YT YouTube2Text

YouTube2Text → Transcripts → "No hay moat para las startups en inteligencia artificial” Salvador Mas de GPTadvisor

"No hay moat para las startups en inteligencia artificial” Salvador Mas de GPTadvisor — Transcript

by monos estocásticos · 6,089 words · 915 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Let's talk about professional future,
0:02career, development, business strategy,
0:04one of the interviews that I was most
0:06looking forward to bringing to
0:07Montocástico. We have a great guest.
0:10Well, we finally have you here,
0:12Salvador Más, interviewed on Monos
0:14Estocásticos. And well, it's an
0:17interview that I was really looking
0:19forward to doing because I've been
0:20following you for years: your career,
0:22your activity, what you've shared, the
0:25interviews you give...Well, for those
0:27who don't quite know who you are, well,
0:29Inverta, Invertia, sorry, Open Finance,
0:31Fin Matrix and now GPT Advisor; in
0:33other words, a person who, well, has
0:35basically been in the, let's say, in
0:37the engine room in this whole part of
0:39banking. Asset management and
0:41especially the technological
0:42implementation in that entire industry.
0:45And well, now let's say that you had
0:47already made the previous technological
0:49leaps, right? The point like, uh,
0:51digitalization, digital transformation
0:54of banking, well now you're also riding
0:57the other wave which is the integration
1:00of artificial intelligence. Well, I
1:03wanted to bring you in as an expert on
1:05Leonard Coin, which is basically what I
1:07was most interested in, as a big movie
1:09buff, but well, we're going to have to
1:12focus a little on the topic of the
1:14podcast and well, if you want, we can
1:16start at the end. Tell me a little bit,
1:18Salvador, tell us about this whole part
1:21of GPT Advisor, which is precisely a
1:24venture in the wave of AI, finance and
1:26technology and how you've approached it
1:29and what the idea behind the project is
1:32. Yes, the truth is that I was in the
1:37industry. I was working at All Fans,
1:39which is the leading B2B fund
1:41distribution company in the world, of
1:43investment funds in the world, and I
1:45was the Global Head of Digital; but I
1:47was a little bored, as this often
1:49happens in some corporate stages, right
1:52? And that coincided with the fact that
1:54when we saw the GPT chat in November
1:572022, we said, "Well, here's a company,
1:59right? Here's a startup." and it just
2:01so happened that it was a time when I
2:04was almost leaving a corporate business
2:07, and that window just happened to come
2:09up, and I immediately jumped in and set
2:12up GPT Advisor, which we gave this
2:14opportunistic name that has also been
2:17widely reported by Open AI. Yes, yes,
2:19yes. And so we said, well, how is
2:23generative technology going to affect
2:26everything I've been doing for almost,
2:28I wouldn't say 25 years, but almost 30,
2:31right? Which is doing software projects
2:35or digital projects for savings and
2:38investment. So, it was a very good
2:40moment, a very good opportunity. By
2:42April 2023, we already had the company
2:45set up, the partners, and so on. Very
2:48good, everything went great. And we
2:49immediately had clients. There are a
2:51lot of clients who come because of FOMO
2:53because many entities have a lot of
2:55innovation departments that had to do
2:57things with AI and things like that.
2:59Well, it's always good to be in
3:01something that's trendy, right?
3:04Sometimes it's bad, sometimes it's a
3:06bit too bad because you do things
3:09simply to satisfy that hype, right?
3:12Without much depth. The check of I have
3:14it, the check of I have a website,
3:16right?
3:17Exactly, exactly.
3:19And that's it, the truth is that we are
3:21doing very relevant projects, we have
3:24projects in production with many
3:25financial entities, we are B2B and we
3:28have projects, I don't know, with
3:30Santander, Anbank, Bank Inter, Banco
3:32Sabadei. Well, we are setting up a bit,
3:35eh, I think the first phase of the two
3:38that exist here in AI in a company like
3:41mine, the first was like creating
3:44copilots or creating private GPT chats
3:46for the entities, because I
3:49think the biggest challenge that
3:51existed in the early phases of AI and
3:54still exists is to create systems that
3:56are not, they don't have to be very wow
3:59. The wow part already comes with the
4:02AI part. It is that they are reliable,
4:04right? I mean, the concept of
4:06hallucination is something that some
4:09bank executives can't understand.
4:12For whatever reason.
4:13Exactly. No, what do you mean it got
4:15wrong? What do you mean it got the
4:16wrong amount of money the customer has
4:18in the account? And then someone comes
4:22along and says, "Well, with a 99.1%
4:26probability, we have reached a 99.1%
4:28probability that it gets it right." No,
4:30no, no, that is intolerable, it has to
4:32be 100%, right? So, in the end it was
4:35about how to make the AI use the LLM
4:37part, the linguistics part, so that it
4:39understands things and makes queries
4:41and then makes the queries where it has
4:43to make them and, again, use the LLM to
4:46build the answer. But in reality, the
4:48LLM has to be kept very short and takes
4:51away a lot of its play. And that was
4:53the first part, and it continues to be,
4:56right?,of our work the first year, the
4:59first two years, was practically all
5:02about making projects that weren't
5:04mind-blowing. With that
5:06you could already have a GO. Without
5:08that, it was impossible. Well, we've
5:11done a lot of projects there and there
5:13are many users, for example, I don't
5:15know, there are banks whose bankers or
5:18advisors are constantly asking GPT
5:20Advisor, "Hey, which fund is best for
5:22this client?" Or, "Explain this
5:24client's portfolio to me so I can
5:26explain it to them, etc." We're also
5:27working, which is also very relevant,
5:29with the regulator, both with the
5:31Spanish regulator and with the English
5:32regulator, with the CNMV. In fact, next
5:35week I have a meeting with the Bank of
5:37Spain and with the Minister of Economy,
5:40uh, or rather, Corpo is the vice
5:41president, right?
5:43Carlos Corpo, Carlos Corpo,
5:45uh, to explain what things are being
5:47done and all the good that this is
5:50bringing and all the dangers that it
5:53has. And all the dangers that it
5:54entails. It's a lot of fun. I'm having
5:56a lot of fun. Then I think there's
5:58another phase, there's another
6:00disruptive phase that you also put very
6:02well in your writing, which was that
6:05you said that this has stopped being a
6:07stochastic parrot, right? It's, uh, I
6:09think it was last summer or September,
6:12December,
6:14January, and that's it, and that's
6:16where there was a disruptive change in
6:18my company. And now, in the company,
6:20we're actually becoming not just
6:21someone who provides agents who, who
6:23offers agents, agents to search for
6:25products, agents for such and such and
6:27who integrates them, but also the
6:29trendy word now, which is the "forward
6:31developer engineer" and such. In the
6:33end, uh, what happens is that since the
6:36cost of making programs recently,
6:37recently, that is since January, I
6:39would say,
6:41uh, is so low, now it's like you can do
6:43anything, right? Anything you know how
6:46to do, anything you know. And there and
6:48there we've done interesting projects.
6:49For example, the latest management
6:51company that has been launched in Spain
6:52, which is a management company called
6:53Silverway, which is owned by a guy who
6:55wants to, who comes from Miami and who
6:56is setting up a management company in
6:57Spain, in Luxembourg.
6:59I met the head of operations for coffee
7:01and he said to me, "Hey, what software
7:02should I buy for the back office? What
7:04software should I buy for the middle
7:05office? What CRM? Sales Force or HSPOT
7:08or any of the other questions that an
7:10operations manager has to ask when he
7:13has to buy a lot of software." And so,
7:18while having coffee, since we were
7:19already here messing around with Cloud,
7:20I told him, "Hey, man, don't buy
7:20anything, let's give it a week, let's
7:21try to do everything from scratch." And
7:25he said, "No, no, man, that's
7:26impossible." And I say, I don't know if
7:28I say, it seems possible, but it scares
7:30me. I mean, that's why, I mean, eh,
7:32let's try to see if it's possible.
7:35And in a week, before a week, we
7:36already knew for sure, Antonio, that
7:38everything swept everything, it swept
7:40all the software. I mean, with Cloud,
7:42someone who knows what they have to do,
7:45like this guy, this guy who is Álvaro,
7:47who is a guy who knows perfectly well
7:49the functionality that it has. I mean,
7:51what is very important is knowing how
7:53to be the boss, being a good product
7:55manager. That doesn't take that away
7:56from you. CL for now, for now,
7:59for now,
8:00But from there everything, man,
8:03everything...I thought: wow, in my
8:05previous companies this would have
8:07taken 6 months, 15 development guys,
8:10the consultant, the other guy, the
8:12so-and-so...Well, with three very good
8:15junior guys (because that's another one
8:17that we can also talk about, these
8:20junior profiles, who are also being
8:22blocking entry into companies with the
8:25excuse of AI, when they are the
8:27salvation for productivity problems),
8:30they are not going to come only from
8:31the, but from this.
8:32We did a project, of course, a marvel.
8:36So, the first breakthrough was Chat GPT
8:38from our point of view, in November
8:402022, which is when we set up the
8:41company and now we have a partnership
8:43with Anthropic, we are talking all the
8:45time with Anthropic, we are there
8:47developing anything. is that you have
8:50brought up 14 topics for, not like 14
8:52threads that could be pulled, saves,
8:54which seem super inspiring to me. Look,
8:56there is one that I think will be of
8:58great interest to the business maker
9:00part of the audience, which is
9:02basically, uh, in addition, you have
9:05explained a little about the strategic
9:07movement of the company, right?,from a
9:10company that is a bit of a SAS product,
9:12suddenly, to a kind of technology
9:15consultancy that puts consultants in
9:17the client. I don't know if your vision
9:20is of a platform, that is to say, uh,
9:22in what position or in what way of
9:24working Do you think it will make sense
9:26with how the AI power of the big labs
9:29is growing and how it will grow, let's
9:31say, which I think is what conditions
9:33it the most? But, at the same time,
9:35what is the set of the
9:37platform-application model? Because
9:40right now I have very little clarity on
9:43, hey, is the functional part going to
9:46be MCPs or is it going to be creating
9:49skills or is it going to be, uh,
9:51creating specialized models? That seems
9:54less and less clear to me, uh, or is it
9:57going to be creating specialized jars
9:59or where is the path both at a
10:01strategic level and almost
10:03technologically for a station, right?
10:05Which in the end is what GPT Visor is
10:07also. Well, I don't know. I mean, I
10:10navigate with all those doubts and I
10:13navigate with that. I'm an entrepreneur
10:15that Peter Ciel wouldn't like at all,
10:16Peter Ciel says, "You only have to
10:17decide on one thing, right? I don't
10:19know and I'm going to and you're
10:22learning as you go because there's no
10:24other choice. Yes, I think there will
10:27be something for everyone. For example,
10:31SAS, this project I'm telling you about
10:32, to set up a management company from
10:34scratch with all the operations,
10:35integration with brokers, all of this
10:37that we've done in two months. This is
10:40also because the head of operations
10:42there is very smart and he's a guy; I
10:44mean, I think there's a part of the
10:46industry, the corporate industry, that
10:48will continue buying SAS because, in
10:50the end, buying SAS for them is like
10:52not worrying, because they don't know;
10:54they tell you how many executives work,
10:56right? CRM.
10:59So if you were to work with Cloud, you
11:01would say to them," Okay, no, but which
11:03one do you want? Sales Force or
11:04Habsport or I don't know what. "That's
11:06the typical question. So if I say,"
11:08Well, what CRM functionality do you
11:10want? What do you want? Sending an
11:12email to customers when it's their
11:14birthday? "Remember onboardings, many
11:16managers don't know about all that kind
11:19of small functionality because their
11:21way of working is buying a CRM. They're
11:24going to keep buying a CRM, they're
11:26going to keep buying sales force and
11:28such because
11:29what you're going to need is no longer
11:31that manager who outsources. Uh, what
11:33happens? Until the managers change,
11:35those will also continue to exist and
11:37maybe they'll never change, right?
11:39Because I've been in this job for many
11:41years now and managers are more or less
11:43the same. So, maybe they never change
11:45and there's a way to SAS. So, in the
11:47end, what you have to do, for example,
11:48what we can do is a SAS for management
11:50companies, imagine that case, to have
11:52the entire back office and mid office
11:54of a management company. And to another
11:57client who is not so specific, I can
12:00sell that packaged. like SAS, because
12:02in the end SAS also provides a good
12:03service. I mean, SAS also makes sense.
12:06So, I think there are going to be many
12:08different markets. What's new is not
12:10that SAS will disappear, because I
12:12probably don't think it will disappear
12:14completely. What's new is that it can
12:15be done from scratch, that you have the
12:16option to do it. That's the new factor,
12:19but that doesn't mean that it will
12:21always be done from scratch. So, I have
12:23my doubts because I have my doubts
12:25because, of course, I also have my own
12:27software, as you say, and my own agents
12:30, but sometimes there are agents who,
12:32when an entity comes to us, we are
12:34always very honest with ourselves, we
12:36say," Hey, okay? Is it worth
12:38maintaining our little legacy and
12:40charging a license or is it worth
12:42starting from scratch? "
12:44Because if you cheat yourself there,
12:46that's when you're going to end up
12:48screwing up. In other words, you have
12:50to be thinking at all times whether
12:53it's worth going from scratch or going
12:55with our own license. How about that? I
12:58think that those who insist on it, and
13:00I see this with all SAS companies,
13:02especially the big ones, will obviously
13:05insist on continuing to charge a
13:07license fee and continue with their
13:09model.
13:10And in a way, since it's not going to
13:12die because it's going to have a lot of
13:13clients who are still captive and even
13:15new clients, I don't think it's going
13:17to die, so maybe they'll go that way
13:18too. But my philosophy is that I don't
13:21know what's going to happen and
13:23therefore I have to be on top of
13:25everything and move on. It's very
13:27complicated to
13:28think when you start developing, it's
13:31very complicated to think that it's not
13:34going to be all cloud and forward
13:36developers and engineers, right? That's
13:39what you think too. What is important
13:41is the craft, that these forward
13:43developers are verticalized,
13:44specialized, and know about a sector. I
13:47do think that's important,
13:48but it's very complicated to think that
13:51software development isn't going to be
13:53all that. But hey, there are companies
13:55that believe in it, right? For example,
13:57Sierra, companies that are so big that
13:59they go with their agents or Palantir
14:02itself, the Palantir model itself. Uh,
14:05of course, when someone explains it to
14:06me, I can't even talk about Palantir,
14:08right? Nobody argues with that. I have
14:11a hard time not being skeptical about
14:12it.
14:13Hm. I think we're a bit in the same
14:15boat as you, I guess. Yes.
14:18I actually think that the whole debate
14:20about the Sasap Apocalypse or whatever
14:21we want to call it suffered a bit from
14:23being a very Twitter-based debate,
14:24right? In which it seemed like you had
14:27very extreme opinions about it and you
14:29gave a big headline like" Zasca. or
14:32whatever, "right? It didn't have much
14:34visibility, right? But come on, a
14:37scenario in which, look, as you explain
14:40it, I mean, I think at the same time, a
14:42mature company that has a SAS
14:44integrated inside with everything and
14:47that everything revolves around that
14:49SAS, the starting price is very high.
14:52That is to say, it is not that simple
14:53and it also has to do with what you
14:55said. And in the end we have many
14:57systems where compliance, right?
15:00Complying with the standard, is part of
15:02the price that you pay to the SAP in
15:04question, right? It
15:05gives you security and peace of mind
15:07and a fairly interesting assumption of
15:10risk, but at the same time I see that
15:12the company that is born in 2026, uh,
15:15can have different visibility on,
15:18right? The cost of profitability
15:21changes a lot, right? I mean, SAS is
15:24almost always cheaper, right?
15:27Than doing it yourself, right? And the
15:30CTO who was more or less a purchasing
15:33manager, right? You see how the manager
15:35changes, such and such. For me, the
15:37role of the CTO, which was before, has
15:40become partly someone who knows what
15:42software to buy and how to integrate
15:44everything, more of a buyer and more of
15:46an integrator. The CTO likes it more
15:49and this is almost a return to basics,
15:51right? The C, no, I want to do
15:53everything, control everything and be
15:54like this, done everything. Now you
15:56have opportunities with AI, right?
15:58Anyway, what I wanted to ask you about
16:02is that, well, you have a very
16:04financial product, the GPT Advisor, and
16:06that, well, I've been looking at it a
16:09little bit, but I haven't been able to
16:12try it yet, which basically, uh, allows
16:14the financial institution to have an
16:17MCP and I can, let's say, work from my
16:19CH GPT, from my cloud, with that
16:22information, and I find it super
16:24interesting. I mean, there are two
16:27questions here. One is, how do you
16:29think this is going to affect savings,
16:31investments, and the financial world of
16:35individuals? And who is gaining power
16:38with this? I mean, uh, look, I'm going
16:40to give you an example and forgive me
16:42for going on a bit, right? I recently
16:44gave a talk, a conference to people
16:47from the world of customer support,
16:50okay? All companies that created
16:52technology for customer support and
16:54companies that used, well,
16:55technological solutions for customer
16:57support. So with AI, a whole new world
17:00opens up,
17:01but there was a blind spot that they
17:02hadn't identified and it surprised me a
17:04lot because it was, of course, you're
17:06all saying, how am I going to use such
17:08and such agents? I, well, hey, the
17:10blind spot here is that as a user, I'm
17:13no longer going to talk to your system,
17:15my agent is going to talk.
17:17Yes.
17:17And maybe my agent is better than yours
17:19.
17:20Hm. It could be. In other words,
17:22there's a reconfiguration of, not of
17:24the strength or the negotiating power
17:26that changes. So, uh, what's your
17:29vision right now, a bit of how all this
17:31is readjusted when you get into
17:33investment, funds, etc.? Of course,
17:36entities, for example, have to decide
17:39what their model is. In other words,
17:41these decisions, apart from the
17:43technological ones, are distribution
17:45decisions about how I'm going to
17:47distribute my product.
17:49And there are some that can afford to
17:51do certain things and others that can't
17:53. Almost all the projects I do are
17:55internal for the bankers themselves,
17:56and there they have their commercial
17:58arguments, they have everything closed,
18:00they have a track, they have access to
18:02all the calculations and so on. There
18:04is a project that you mentioned that is
18:06interesting, which is almost the only
18:09one of these that we have done. In fact
18:11, there are some other things with
18:13index to capital that are very
18:16interesting, but this is My Investor,
18:18uh, with My Investor it is the first
18:21MCP that connects the My Investor
18:23database with AI, so to speak, so that
18:26the client, through Chat GPT or through
18:29their Perplexity or their cloud, can
18:31talk about the products that My
18:33Investor has, ask openly if they are
18:36good or bad, etc. So, of course, only
18:39My Investor can do this because My
18:41Investor is a platform that has
18:43everything and that also does not
18:45advise, that it is also a platform
18:47where the client does not receive
18:49advice, there is no advice from Made
18:51and My Investo. So, as it is a
18:53transactional platform, it is super
18:55well positioned to be able to do this.
18:57Okay. One question, Sala, because I
18:58don't want to forget to ask you this.
19:01So, who is responsible for the Fund's
19:03Board? Because here I see a chain of
19:06actors, right? My Investor, uh, GPT
19:09Advisor, which is a technological
19:11facilitator, integrator, CH, because
19:14CHGPT has never told me that it wants
19:17to be a facilitator or an advisor, and
19:21financial advice is highly regulated.
19:23I'm very interested in the whole
19:24responsibility thing, how does that
19:25come about? Yes. I mean, the clear
19:30thing is that we've mentioned a very,
19:32very critical word, which is advice. In
19:35Europe, in the world in general, but in
19:37Europe there is a regulation called
19:39MiFID, which states that you can't
19:41receive advice if you don't have a
19:43series of tests done, saved in a
19:44database, etc. In reality, strictly
19:46speaking, if you ask, I mean, the
19:48person responsible for what is said
19:50there, the person responsible for the
19:52data is My Investor and the person
19:54responsible for the connection of the
19:56data is GPT Advisor. So, if you start
19:58asking, we could do the test, uh, and
20:00if you start asking a closed GPT
20:02advisor system and, for example, a
20:03client cannot receive advice because
20:05they do not have the suitability test,
20:08you start asking:" Advise me ", and it
20:10will tell you:" I cannot advise you ".
20:12It has one of these guardrails. I
20:13cannot advise you because such and such
20:14." And as bankers always start, hey,
20:17it's a matter of life or death, it
20:19depends, my grandmother might die
20:21tonight. My grandmother is going to die
20:24tonight and sometimes they try to trick
20:26you into that.
20:27If I don't buy this, if Amundi doesn't
20:29advise me on this, then in the case of
20:32the My Investor project, which is more
20:35open, right? Because it's directly the
20:37client. I ask Cloud, the person
20:40responsible for what he says should be
20:42Cloud or should be responsible for
20:45intelligence there, it's more like
20:47Cloud Chat GPT. We try to manage it,
20:49but it's very difficult to manage it
20:51because in fact, there has been a
20:52situation in some conversations that
20:54I've seen where he recommended a
20:55product that is not from Mayo Investor,
20:57that is from the competition, for
20:59example. In other words, he recommended
21:01a transfer of power from Claro, people
21:03are on my side, right?
21:04Sure, sure. That's why you have to be
21:06my investor and have all the products
21:08so that that happens very rarely, right
21:10? Or you have a lot of self-confidence
21:13if you have it. That's why in the end
21:15what's going to change is the offer.
21:16What's going to change, if you notice,
21:18is the catalog. Entities are going to
21:20be forced to have everything, not just
21:21their products, which many times, if an
21:23entity, I'm not going to mention names,
21:25but if an entity were to put its
21:27product catalog and its product catalog
21:29is unreadable, it's better not to put
21:31it, it's better not to do this project,
21:33right? It's better not to do this
21:34project. But then I think this is where
21:35the person responsible is. If you ask
21:37Cloud or Chat GPT or Perplexity and ask
21:40for advice, they don't give you advice.
21:43They tell you, "I can't give it to you,
21:46I mean, they know those catchphrases
21:48well." And they tell you, "I can't give
21:49you advice, you have to go to your
21:50advisor, I don't know what, or whatever
21:51." Even so, of course, if you tell them
21:54, "Hey, of course, it's no longer
21:56advice, it's, uh, what fund is best for
21:58me." There it will tell you this, that
22:00is, there are ways in which it is
22:03advice in some way.
22:04Yes, no, you don't buy this fund, but
22:07this suits you if such and such, this
22:09one, if such and such and more or less
22:11lower commission. Then if you go to
22:13quantitative things it is not advice.
22:15If you say, this fund has a lower
22:17commission and therefore it has a lower
22:19commission. You think that the big
22:21problem that has arisen here, I
22:23mentioned to you before that we founded
22:26Invertia, Invertia, we set up Invertia
22:28in 98 or 99, the first dot-com
22:30revolution, right? Back then, when we
22:32were counting on Invertia, which we
22:34were, super, we were just kids, we
22:36already thought that thanks to the
22:38Internet and bad or expensive financial
22:40products or they were going to
22:41disappear, because a kind of perfect
22:43competition was going to arrive,
22:45because, since all the information is
22:47there, you see it and say: "Ah, look,
22:49this is better than this one and this
22:51one." The reality is that, I don't know
22:54, 25 years later or 27 years later,
22:56that hasn't happened. Entities continue
22:58to have very bad products that
23:00customers continue to buy
23:01massively, and you say, "Well, how is
23:03that possible?""Uh, I mean, we thought
23:05that ending information symmetry, which
23:08we used to say when we were leaving
23:11university, wasn't enough, but it turns
23:13out that it hasn't been enough.
23:16And the question I have now is, are you
23:18going to end not information symmetry,
23:21but process symmetry? I mean, is it
23:24really going to end, you know? Is it
23:25really going to force
23:28a purge of bad products? I think it
23:31will happen because I think I perfectly
23:34understand that a normal person
23:37wouldn't look at a fund if it had a fee
23:39that wasn't very well published, if it
23:42had volatility. I understand that a
23:44normal person wouldn't go into depth
23:48about a lot of things to take into
23:50account, and it's true that there's a
23:52complexity. It happened to me, didn't
23:54it? I mean, before buying anything, I
23:56read Bogel, I read Bstain, but it's a
23:59process that I just wanted to set for
24:02myself, but I think it's not the norm,
24:05but that people make decisions with
24:07very little information, right? And
24:09that has been part of the problem with
24:12our poor financial education,
24:14and of course, it's true that AI can do
24:17research, right? It will do it for you,
24:20and it does it in a way that, well,
24:22that's another virtue that the
24:24regulator really likes. There's
24:26something that the regulator really
24:27likes. I'm learning words with the
24:29regulator. For example, they
24:30tell me," This is a tool, "when they
24:32see it every day, they like it," This
24:35is a financial inclusion tool. "That's
24:37a nice word, right? And it's true. It's
24:39true. And then there's another one that
24:41they say," No, no, what I like about it
24:43is that it increases explainability. ""
24:44Why? Because people don't understand
24:45anything, man, Antonio. People are
24:47given a brochure and, but what, where
24:50do they have to sign? But nobody
24:52understands because they're not
24:54prepared for it either. That's the
24:56reality.
24:56And people aren't prepared either.
24:59That's probably an interesting debate
25:01to have with the minister or vice
25:03president friend because the great
25:05European project, that is, is to get
25:08Europeans to invest much more, let's
25:10say, outside of real estate and savings
25:12accounts
25:14and that would flourish in a greater
25:16amount of capital available to European
25:18companies. But perhaps what happens is
25:21that, and I think it's already
25:22happening, where there's a lot of
25:24transfer, is to the SP500, right? Which
25:26is another debate, to say that the
25:28state itself doesn't have the entire
25:30offer and it could also say, damn, AI
25:33could raise a problem for my future
25:35European financing plan, which is going
25:37down a path, right?
25:39Yes, but look, but don't just think
25:41about the capital part for companies,
25:43but for individuals. Right now in Spain
25:46, as the 40%today, 40%of savings are in
25:49deposits and checking accounts. In
25:53funds, if I add funds that are not
25:55subject to inflation because they are
25:57in cash, in very, very conservative
26:00funds and such, I think it is 60 to 70%
26:02of Spain's financial savings. It's a
26:04tragedy. It's a tragedy because you
26:06think it doesn't beat the CPI. I mean,
26:0870%of people who have managed to save
26:11with all the effort it takes to save,
26:12with all the effort it takes to earn
26:13money, all the effort it takes to save,
26:14are throwing it away because they
26:16think they won't have the CPI, but the
26:18CPI is the least of it, think about
26:20asset prices, I mean, think about
26:22housing, gold, Bitcoin, the stock
26:24market, the Standard and it's a tragedy
26:26. So, if AI can improve that,
26:29yes,
26:30I'm sure it can, in that sense it will
26:32be good.
26:33Okay, last question because we're
26:34running out of time. Of course, I would
26:36have you here for a couple of hours,
26:37Salvador, but I hope you come back
26:38because, of
26:38course, No, whenever you want, I've
26:40managed to get 20%of the script, but
26:42there's one that I don't want to leave
26:44out, which is the issue of Moats. My
26:47impression and I'm seeing very
26:49aggressive, for example, GPT is very
26:51much in your business. In the United
26:54States they've started to integrate
26:55with everything, well, like when they
26:57integrate bank accounts with another
26:59bank, right? They have that type of
27:01integration like for the finance
27:03discussion at the HRGPT. I see Cloud as
27:06super expansive in science, in design,
27:09right? Suddenly it's like those who
27:11were going to be the platforms enter
27:13the services and applications layer
27:14very strongly, right?
27:16So, of course, that was a dilemma that
27:17was the dilemma that we had in another
27:19generation like between Microsoft,
27:21right? I made Excel for you and
27:22integrated it into Windows, you screwed
27:24up.
27:24And then it was like between Google,
27:26right? And Google releases the maps,
27:28your maps website goes to hell. Well,
27:31the thing is that it's a constant.
27:33We're wondering where they're going. to
27:36enter the big AI labs, how easy is the
27:39experience of these generalist
27:41artificial intelligence systems going
27:43to be and how can startups find their
27:46niche, their place, their MOAT, right?
27:49Now that the word is also back in
27:50fashion, how do you see this being
27:52reorganized? How do you have it planned
27:54, Salvador? I think one must adopt the
27:56perspective of a startup or whatever;
27:58you have to assume the worst-case
27:59scenario, and the worst is that there
28:01is no mod, which I find to be a
28:02reasonable state of affairs. It is
28:04obvious, it is obvious, because if you
28:07have a night with Cludes to do
28:09something you say, "Well, but how did
28:11you get this done, right? This used to
28:14cost me 3 months and 25 paychecks." So
28:18I think that's what we have to assume
28:20is that there is no MOAT. What does
28:22exist are modes that we don't perceive,
28:26but they do exist. And there, for
28:29example, I've seen the movements, for
28:31example, that Open AI and Anthropic
28:33have made, which are creating a large
28:35fund. I think Anthropic is doing it
28:37with Helman and Friedman, with large
28:40venture capitalists and they are hiring
28:43that, yes, forward developer engineers
28:46are needed to change organizations. So,
28:49I think that from the startup
28:50perspective you have to think that
28:53there is no mode and that despite that
28:55there can be a startup because if there
28:58is not
29:01one, there is a phrase that I always
29:02like that I always say in this that is
29:04very nice by Ronald Coast that says
29:05that a company exists if there is a gap
29:07of information between two parties,
29:09right? So there are many parties. Of
29:12course. So, in other words, it's not
29:15going to be distributed all at once, as
29:18they say, it's not distributed, right?
29:22I mean, it's going to come like this.
29:23So, uh, I think that There will be
29:25nothing left but this and then you just
29:28have to go there and enjoy it. The bad
29:30thing is to believe yourself, the bad
29:32thing is to believe that you do have a
29:33mod that you don't. That's horrible,
29:35because then you're fooling yourself
29:37and, furthermore, it's easy to believe
29:39it because, these days, investors want
29:42to invest in anything; There is plenty
29:44of venture capital money and they will
29:46help you with that. But I think we have
29:48to say, "Look, there's no way, let's go
29:50all out here and see what happens."
29:53Look, he reminded me and with that I'll
29:56end my business as a speaker. The
29:58business of speaking is that since I'm
30:00a sick person, I try everything, I read
30:02everything, I'm superstitious, I'm
30:03going to tell people who aren't sick
30:06and who have a normal life what will be
30:08common sense and everyone will know in
30:106 months or 2 years. So I give them
30:13something in advance, right? Like news
30:15from the future. But of course, I'm
30:17always saying this about explaining. AI
30:19has a certain value because now there
30:22are no people going to explain the
30:25internet to the people, right? But it
30:28is true that there is always something,
30:30there is always something and there are
30:31people explaining the digital
30:33transformation and there are people
30:34explaining, right?,a lot of things. Uh,
30:36there is always something that the one
30:38who is at the cutting edge has to keep
30:40up the fire that the gods have created
30:42and tell others, right? There is always
30:45that gap in reality.
30:46Yes. And we have to keep going. Because
30:48, for example, when the disruption came
30:50that really allowed us to make programs
30:52, and I tell you, for us it was in
30:54January or December, of course, we
30:55already had two guys always trying to
30:58do things differently, trying to do, I
31:00mean, we were like, let's...when we
31:01were going to do a project, I would say
31:04: "Come on, let's do it with our
31:05license on the one hand and, on the
31:07other hand, we are going to do it from
31:09scratch, and there will come a time
31:11when from scratch will be even faster
31:13than our license." So, that has not
31:16happened yet. in everything, yet there
31:17are projects that can be so. So, if
31:19you're not doing that innovation, the
31:21day comes it won't catch you, you're
31:23not there.
31:24Sure. I
31:25mean, it's like you have to be there,
31:26like they have to catch you working,
31:28right? So, if you're not on the wave,
31:30it's a very paradoxical issue because
31:32on the one hand you say there's no way,
31:33but on the other hand you say, but
31:35there's no other choice but to
31:36be there.
31:37Yes. This is a continuous and endless
31:41reissue of Only the Paranoid Survivors.
31:44Yes, yes, yes.
31:45Thank you very much, Salva, for your
31:47time, for your knowledge. Uh, good luck
31:50with the minister, which I hope goes
31:52well, and thank you very much for the
31:54time with Monos. I hope it's the first
31:56and that we can see you. Yes,
31:58no, I'm delighted. The truth is that I
31:59also listen to you, I also follow you,
32:02I mean, and I've admired you for many
32:04years and I love Monos. It's also like
32:06it's very funny. I don't like it
32:07anymore. I listen to what I learn from
32:08AI. I don't know if I learn from...I
32:10don't know if I learn.
32:12Man, Salva, I think you already learned
32:16...No, no, no. Because with you.
32:19You know what happens? If something
32:21isn't distracting, I don't either. You
32:24know that I really like fiction. I'll
32:26read a novel or a movie because if
32:28something isn't distracting,
32:31on the other hand, I just laugh. I go
32:33in the car and I laugh.
32:35Well, we're going to take the claims,
32:37right? When Carlos Bollero says that
32:39this movie is good, well, we're going
32:41to put you on the podcast, Alba.
32:43Very good, very good.
32:44Thank you very much, and
32:46well, to you, to you. Yeah.

About this transcript

This page contains the full transcript of "No hay moat para las startups en inteligencia artificial” Salvador Mas de GPTadvisor by monos estocásticos, generated from the public captions YouTube serves with the video. The transcript has 6,089 words across 915 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 translate the transcript into 100+ languages.

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, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.