Transcript: Why the Future of Enterprise AI Is Controllable Autonomy — Amnon Morag, VP Product at AI21 Labs

Eyal David: [00:00] Hi friends, this is Eyal. You've reached Product Builder, and here you can listen to the most interesting product conversations. My journey began in 2012, when I learned what product is at my own startup. Since then I've worked in the industry, and since 2017 my company has been consulting for a variety of companies. My drive is to help all of us improve as product people. And before we begin, if you enjoy listening to Product Builder, please rate us five stars on Spotify. It'll give me great feedback that the podcast delivers value to you, and it would make me really happy. So let's go, we're starting! Hi Amnon, how's it going? Great, good to be here. A pleasure to have you — here with us is Amnon, VP Product from AI21. Amnon, I'd like you to tell us a bit about the company, what do you do? Great, so AI21 is an AI company, it's been around for seven years already.

Amnon Morag: [00:52] We build large language models, LLMs; we're actually one of the only companies in the world that have the ability to do this, and also Israel's representatives on the map. And you also kind of started back in 2017, when this wasn't a commercial thing, right? Totally. We did — we were actually the first company after OpenAI to release an LLM the size of GPT-3, this was before the ChatGPT Moment, before the Hype, and we were also the first company to release a product based on Generative AI — Wordtune, a product that helps with writing, and really, before ChatGPT, it was very innovative.

Eyal David: [01:37] Maybe tell us about it in a couple of words — who is it for, a bit about what the product does, for those who don't know? Yes, so Wordtune helps with writing and also with reading,

Amnon Morag: [01:45] it also knows how to summarize text, and in writing, the central feature is essentially the ability to rewrite your text in a very creative and open way, that gives the user a lot of control; it's not, for that matter, like today in ChatGPT where we put in text and get one rewrite of it — as a user you actually have the ability to really choose the part, the paragraph, the sentence, the part within the sentence that you want to rewrite, and get many rewrite suggestions; it leaves a lot of control with the user, and creates a very interesting collaboration between human and machine.

Eyal David: [02:22] That's amazing, and it also really started long before — so tell me for a second, who is the real user of this thing? Of Wordtune, we have tons of types of users,

Amnon Morag: [02:33] there are students, there are professionals, tons of non-native English speakers, also tons of native English speakers, so we have many segments, we approach each one a bit differently and adapt the product to them, but we really target General Purpose for Writing.

Eyal David: [02:49] And it's also a Revenue Source, right? Absolutely. Okay, so let's go back to Wordtune and continue with the overview you were giving about the company.

Amnon Morag: [02:58] Yes, so we really, like I said, started with large language models, and we were among the world's pioneers in the field and also in the application, and really since the ChatGPT Moment, as everyone knows, this market exploded, became very dynamic, very competitive, and we continue to operate in it, and we look further ahead, and... and we're building the future of this revolution.

Eyal David: [03:28] So let's talk about the future of this revolution, because I don't know where you're taking it, but really since the GPT Moment, a big buzz was created in the world. A buzz that maybe doesn't quite reach your customers, the enterprises, right? Because they actually have a certain set of needs that this very moment doesn't answer, right?

Amnon Morag: [03:49] You could say — I think the buzz definitely reaches the enterprises too, and we even have the... the unique challenge of a market where the customer wants the technology, and this, I mean. They're just not entirely sure what for. And we see that today in enterprises, we're in the midst of the transition, from some mass-experimentation with the technology, to the beginning of significant adoption, and this is happening through players that are very sophisticated, and in certain areas that are a bit more low-hanging fruit. But we think that to reach the stage where AI really brings significant value to the enterprise, another revolution is needed within this technological revolution, and that's the transition to autonomous AI systems around the language models.

Eyal David: [04:48] So maybe let's actually explain why, explain who those bodies are and maybe what their problems are, and then maybe I'll shed a little light on this. Yes, so today, if I paint a picture,

Amnon Morag: [04:56] if we look at an enterprise organization, usually whoever lives there on the adoption of Gen-AI is R&D. Say a bank, for instance? At a bank, a financial institution, a medical institution, there's a CTO or an AI Center of Excellence, Head of AI, and they receive use cases from across the organization, and they try to solve them, and today they have dozens, dozens of use cases, and in the end they manage to bring to production a fairly limited number of use cases, and not necessarily the complex ones among them. And really what the enterprise mainly lacks, is a solution to a problem we all know, if we've also played with AI tools, also as consumers, and that's the reliability problem; in the end a language model does a lot of smart and wonderful things, but it's very hard to rely on it, hard to rely on the answer it provides me, and for an enterprise, even if a model does something excellent 90 percent of the time, that's not something, that's unacceptable. So you started talking about use cases,

Eyal David: [06:12] and earlier you also talked about low-hanging fruit, and so today you create some synergy between these things, you actually help these bodies find that value that's easy?

Amnon Morag: [06:22] Exactly. So we both serve the organizations, our customers, and with some of them we also go hand in hand and help them implement, but we're also building, as I already hinted, the next stage in the revolution, which we think will deliver AI to the enterprise, that is controllable, reliable and efficient.

Eyal David: [06:47] Which is completely different from what GPT and Claude and their like do, right? Yes. I think that, relative to those organizations, we're

Amnon Morag: [06:55] much more significantly focused on the enterprise and on the needs of enterprise organizations. We're not trying to build AI in order to build Artificial General Intelligence, we want to create AI that solves real problems today and brings a lot of value. Okay, Amnon, can you explain in a few words about the Jamba models? Yes, so Jamba is the family of language models we recently released, Jamba 1.5 is Large and Mini, and what's special about them is, first of all, their architecture; it's an architecture that isn't based only on Transformers like the other models we know, and its architecture actually allows the model to be the most efficient of the models that exist out there today, at long context. There's a context window of 256K, and the longer this context is, there are even more gains, in efficiency and run speed, together with performance that is generally comparable to and better than some of our competitors.

Eyal David: [08:05] Which, it's important to note, is exactly the opposite of how it was until now, right? The more I had, the longer my prompt was, the more it would lower the quality, right?

Amnon Morag: [08:14] First of all it lowers the quality, and we also show in the model's evaluation that we succeed at tasks that are based on context, no matter where the most significant information is located within the context window, and we show that we actually have a higher context window than all the competitors, certainly those that are also open models, and all of this together with really strong performance on their regular tasks, where they're measured — they're right there at the top.

Eyal David: [08:48] Wow, that's bursting with tons and tons of possibilities. Where do you think this will go in the end, like what's the vision? So we think efficient models will keep becoming more and more important, and in particular for enterprises,

Amnon Morag: [09:03] we're identifying the beginning of engagement also with resource management and the cost of Gen AI, at first the goal was just to manage to stand up something that works, and now we want to stand up something that works and is also efficient, so efficient models. That's a very important theme, and in addition to it also the long context window. The long context window, if we also connect it for a moment to the agentic vision that the world is currently chasing, then they'll have a very, very significant role. In the end an agent's memory, the ability to learn from past examples and improve performance over time, all these things, together also with analysis of a much larger amount of data, will require models with a long context window, and in particular ones that are efficient when they do it. So where is the future of this going? What does it look like? So the future is in autonomous AI systems that wrap around the language models, that use the language models as one more tool, as part of a broader tool set. And the goal of these systems, some call them Agents, is essentially to enable autonomy, controlled autonomy, the ability of a system to act flexibly, in order to solve a problem, and it breaks down a complex, multi-step problem, but does it in a way that's reliable, that provides guarantees an enterprise can rely on.

Eyal David: [10:37] Amazing, so maybe let's even give an example of this — you and I talked about it earlier, we gave an example about the bank, right? So the bank now wants to put out,

Amnon Morag: [10:45] let's — which use case can we use, the example where a language model would fail — say a bank wants to build, for its financial-analyst department, a tool that would let its financial analyst ask complex questions, comparing financial performance between different companies. And mistakes aren't allowed, we said. And mistakes aren't allowed, absolutely. So if we look at the technology available today, you'd need to take a language model, or even a language model that has a web-search tool available to it, or file search — if you ask it a question, like how would you compare the financial performance of a certain bank relative to its competitors, then what the system will do — the model will either answer from its parametric knowledge, and that will usually be wrong, or it'll know to use some retrieval engine from the internet or from files, bring the relevant context, in this case it would simply be documents that maybe already performed this analysis, and it'll summarize them. Okay. If it brings, maybe — if we take GPT-4o today, it'll go bring the top five documents from Google that perform the comparison between the bank and its competitors, and summarize them. The problem with this is that it's essentially some retrieval of information that was already created, in its summary, but what we want as an enterprise, what this bank wants, is for the AI to actually do the work, for it to save the time of analysts who sit and gather information. What would have taken it a long time — for it to take it in seconds, right? Exactly. To actually do the work in this context — if I want to compare a bank's financial performance, then I need to understand, first, who am I comparing this bank against, and on which comparison parameters am I going to compare it. Maybe I choose a set of five parameters. And then I want to go and make sure that I'm retrieving information that really finds that parameter, for each of the banks, for each of the competitors, performs some analysis on it, and in the end produces a summary that's grounded, gets into this analysis, really performs the analysis. That's something that today no AI-based system

Eyal David: [12:57] does reliably. It's funny, exactly, like we talked about earlier — so we said that a human, their difference is that they'll start from the end, start to cut out the picture, and exactly the model you need to predict the next word, that's exactly the opposite of that,

Amnon Morag: [13:11] it doesn't converge, after all. True. We as humans know not only how to plan how we're going to execute, and also to make sure we're following through on the plan, but we also know, while we're, for that matter, performing a search, to correct ourselves, and we know how deep to search, whether we should keep searching, because maybe we'll find more relevant information, or whether we've already exhausted this search and we'll move to the next one. All kinds of things that language models, as something that in the end predicts the next word,

Eyal David: [13:43] one token at a time, just don't know how to do. Wow, what an interesting job you have. So where do you think these companies are going? Which companies? Companies of all the Gen AI that focus on probabilistic models, let's say. So language models can still be improved.

Amnon Morag: [13:59] There are still many opportunities, in architecture, in size, in data. There will be a limit to it, but we're not there. But definitely most of the value we'll extract from AI, in the coming years, will come from systems that wrap the LLM and not necessarily from improvements in the LLM itself.

Eyal David: [14:24] Okay, all that's well and good. So tell us a bit how the company started, how you actually started there in 2017, right? Right, so we started, and we still are, a Deep Tech Company.

Amnon Morag: [14:37] The goal is to create unique and deep technology. And when we set out there was a technological vision, which has since only become even more relevant, significant, and connected also to the topic of the systems around LLMs. And that's the vision of what we call The Neurosymbolic Future, which is essentially meant to connect the values we get from statistical models, from the neural networks, and more logical Reasoning. We think that connecting these two kinds of Reasoning together is what leads the... will bring AI to the Next Level.

Eyal David: [15:21] Wow, it's amazing that in 2017 you were already thinking about this thing, even before the GPT moment, as happened — you, right?

Amnon Morag: [15:28] What actually was, what was in the market at that time? At that time, GPT-2 was out there, few knew it, we, we played with it quite a bit, and I think we saw that there really are, there are limitations, this was of course before the model grew very large, and already brought very impressive performance. But essentially the plateau of what we can get from the statistics — in the end we understood that we really weren't at it in 2017,

Eyal David: [16:06] and we think we're very close to it today. It's amazing, because usually a company is created to solve a problem, and you kind of didn't start from there,

Amnon Morag: [16:13] because there wasn't a problem yet. Right. It was a forecasting of the future, you could say. Really. And I think if we go back to product to also talk about the product, then the challenge of product within a deep tech company is exactly this kind of challenge. Essentially we don't always start from some problem we identify in the market and try to solve it. Sometimes we also want to invent a technological vision. We also want to refine the technological vision in order to extract value from it. In the end, there's some overlap between technology and value, and you have to find it. You have to connect both sides, you have to turn to the market and to the business and marketing stakeholders, and on the other hand, also to the technological vision and the deep research, and manage to find the optimal common denominator.

Eyal David: [17:16] Amazing, totally. And also, probably, the VCs and everyone believed in you, because you've raised quite a bit of money since then.

Amnon Morag: [17:22] Yes, over the years we've raised more than 300 million dollars, among investors like Google, Nvidia, Intel.

Eyal David: [17:33] And do you want to tell a bit about how you identified this potential? I mean, in the end, Amnon, in 2017,

Amnon Morag: [17:42] you identified that there's something here you want to take part in? Yes, so honestly I think I more tumbled — I tumbled into the thing, whoever... Maybe tell us even a bit about your bio, to give a bit of context actually. Yes, so back then in 2017 I was still a student, I studied law, economics, philosophy, history, and I actually worked at a political nonprofit, through which I got to know one of the company's founders. And when the company was founded, there really was the opportunity to join. When I joined, I didn't know that the company would tumble to where it would tumble, and not how I would tumble within it. And in the end it was an extraordinary experience.

Eyal David: [18:27] I think the future will continue to be like that. Yes, an amazing decision. And so you came from a product background? You're saying, I was in law, I was at a nonprofit. How did it get to you being VP Product now? So I really grew up within the company.

Amnon Morag: [18:44] At first I was, for some time, the only one in my kind of role in the company. And my role was to do business research, market research, and really already from then the goal was to look at a technological vision, and search for applications for it, and to address the complexity of the market and of the applications in various aspects. And slowly this really turned into a product role, it was the first product role in the company. And then I was actually the PM of Wordtune, which in itself was an interesting experience. I was PM of Wordtune for almost a year before the launch and almost a year after.

Eyal David: [19:32] Let's stop for a second, put a pin in this and come back for a moment. What a great way to enter the company, especially in a company of this kind where you come in and do research on the entire industry, probably understand where everything is going, make a digest of all this data, and now set out. So how long, actually, did you spend understanding? You actually worked until you moved to product, remind me? About a year. And how was that process? Did they just tell you, come to product, this is the thing you wanted — how was it?

Amnon Morag: [19:56] I wanted to, I understood this is an interesting position, and certainly in a company like this,

Eyal David: [20:04] and it very quickly tumbled there. And how did the whole Wordtune thing come about, when the company decided it was going in that direction?

Amnon Morag: [20:10] The company's vision was that we want to change the way people read and write. We have questions — it's a very significant and really also prophetic application of the way AI is going to affect our lives, and we started trying to understand what it means, to change the way people read and write, to understand in depth what writing means, what the challenges in writing are, and that's how we arrived at Wordtune, and honestly Wordtune was like magic at the time. It was also an experience for everyone who encountered it, because it really was a new beast, a product based on a language model, and essentially from that moment you functioned as product and continued at the company, or was there some moment you left? How was it exactly? So I left about a year after the launch of Wordtune, although I was very attached to the company and very attached to the product, I chose to go and do a master's and doctorate in law at Harvard, I moved there with the family, finished the master's, started the doctorate and decided that I don't want to do a doctorate, I came back home, back to the company — asking if there was some arrangement one could work in — so I kind of came back to work one day a week and did product projects, but really the company also changed a lot during the period I was there; when I left we were sixty and when I came back we were 200,

Eyal David: [21:38] you could say that day blew your mind, but it's amazing that you actually stayed there and kept plugging away one day a week. Yes, it's demanding, complex,

Amnon Morag: [21:47] together with the studies, but I couldn't, I couldn't leave completely. And now, if we look for a second at your challenges as Amnon, what does it look like? So the challenges of product in a company of this kind are really enormous challenges, the market is very dynamic, very competitive, and in the end, to succeed, we need to, we need to imagine the next step, we can't only serve the needs we identify in the market, we also need to identify the needs the market will have in a year and in two years and that the market today doesn't know how to tell us about, and that's a very, very significant challenge for product, we're constantly in this duality — on one hand we want to be very evidence-based and very much look at the market and solve real problems that exist today, and on the other hand we must have the vision, the leap of faith, the looking forward, essentially to help ourselves and the world imagine things that don't exist today.

Eyal David: [22:56] Wow, and it's even, like you said, at a vertical level, right? It's much more complicated, like I need to look at what banks and where banks will be versus other bodies, like, and versus the industry in general. Absolutely, another dimension of complexity is of course that we're looking at AI, also the expectation from AI to be general purpose, essentially to be something that's material in your hands,

Amnon Morag: [23:17] to build with, to use, and we need to move very versatilely between the use-case level, where a certain persona with a certain problem and there's a use case, and the need to build something general, generic enough that any builder in the enterprise will be able to build a solution for their own use case.

Eyal David: [23:38] Tell me, and you're surely hiring now or in hiring processes of one kind or another, to find product managers who fit your fit, essentially.

Amnon Morag: [23:47] Even understanding it is hard, now it's a challenge, it's a challenge. I think in Israel it should also be said, there aren't many more companies like AI21 that have this particular skill set, but there are tons of talented people, and we're really looking for people who will also reach a very deep technological understanding that's critical to succeed in a product role at AI21. And in addition to that, people who really know how to do the combination between being data-driven, evidence-driven, looking at the market, understanding the market, and on the other hand also vision, knowing how to imagine.

Eyal David: [24:32] Like, this transition between broad and narrow is really important, and generally the shift to broad. So what, so essentially you're looking for, you're saying. Already along the way — I'm the type whose skill interests you, but we're not dismissing it — but what you're saying. Essentially, it's that maybe because there aren't enough people who are product people with this expertise, you actually train up people who don't come from product in order to enter these roles, a lot, kind of like that.

Amnon Morag: [25:02] Less so today. We do take people from law and generally examine them.

Eyal David: [25:08] Yes. So today you actually accept people who are without product experience but, say, with technological experience, who integrate very well.

Amnon Morag: [25:16] Experience, technological experience and a product mindset when we can identify it in candidates, yes absolutely.

Eyal David: [25:24] Yes, so like if someone is listening to us and actually comes from this place, then what really are the product skills that in your eyes are most important? You touched on it a moment ago but I see you want to complete the thought.

Amnon Morag: [25:33] Yes, so really as I noted you need the ability to connect the two sides of the equation, very deep technological understanding and very deep understanding of the market, and also the ability to think about the here-and-now, about tomorrow morning's scorecard that delivers value to our current customers, and also about what the world looks like two years from now. So I think that the ability to move across these dimensions can be very important. Beyond that, product people need to be the kind who make things happen, who are able to manage interfaces, interfaces in an informal but very effective way, and it's simply people skills that we have to see, in whoever we hire.

Eyal David: [26:29] Amazing. With you too there's actually the matter of a lot of dialogue with the customer, right? You need to really understand tons of use cases, right? Right, living the market also means getting on calls with customers, traveling to meet customers, absolutely. And how is this divided at your place? By what? By verticals, by products — both — by modules within products and different product lines, you know. Tell me, and over your career, when you actually talked about this a few times, that this is what you actually expect from a product person who works with you, for such a skill, to find this overlap essentially between the value and the technological, how do you do it, how do you do it at your place?

Amnon Morag: [27:10] So it's a very iterative process, we actually — it starts, really, with mapping and understanding first of all at the conceptual level the technology, hypotheses about what it does and why it's good, and on the market side also, in a very conceptual way, dividing the world into the different segments and the different issues that represent pains. And then we're essentially in some balance that recurs all the time, we formulate some product vision that dialogues with what we learned both from the technology and from the field.

Eyal David: [27:57] So like, say you take your roadmap now going forward, so do you already know how to divide it essentially in an even way between the here-and-now or a year out, or the near quarter versus two years out?

Amnon Morag: [28:08] We mainly try to create a clear north star, that looks more than a year ahead. And to derive the here-and-now from it — we try, in what we build for the here-and-now, to almost never build something we don't think is on the way to the north star, and it's interesting because, like, this north star is dynamic like that, a north star. So what does it look like, a north star? It's an exploration, yes. There's no — we picture it as, there's a north star but it's covered by a fairly large cloud and maybe it also moves from side to side, and we take it one step at a time; the goal is essentially to create implementations that let us rub up against the world, rub up also at the level of the technology, meaning we're really building the technology and we learn a lot, and also rub up against the market,

Eyal David: [29:06] when there's something that's our jump for a moment into reality, we learn and then we can correct course. Is there some case you remember from recent times, the last few years, where the market really caught you like that, the market in some sense, where you really had to move it like you said?

Amnon Morag: [29:26] Yes, definitely, we've been building AI systems for quite a while now, and we have already experimented ourselves really with building things similar to what enterprises build for themselves, and that contributed a lot to understanding the challenges and where the technology today falls short.

Eyal David: [29:52] And now a simpler question, where do you think this whole world is going, the world of Gen AI, using AI in general?

Amnon Morag: [30:01] So I think we're only at the beginning, like I said I don't think we're overhyped, I think we're underhyped, we're going to see systems, autonomous ones that know how to perform complex, multi-step tasks in a controlled way, I think at least for the enterprise this is going to be the game changer that really creates the disruption expected within the enterprise, and also in the consumer world, it's going to be an arms race over the personal assistant, but it's a race that's also interesting to watch from the sidelines. And what do you think this will do to the job market? I think it'll create upheavals, I have no doubt about it, I think it's not going to be something singular in history, I think it is going to be similar to previous technological revolutions, and to the race, and we'll simply see change, I don't think we're approaching a future where you don't need human workers. It depends where, right? In this enterprise they'll probably cut a lot, this is actually interesting, I think maybe they'll cut in certain areas and increase in others.

Eyal David: [31:25] Yes, for example Elon from Gong sat here and argued, and I very much agree with him, that simply a lot more problems will be solved, which seems to me the direction this is really going, and when you suddenly have the thought about an agent within the story — you, as a legal person, I'm curious to know how you think about this, essentially that there's suddenly AI that knows how to make decisions on its own, how do you wrap this whole story, especially facing this enterprise,

Amnon Morag: [31:52] it's very intriguing. So of course when we talk about technology that is agentic technology, of an agent, our intention is not that we're creating an entity like a human, that can essentially make any decision within an environment and to take action; the technology in its essence needs to be exactly the kind that enables controllable autonomy — how do we create an autonomous system that's flexible enough to solve problems that have a high level of uncertainty and variance, but do it in a way that's controlled and that gives guarantees. That's exactly what enterprises need, and they're not going to adopt technology that doesn't give them that, and therefore I don't think there's here a risk of some agent that you simply release into the wild and wait to see what happens.

Eyal David: [32:57] Okay, another question, do you work with models that deal with more than just text, right?

Amnon Morag: [33:05] No, we only deal with text, and text-like things, that includes tables and code too, but we're not multimodal today in the sense of image and video processing. Is that because of the challenge in the thing? Again, we focus first of all on the enterprise market and we identify that there, right now, text and structured relational data is the most significant thing

Eyal David: [33:32] for the organizations. For the organizations. How fast do you release a feature or present it?

Amnon Morag: [33:40] So it depends — it's a feature; there are features we put out very fast, there are much heavier things, for example a language model, naturally, which has a substantial release protocol, but when we put out a feature, then we also tend, where it's possible, to let customers too touch it and use it, to give really early feedback, to gauge opinion, before there's even a wrapper around it? Yes, we try not to be surprised, still that... of course sometimes there's no choice and we release

Eyal David: [34:19] it. Okay, there's one more question here, what's the way to learn the field for someone who wants to get into it? Many people I see getting into the field,

Amnon Morag: [34:27] without prior knowledge, first of all simply play with all the tools, all the models that are out there, in the end there's in this field an experience of exploration with the technology, that's pretty amazing, and by the way I really see, when you see candidates who come to product who maybe aren't coming from the field but they played a lot with the technology and the products, they already grasp a pretty good understanding of what the limitations are and what the opportunities are, so that's something one can do,

Eyal David: [35:03] and of course there's no shortage of content. No shortage of content to read. So there's one more question here, really the last one, what's the company's rationale for continuing

Amnon Morag: [35:11] to share the models as open source? So we really share the models, and let you download the model's weights under a license that permits making use, certainly experimental and also commercial up to a certain cap, and we essentially only sell a license to the model to organizations that have revenue above a certain threshold. The reason we do this is that this way we get, first, community recognition, and a great deal of learning, we see how the model is used, how it's also fine-tuned, for all kinds of needs, and we also allow, with very little friction, organizations to start and experiment, so as a general strategy we think it serves us, and in a broader sense it serves the community — the more open models there are out there, we'll all advance.

Eyal David: [36:13] And in this respect, actually, are there enough Israeli companies that you feel are playing with and touching the models,

Amnon Morag: [36:20] or would you want more? How does it look relative to others? So there are quite a few in Israel who touch them, I think we can reach more organizations in Israel, there's a very, very friendly community here, but yes, we're in fairly deep interaction with the community.

Eyal David: [36:41] We'll put a link in the show notes and you can access it easily. Okay, so let's wrap up the episode, Amnon, what do you actually wish for yourself,

Amnon Morag: [36:48] ten years from today? Wow, a question that's hard to answer, within the situation — but for sure, not a doctor of law. Not a doctor of law? No. Although... who knows. Okay, but I think I'll continue to touch interesting things, that are cutting edge — I don't know if it's in technology or in other fields, but yes, that's what I wish for.

Eyal David: [37:16] So best of luck with it, thank you very much, and I think that if people want to get in touch with you, how do you suggest? There's LinkedIn, no problem, we'll put a link. Cool, Amnon, thank you very much for coming, I had fun, alright, bye for now.

Amnon Morag: [37:32] Bye bye.

Eyal David: [37:35] Friends, thank you for listening. If you found this podcast valuable, you can subscribe to follow us, of course, for more episodes on Spotify, Apple Podcasts or any other app; of course, if you didn't find us on some app, I'd love for you to write to us, we'd be very happy for five stars on any platform and for you to follow us so that more listeners can be exposed to us and find the podcast. You can also find the previous episodes on any app or on the YouTube channel; we have links in the description. Until next time, come on, be efficient, and bye bye.