind of courage is needed, the courage of
00:15:35perseverance, the courage of...
00:15:38The courage of resistance, it is a significant moral
00:15:42problem, it is precisely that which brings
00:15:45us closer to the true concept of a revolutionary,
00:15:48to the concept of man, and more than of man, of
00:15:51a revolutionary, who, as Che said, is the most
00:15:55honorable title, the highest rank a man
00:15:58can hope for, to be a revolutionary, Fidel
00:16:02Castro. November 4,
00:16:051969, speech delivered at an event
00:16:08with members of the Revolutionary Armed Forces, FAR,
00:16:12held at the Ministry of the Revolutionary Armed Forces,
00:16:15Havana,
00:16:17Cuba. Instant news, with a
00:16:26complete vision of Latin
00:16:29America and the Caribbean, wherever you are, with the Telesur
00:16:33app, you have access to live broadcasts,
00:16:36exclusive videos and in-depth
00:16:40analysis, be part of the community that builds the future of the Global South.
00:17:25Thank you for following us in text and context, today talking about
00:17:28artificial intelligence, are we reading more with
00:17:32AI, are we writing more with AI, that is the question and the
00:17:35debate today, María Paula Herrero, professor of the artificial
00:17:39intelligence career at the Humberto Fernández Morán University,
00:17:43also joined by Lino Meneces from the General Directorate of Artificial Intelligence for the
00:17:46bias, in what some say, you also spoke
00:17:49about the way it is also affecting the cognitive
00:17:53process as well, well, critical thinking, we are
00:17:57facing that important debate with artificial intelligence
00:18:00taking into account how language models and all
00:18:04these applications are allowing, for example, reading a book in 10 minutes.
00:18:09Indeed, I believe that artificial intelligence is posing a very
00:18:13serious epistemological problem, which could even
00:18:16have antecedents in that neopositivist trend of the Vienna
00:18:19Circle, when the only thing that mattered was that the protocolary
00:18:23statements were logically well articulated, it didn't
00:18:26matter if they corresponded to reality or not, in such
00:18:30a way that artificial intelligence can build us
00:18:33an entire apparent framework that is very
00:18:36congruent, but... that does not really correspond to the
00:18:39objective world and therefore gives a totally
00:18:43biased image of reality. The other element that
00:18:46I agree with the professor on is its Eurocentric nature, because
00:18:50almost everything is in English first of all. The other is that
00:18:53it ignores other types of knowledge and since knowledge must have a
00:18:57multifactorial basis, it cannot be determined solely by one
00:19:01source, even if artificial intelligence is
00:19:04capable of receiving an infinite
00:19:08amount of information, even Noam Chomsky says that this
00:19:12artificial intelligence is a kind of authorized plagiarism, isn't it?
00:19:15Well, it feeds on many sources, but sometimes those sources are not
00:19:19reliable, they are totally false, right? And therefore, in this
00:19:23era when the so-called post-truth is talked about, which is not such post-truth, it is structured
00:19:27lies, I believe that artificial intelligence is posing
00:19:31a serious problem to education and scientific research. I
00:19:34just gave a conference in Mexico about
00:19:38educational technologies, about artificial intelligence and scientific
00:19:41leadership in which I raise that. The book
00:19:45has just been published in Spain by Dickinson
00:19:48because it is a very serious problem for students
00:19:52as well as for professors as well as for scientific
00:19:55researchers. We are returning to that same
00:19:58concern in academia about
00:20:02Eurocentrism in all studies, in all
00:20:05publications and books that have
00:20:09been seen over many decades, that same
00:20:12concern is also raised with artificial intelligence and its
00:20:16content. I think so, that of course we are
00:20:20carrying a whole Eurocentric heritage
00:20:23that underestimates these values of indigenous
00:20:27peoples, of other forms of knowledge, of other
00:20:30emotional, affective, volitional
00:20:33elements that make up the contradictory human condition that cannot
00:20:37be reduced simply to mathematical algorithms,
00:20:40perhaps artificial intelligence is a consequence of that
00:20:44hyper-rationalism that has characterized modern philosophy
00:20:47from René Descartes to today and that can
00:20:51lead us to a dead end of the
00:20:54robotization of all educational
00:20:57training of new generations and that is
00:21:01dangerous, although I repeat, artificial
00:21:05intelligence can be used in a very beneficial way and has
00:21:08demonstrated an infinite number of potentialities in
00:21:12education, but everything depends on how it is used, everything depends on
00:21:16how it is used and we talked Lino recently about
00:21:20how companies like Anthropic are taking books,
00:21:24destroying them and putting them directly
00:21:28digitally into a computer in which
00:21:31artificial intelligence remains there in a limbo, we don't know what
00:21:35it does there to then
00:21:37extract information from it, how are you seeing that process, obviously it's a debate that is not
00:21:41only happening in countries where artificial intelligence is already being
00:21:45cataloged as a superintelligence, but
00:21:48also within the United Nations, where we have seen recently the debates that it has been one of
00:21:52the central themes, but the way it is destroying books, as we saw
00:21:56in recent information and the way it is also
00:21:59being used in that sense, how do you see it? Yes, well,
00:22:02the topic of Anthropic and books starts from an ethical problem
00:22:06from the beginning. In the end, they are buying the books, but the utility
00:22:10they need is to fill the model with the largest
00:22:13amount of information,
00:22:16benefit? What they need is to extract words from the model
00:22:20using the professor's words, plagiarism, correct, they
00:22:24need those words that are linked to have existed
00:22:27before, nothing that AI says is
00:22:29totally its own, it already existed somewhere,
00:22:33we have the practical example right now with the resolution of
00:22:37the millennium problem, yes,
00:22:40the resolution of the problem had used artificial intelligence to
00:22:44make certain statements
00:22:47and the company took
00:22:50that chat, because they weren't even public ideas,
00:22:54those chats it had with the person, and
00:22:56used those ideas to solve the problem before
00:22:59them, because it had more computing power, because it's a mathematical problem, and spending
00:23:03more money, and spending much more money, yes, of course,
00:23:06so authorized plagiarism exists
00:23:11AI, all that series of words that are linked
00:23:14already existed at some point, that's why the more information it has
00:23:18it will answer better, but more information of what we would enter into the
00:23:21discussion, what does the company want, because we are talking
00:23:24about American companies
00:23:27mainly and Chinese open source models, what does the
00:23:30company want exactly, to sell you AI, to sell you the answer that AI has,
00:23:34to sell you tokens, yes, what generates the answer, so they
00:23:38need more information to generate more tokens and make them sound more
00:23:41real, and we omit in that process because there is not much
00:23:45digitization in our countries, the answers that
00:23:48we ourselves have... Now going back a bit to reading
00:23:52it leads us to another problem, if
00:23:54we start reading only what AI gives us, whether
00:23:58in practical cases, in cases of
00:24:01makes us reflect and grow as people, then
00:24:05we will have another problem that we will not have no we will not
00:24:09find answers from our territories, from our
00:24:12realities, but we will also start copying,
00:24:16importing solutions that appear in the books with which those
00:24:20models are trained, which is limited, of course, as we see that and as you mention,
00:24:23they need to destroy the books to be able to feed it
00:24:29for example in Latin America and the cultural
00:24:31richness we have here on the continent is not placed there then
00:24:35we are not doing exactly, they are very different realities, the answers of the
00:24:39models and one sees them day by day in our jobs, the answers of the models
00:24:42are adapted to other realities because they were trained
00:24:45with those other realities, when we use it in practical cases of
00:24:51retraining, looking for information, training other models
00:24:54to be able to adjust to what we need our problems, the resolution of our
00:24:58reality, of course, now Maria, uh, taking into
00:25:01account this whole scenario, what do you recommend for a
00:25:05person who is, of course,
00:25:08looking for information, trying to find information in artificial
00:25:12intelligence, well, taking into account all these aspects that we have
00:25:15maintained, what suggestions can a person have who says, "well, I'm going to look for
00:25:18information, maybe it will serve me or not serve me," look,
00:25:22I think it's a good
00:25:25time to rescue or to start
00:25:28working with a lot of force on critical thinking, that
00:25:31is, AI is a tool and it is a
00:25:35useful tool, the truth is that it helps a lot, but you cannot take what it
00:25:39tells you at face value, you have to think,
00:25:42analyze, reflect, look for sources, compare,
00:25:46this, many people use it as something fast and and so that it...
00:25:50fuses everything immediately and
00:25:54lies, because in the end it helps you in many things, but it also
00:25:59it is giving you is correct, you also have to take into account. that it has a
00:26:03bias, sometimes it's not even a geopolitical tool that
00:26:07has the bias, it's that the data with which it was trained
00:26:11was not sufficient and then this is
00:26:14not only the bias of the amount of data but that the data that
00:26:18are minority flattens them, removes them because it is not the most
00:26:21frequent so well use the tool
00:26:24but use it with critical thinking, ask yourself if that is correct, check
00:26:28the sources, especially the students, I for
00:26:32example. I have no problem with them doing homework using AI, but they have to
00:26:35analyze what AI is giving them,
00:26:38because it doesn't make sense to grab, copy, paste and deliver, no, because
00:26:42besides, the idea is that it is a tool, you also have to learn to
00:26:46think and reflect, because if not, we are not going to get anywhere,
00:26:49no, where will the new ideas come from, where will the new
00:26:52contexts come from, there is a very famous computer scientist,
00:26:56Grady Bush, who tells you that AI is not
00:26:59inventing anything, everything that is there is human knowledge.
00:27:02Do not put an AI only with books from 1800 to see
00:27:06if it advances with all the technical
00:27:10the human mind so use it but it is
00:27:13a tool especially now that there is so
00:27:17much publicity and a lot of boom because they are trying to sell it to you as
00:27:20a service there is a lot of smoke and one has
00:27:23to stop seeing those grandiloquent comments that
00:27:26are in the media and and realize that it is a tool
00:27:30that will help you, of course. and it is a challenge for students, for
00:27:34teachers, for families, obviously in the way, as Professor
00:27:38Pablo says, of how it is
00:27:41academia, through work, through, for
00:27:45example, education, very important also, uh, this
00:27:48aspect within artificial intelligence, unfortunately my
00:27:52time is up, thank you very much to you, Professor Pablo, thank you very much for your
00:27:56time with us, thank you, thank you
00:27:59all, so we finish www.
00:28:02surtv.net also on telegram the telesur news service 24
00:28:06hours a day, so we finish, it will be until a new opportunity, thank you for following us.
00:28:38After the triumph of the Sandinista Revolution, Nicaragua
00:28:41celebrated its first democratic elections organized by the
00:28:45FSLN. With popular support, candidate
00:28:49Daniel Ortega obtained a resounding victory,
00:28:52despite the opposition boycott and pressure from the United States,
00:28:56international observers validated the process,
00:29:00consolidating a project focused on social
00:29:03justice, literacy and national sovereignty.
00:29:07November 4, 1984,
00:29:10the FSLN wins the presidential
00:29:13elections.
00:29:17Plastic arts, literature, cinema and
00:29:20more. How much do you know about the contemporary cultural
00:29:23scene? Discover the connection between art and
00:29:27fiction beyond the conventional. In the company of
00:29:30artists who are redefining the multidisciplinary
00:29:32field. Neta fiction. The
00:29:36cultural magazine that connects you with alternative and emerging
00:29:40reality. Grand premiere.
00:29:43Saturday. Only on Telesur.
00:30:15What makes parents avoid the topic of sexuality with their
00:30:18children and what is its consequence?
00:30:22And to think that it is too early to talk about sexuality
00:30:25is rather based on my fear and my prejudice. And if that leads
00:30:29me to silence, then I become an accomplice. The
00:30:32lack of sex education that exists in families and in society
00:30:36in general, get guidance healthily with
00:30:39Aisa García and her guest, Dr. Vicenta
00:30:43Hernández,
00:30:45Saturday, only.