This discussion examines integration of large language models with deterministic reasoning to deliver trusted enterprise artificial intelligence. The conversation explores hybrid architectures, mixture of experts strategies and auditable workflows that support regulated processes and operational governance.
Dr. David Ferrucci of Unqork, product and AI officer, joins theCUBE Research hosts John Furrier and Dave Vellante in the NYSE Wired studio to discuss the LLM sandwich and practical approaches to combining probabilistic and deterministic systems. Ferrucci brings decades of AI research and product experience; they examine how LLMs, formal reasoning engines and auditable workflows intersect for complex enterprise reasoning.
Ferrucci argues that LLMs alone cannot guarantee the precision required for regulated enterprise processes and must be integrated with deterministic engines to produce auditable outcomes. They recommend that organizations use LLMs to capture tacit knowledge and generate validated code or deterministic processes while prioritizing openness, governance and platform architectures that indicate when probabilistic or deterministic approaches apply. The discussion emphasizes governance, explainability and platform design as critical factors to consider when deploying AI at scale.
This episode addresses practical considerations for enterprises adopting hybrid AI architectures, including validation testing, monitoring and audit trails to support compliance and risk management. Viewers gain strategies for building auditable, trustworthy enterprise AI that balances probabilistic innovation with deterministic accountability.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Mixture of Experts Series
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Mixture of Experts Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Mixture of Experts Series
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Mixture of Experts Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Dr. David Ferrucci, Unqork
This discussion examines integration of large language models with deterministic reasoning to deliver trusted enterprise artificial intelligence. The conversation explores hybrid architectures, mixture of experts strategies and auditable workflows that support regulated processes and operational governance.
Dr. David Ferrucci of Unqork, product and AI officer, joins theCUBE Research hosts John Furrier and Dave Vellante in the NYSE Wired studio to discuss the LLM sandwich and practical approaches to combining probabilistic and deterministic systems. Ferrucci brings decades of AI research and product experience; they examine how LLMs, formal reasoning engines and auditable workflows intersect for complex enterprise reasoning.
Ferrucci argues that LLMs alone cannot guarantee the precision required for regulated enterprise processes and must be integrated with deterministic engines to produce auditable outcomes. They recommend that organizations use LLMs to capture tacit knowledge and generate validated code or deterministic processes while prioritizing openness, governance and platform architectures that indicate when probabilistic or deterministic approaches apply. The discussion emphasizes governance, explainability and platform design as critical factors to consider when deploying AI at scale.
This episode addresses practical considerations for enterprises adopting hybrid AI architectures, including validation testing, monitoring and audit trails to support compliance and risk management. Viewers gain strategies for building auditable, trustworthy enterprise AI that balances probabilistic innovation with deterministic accountability.
>> Palo Alto Studio Connection, Silicon Valley and Wall Street. I'm John Furrier, co-hosting here with Dave Vellante, my co-host.
David Ferrucci
>> Hi, everybody.
Dave Vellante
>> I'm Dave Vellante. Welcome back to theCUBE's NYSE Wired studio, high above the Options Exchange. Dr. David Ferrucci is an important figure in modern AI, and is widely recognized as the inventor of IBM Watson. He initiated and led the team that achieved that amazing landmark Jeopardy! victory in 2011. And then he later founded Elemental Cognition, where he focused on combining large language models with formal reasoning agents to solve complex problems. And the key is with transparency and rigor. And we're going to talk about that today. He's today the chief officer at Unqork, and he's responsible for the strategic application and integration of reliable AI across that platform, and he contributes to the AI research agenda in the Center for Global Enterprise's Institute for Advanced Enterprise AI. So what makes this conversation timely for our Cube audience is Dave's view is that LLMs alone are not sufficient for complex enterprise reasoning. That's probably not going to surprise you, but he's argued that businesses increasingly expect AI systems to solve problems that require precision. So how do you do that? You've got to have formal logic and accountability, but LLMs alone can't meet that bar strictly on how they work. So he's coined a term, the LLM sandwich, which we're going to talk to him about today. So basically the idea is using large language models as a fluent natural language interface to formal systems that are capable of reasoning. So in other words, think LLMs become the human interface, but deterministic reasoning, you've got to have engines to provide that accuracy. David, welcome to theCUBE. Thanks so much for coming on.
David Ferrucci
>> Thank you.
Dave Vellante
>> So I hear all the time, all the time that we're in the early innings of AI. When you hear that, what do you think? You've been at this for how many decades?
David Ferrucci
>> Exactly. When I was trained in AI, people used to talk about John McCarthy coining the term in 1954, if you can imagine, and that AI has been in the lab for many, many decades and there's actually been an enormous number of advances throughout the years. But once AI started tackling language as fluently and as capably as we've seen with things like large language models and with GPT, and it became a social phenomenon, Everything just took off and it became a common household word. Everyone's using it, not just in their personal lives and social lives, but it's becoming a huge force in industry because of how much it can accelerate the work we do.
Dave Vellante
>> When you saw something like ELIZA back in the day, did it sort of— it was sort of, in your mind, a harbinger?
David Ferrucci
>> In the audience, you know what ELIZA was?
Dave Vellante
>> ELIZA was kind of the original AI chatbot. Chatbot, correct. Right? and at the time, I think people felt like, wow, we could help people that were suffering from depression and have another companion. It was really groundbreaking. It ran on IBM mainframes. But at the time, did you see it as a harbinger for the ChatGPT moment? we went through the AI winter and through two AI winters. Yeah.
David Ferrucci
>> No, I think what happened— there was always this expectation that computers were ultimately going to tackle this problem, that we can get them to treat language as fluently, as capably as we can. And in many ways, that was kind of a measure of intelligence. Alan Turing set up the Turing test. This is, put a human and a computer behind the curtain. You speak to them. Can you determine which is which? Nobody thought for a very long time this would ever be achieved because of the struggle we had in dealing with language computationally. And now we're in a situation where I can always tell if it's a human or a computer. The computer is the one that's more intelligent and uses language more fluidly.
Dave Vellante
>> Yeah, right. It writes really well.
David Ferrucci
>> Completely past that.
Dave Vellante
>> Yeah.
David Ferrucci
>> So it's been— and really these advances have happened, very quickly. So the Foundations were laid over many, as you said, many, many decades. But, you know, in the last, you know, 5 to 10 years, things just absolutely took off in terms of, you know, the ability of large language models, these statistical techniques to process language and the impact that that's having on industry and society in general is dramatic.
Dave Vellante
>> How should we— how should the average person think about AI when you hear The frontier model companies saying slow down, we need regulation. There's a chance that it could wipe out humanity. And then on the other end of the spectrum, you have, you know, Jensen saying there's 0% chance. Does anybody know? I mean, for sure. How should we think about that?
David Ferrucci
>> I mean, the way— so I don't know that anybody actually knows. And I think there's a lot of different incentives to exaggerate some of these claims for a variety of different reasons.
Dave Vellante
>> Good marketing.
David Ferrucci
>> It's good marketing in a way.
Dave Vellante
>> There's no such thing as bad press.
David Ferrucci
>> So on one hand, we're being asked, you know, use them aggressively. On the other hand, we're being warned about how much damage they can do. Look, I think we will. We are already and will continue to learn how to control them. I think that underneath the hood, they're not as much of a mystery as most people think they are. I think that technology will continue to get easily reused and replicated. I believe the cost of AI will go down. I think it's within everyone's interest to open the box, if you will, and to make sure that the world at large can exploit these things and leverage these things. There's always risk, but I think that that risk, we're going to learn how to manage that risk. And the same technology that generates that risk, like it is in so many things, can help us guard against that risk. I think the larger risk is that we create monopolies. I think the larger risk is it's centrally controlled and managed. It really has to be open, in my opinion. It has to be allowed to be advanced in ways that are beneficial for humanity. And we want to open those doors up, if anything.
Dave Vellante
>> Well, don't you think that movement, the open source, open weight movement, is sort of heading in that direction? And to your point, I was talking to George Kurtz the other day in theCUBE, and he's very confident that we can use AI as a defense mechanism.
David Ferrucci
>> Correct. And also, in fact, it's our only chance. It's our only chance. these things are powerful in the types of inferences that they can make. And the enormous amount of space that they can navigate very, very quickly because of how much we invest in the underlying models themselves, how much data has been compressed in these models. And so that data can be used to generate different alternatives and so forth very quickly. It's very powerful. Yes. The best way to combat that with other kinds of models, we'd be defenseless if we couldn't advance those models from a defensive perspective.
Dave Vellante
>> So I think opening them up is sort of essential and you can use them for offensive, like red teaming as well, which can, expose things that you might not have seen before. Which brings me to determinism. Yeah. The discussion, of course, in the industry is we've got to bring the stochastic and the deterministic together. It's the only way we're actually going to be able to deploy AI in the enterprise, trust it and trust agents. And it's a journey.
David Ferrucci
>> Well, yeah, it's a journey and it's very ambiguous right now for a variety of reasons. One is I think that many people who are quickly becoming practitioners of AI don't necessarily have the background to even understand the significance of a deterministic algorithm and a probabilistic algorithm. And as it becomes easier and easier to create agents, which means programs that just through prompting and through probabilistic token generation can make decisions that you might not have anticipated, you might not have thought of. In some cases, you may be happy with them. In other cases, you may not be happy with them. But they weren't carefully planned and deterministically modeled, meaning that I know exactly how this is going to behave under these inputs. The classic investment in computers is exactly that, to give me that precise deterministic response. That's what we count on for regulatory systems. That's what we count on for transactional systems. Anything where there's any kind of asymmetric risk with a small mistake on one number can have catastrophic effects, we fully rely on determinism. At the same time, it's very hard, for example, to predict everything you might say if you're talking to a computer on the other end of a chatbot. Right. So deterministically trying to map out every one of those pathways seems overwhelming. So this is great news for AI, provided you govern that and you understand that in the end, I'm not going to allow that system to do anything that is not deterministically validated. So it is only hybrid architectures that combine agents with deterministic algorithms that will succeed in this space. That's the only way to deliver it. And you have to know that you're even doing that.
Dave Vellante
>> So I want to get your opinion on this, because to me, what's exciting about AI is so much of business is tacit knowledge. people call it tribal knowledge. Yes, I've got my deterministic systems. I trust my whatever Oracle or IBM database to do the right thing because it's deterministic. But there's so much human interaction where decisions are made. And right now it seems like AI is very good at mundane data entry, data-oriented tasks. It's not great at capturing that tacit knowledge, learning from the reasoning traces of humans and being able to act in a deterministic way. Thoughts on that and how do we get there? Am I off base in that
David Ferrucci
>> regard?I would argue that AI is better than BI today. Large language models are better than any technology in the past at actually being able to, through conversations, through dialogue, or through munging through the huge amount of knowledge inside of a corporation and getting at that tribal knowledge, understanding the intent, getting that past information. The question is what happens then?
Dave Vellante
>> Yeah. Okay.
David Ferrucci
>> And so turning that into reliable, deterministic processing is the next step. And so I tell people, half the time you want to tell an agent to do something, you should be actually using AI to help you write code that ensures that you're doing that deterministically. And you need to know the difference when a generative AI solution is actually beneficial over writing the code. So knowing that difference and having platforms and architectures that can both advise on when to use probabilistic agents and when to generate deterministic code that can be validated and tested becomes more and more valuable, right? So they become part of— we just don't look at them to fire off agents. We look at them as architectural systems that help us make these decisions and apply the right technology to the right place.
Dave Vellante
>> I think you're describing the LLM sandwich. So why don't you explain what that is, what you're doing?
David Ferrucci
>> Your work today was an interesting marketing theme around the concept that at the center of any hybrid architecture, you have a deterministic process, one that can compute answers through a formal inference mechanism that can be traced understood and mapped onto a known algorithmic problem-solving approach. But it also acknowledges that capturing that knowledge, that tacit knowledge, capturing those requirements— those requirements are so often written and understood by humans in natural language. And it's the translation from one to the other that matters, is can I take your requirements through a conversation and make sure that I formalize them in a way that I can guarantee those results. I don't necessarily want to take your language and put it in another language processing machine, which is an LLM. I want to put it in a machine that actually follows a process I can count on, I can validate, I can ensure that it delivers the business outcomes that you want.
Dave Vellante
>> And you're building that machine.
David Ferrucci
>> We are building that machine. It is that translation and doing that well with an architected auditable outcome. That's ultimately what matters. So LLMs have a huge role in that, both in the process of acquiring those requirements, teasing them out, helping to make sure they're complete, helping to make sure they're consistent, helping to make sure the human agrees with them, and to potentially formulate the output in a way that humans can consume it. But in the middle, you have to be very cautious, that you are actually following traceable, auditable, deterministic approaches.
Dave Vellante
>> Last question. What do you want to be able to say a year from now, year, 18 months from now, that you're not able to say today?
David Ferrucci
>> I want to say that we can trust the software we're all writing. That's what I want to be able to say. You know, we come from a world where you wrote something, you tested it, and man, you could be 99.99999% sure it's going to give you the answer that you expect. We are rapidly going into a world where that's not the case. It was just a couple of years ago I was working with a big airline and we were talking about developing a chatbot for them based on this deterministic process. But it still had an LLM front end. And they said every single phrase that that chatbot outputs has to be approved by corporate before we can ever approve this. That was just a couple of years ago.
Dave Vellante
>> Well, okay, so it sounds like we're going in the wrong direction right now.
David Ferrucci
>> Well, I don't think we are. I don't think we understand the speed at which we're moving and what the implications are for that.
Dave Vellante
>> Dave, thanks so much for coming on. Appreciate it. It was a pleasure to meet you.
David Ferrucci
>> Thank you.
Dave Vellante
>> All right. Thank you for watching. This is Dave Vellante for the NYSE Wired theCUBE coverage. We'll be right back after this short break.
>> Palo Alto Studio Connection, Silicon Valley and Wall Street. I'm John Furrier, co-hosting here with Dave Vellante, my co-host.
David Ferrucci
>> Hi, everybody.
Dave Vellante
>> I'm Dave Vellante. Welcome back to theCUBE's NYSE Wired studio, high above the Options Exchange. Dr. David Ferrucci is an important figure in modern AI, and is widely recognized as the inventor of IBM Watson. He initiated and led the team that achieved that amazing landmark Jeopardy! victory in 2011. And then he later founded Elemental Cognition, where he focused on combining large language models with formal reasoning agents to solve complex problems. And the key is with transparency and rigor. And we're going to talk about that today. He's today the chief officer at Unqork, and he's responsible for the strategic application and integration of reliable AI across that platform, and he contributes to the AI research agenda in the Center for Global Enterprise's Institute for Advanced Enterprise AI. So what makes this conversation timely for our Cube audience is Dave's view is that LLMs alone are not sufficient for complex enterprise reasoning. That's probably not going to surprise you, but he's argued that businesses increasingly expect AI systems to solve problems that require precision. So how do you do that? You've got to have formal logic and accountability, but LLMs alone can't meet that bar strictly on how they work. So he's coined a term, the LLM sandwich, which we're going to talk to him about today. So basically the idea is using large language models as a fluent natural language interface to formal systems that are capable of reasoning. So in other words, think LLMs become the human interface, but deterministic reasoning, you've got to have engines to provide that accuracy. David, welcome to theCUBE. Thanks so much for coming on.
David Ferrucci
>> Thank you.
Dave Vellante
>> So I hear all the time, all the time that we're in the early innings of AI. When you hear that, what do you think? You've been at this for how many decades?
David Ferrucci
>> Exactly. When I was trained in AI, people used to talk about John McCarthy coining the term in 1954, if you can imagine, and that AI has been in the lab for many, many decades and there's actually been an enormous number of advances throughout the years. But once AI started tackling language as fluently and as capably as we've seen with things like large language models and with GPT, and it became a social phenomenon, Everything just took off and it became a common household word. Everyone's using it, not just in their personal lives and social lives, but it's becoming a huge force in industry because of how much it can accelerate the work we do.
Dave Vellante
>> When you saw something like ELIZA back in the day, did it sort of— it was sort of, in your mind, a harbinger?
David Ferrucci
>> In the audience, you know what ELIZA was?
Dave Vellante
>> ELIZA was kind of the original AI chatbot. Chatbot, correct. Right? and at the time, I think people felt like, wow, we could help people that were suffering from depression and have another companion. It was really groundbreaking. It ran on IBM mainframes. But at the time, did you see it as a harbinger for the ChatGPT moment? we went through the AI winter and through two AI winters. Yeah.
David Ferrucci
>> No, I think what happened— there was always this expectation that computers were ultimately going to tackle this problem, that we can get them to treat language as fluently, as capably as we can. And in many ways, that was kind of a measure of intelligence. Alan Turing set up the Turing test. This is, put a human and a computer behind the curtain. You speak to them. Can you determine which is which? Nobody thought for a very long time this would ever be achieved because of the struggle we had in dealing with language computationally. And now we're in a situation where I can always tell if it's a human or a computer. The computer is the one that's more intelligent and uses language more fluidly.
Dave Vellante
>> Yeah, right. It writes really well.
David Ferrucci
>> Completely past that.
Dave Vellante
>> Yeah.
David Ferrucci
>> So it's been— and really these advances have happened, very quickly. So the Foundations were laid over many, as you said, many, many decades. But, you know, in the last, you know, 5 to 10 years, things just absolutely took off in terms of, you know, the ability of large language models, these statistical techniques to process language and the impact that that's having on industry and society in general is dramatic.
Dave Vellante
>> How should we— how should the average person think about AI when you hear The frontier model companies saying slow down, we need regulation. There's a chance that it could wipe out humanity. And then on the other end of the spectrum, you have, you know, Jensen saying there's 0% chance. Does anybody know? I mean, for sure. How should we think about that?
David Ferrucci
>> I mean, the way— so I don't know that anybody actually knows. And I think there's a lot of different incentives to exaggerate some of these claims for a variety of different reasons.
Dave Vellante
>> Good marketing.
David Ferrucci
>> It's good marketing in a way.
Dave Vellante
>> There's no such thing as bad press.
David Ferrucci
>> So on one hand, we're being asked, you know, use them aggressively. On the other hand, we're being warned about how much damage they can do. Look, I think we will. We are already and will continue to learn how to control them. I think that underneath the hood, they're not as much of a mystery as most people think they are. I think that technology will continue to get easily reused and replicated. I believe the cost of AI will go down. I think it's within everyone's interest to open the box, if you will, and to make sure that the world at large can exploit these things and leverage these things. There's always risk, but I think that that risk, we're going to learn how to manage that risk. And the same technology that generates that risk, like it is in so many things, can help us guard against that risk. I think the larger risk is that we create monopolies. I think the larger risk is it's centrally controlled and managed. It really has to be open, in my opinion. It has to be allowed to be advanced in ways that are beneficial for humanity. And we want to open those doors up, if anything.
Dave Vellante
>> Well, don't you think that movement, the open source, open weight movement, is sort of heading in that direction? And to your point, I was talking to George Kurtz the other day in theCUBE, and he's very confident that we can use AI as a defense mechanism.
David Ferrucci
>> Correct. And also, in fact, it's our only chance. It's our only chance. these things are powerful in the types of inferences that they can make. And the enormous amount of space that they can navigate very, very quickly because of how much we invest in the underlying models themselves, how much data has been compressed in these models. And so that data can be used to generate different alternatives and so forth very quickly. It's very powerful. Yes. The best way to combat that with other kinds of models, we'd be defenseless if we couldn't advance those models from a defensive perspective.
Dave Vellante
>> So I think opening them up is sort of essential and you can use them for offensive, like red teaming as well, which can, expose things that you might not have seen before. Which brings me to determinism. Yeah. The discussion, of course, in the industry is we've got to bring the stochastic and the deterministic together. It's the only way we're actually going to be able to deploy AI in the enterprise, trust it and trust agents. And it's a journey.
David Ferrucci
>> Well, yeah, it's a journey and it's very ambiguous right now for a variety of reasons. One is I think that many people who are quickly becoming practitioners of AI don't necessarily have the background to even understand the significance of a deterministic algorithm and a probabilistic algorithm. And as it becomes easier and easier to create agents, which means programs that just through prompting and through probabilistic token generation can make decisions that you might not have anticipated, you might not have thought of. In some cases, you may be happy with them. In other cases, you may not be happy with them. But they weren't carefully planned and deterministically modeled, meaning that I know exactly how this is going to behave under these inputs. The classic investment in computers is exactly that, to give me that precise deterministic response. That's what we count on for regulatory systems. That's what we count on for transactional systems. Anything where there's any kind of asymmetric risk with a small mistake on one number can have catastrophic effects, we fully rely on determinism. At the same time, it's very hard, for example, to predict everything you might say if you're talking to a computer on the other end of a chatbot. Right. So deterministically trying to map out every one of those pathways seems overwhelming. So this is great news for AI, provided you govern that and you understand that in the end, I'm not going to allow that system to do anything that is not deterministically validated. So it is only hybrid architectures that combine agents with deterministic algorithms that will succeed in this space. That's the only way to deliver it. And you have to know that you're even doing that.
Dave Vellante
>> So I want to get your opinion on this, because to me, what's exciting about AI is so much of business is tacit knowledge. people call it tribal knowledge. Yes, I've got my deterministic systems. I trust my whatever Oracle or IBM database to do the right thing because it's deterministic. But there's so much human interaction where decisions are made. And right now it seems like AI is very good at mundane data entry, data-oriented tasks. It's not great at capturing that tacit knowledge, learning from the reasoning traces of humans and being able to act in a deterministic way. Thoughts on that and how do we get there? Am I off base in that
David Ferrucci
>> regard?I would argue that AI is better than BI today. Large language models are better than any technology in the past at actually being able to, through conversations, through dialogue, or through munging through the huge amount of knowledge inside of a corporation and getting at that tribal knowledge, understanding the intent, getting that past information. The question is what happens then?
Dave Vellante
>> Yeah. Okay.
David Ferrucci
>> And so turning that into reliable, deterministic processing is the next step. And so I tell people, half the time you want to tell an agent to do something, you should be actually using AI to help you write code that ensures that you're doing that deterministically. And you need to know the difference when a generative AI solution is actually beneficial over writing the code. So knowing that difference and having platforms and architectures that can both advise on when to use probabilistic agents and when to generate deterministic code that can be validated and tested becomes more and more valuable, right? So they become part of— we just don't look at them to fire off agents. We look at them as architectural systems that help us make these decisions and apply the right technology to the right place.
Dave Vellante
>> I think you're describing the LLM sandwich. So why don't you explain what that is, what you're doing?
David Ferrucci
>> Your work today was an interesting marketing theme around the concept that at the center of any hybrid architecture, you have a deterministic process, one that can compute answers through a formal inference mechanism that can be traced understood and mapped onto a known algorithmic problem-solving approach. But it also acknowledges that capturing that knowledge, that tacit knowledge, capturing those requirements— those requirements are so often written and understood by humans in natural language. And it's the translation from one to the other that matters, is can I take your requirements through a conversation and make sure that I formalize them in a way that I can guarantee those results. I don't necessarily want to take your language and put it in another language processing machine, which is an LLM. I want to put it in a machine that actually follows a process I can count on, I can validate, I can ensure that it delivers the business outcomes that you want.
Dave Vellante
>> And you're building that machine.
David Ferrucci
>> We are building that machine. It is that translation and doing that well with an architected auditable outcome. That's ultimately what matters. So LLMs have a huge role in that, both in the process of acquiring those requirements, teasing them out, helping to make sure they're complete, helping to make sure they're consistent, helping to make sure the human agrees with them, and to potentially formulate the output in a way that humans can consume it. But in the middle, you have to be very cautious, that you are actually following traceable, auditable, deterministic approaches.
Dave Vellante
>> Last question. What do you want to be able to say a year from now, year, 18 months from now, that you're not able to say today?
David Ferrucci
>> I want to say that we can trust the software we're all writing. That's what I want to be able to say. You know, we come from a world where you wrote something, you tested it, and man, you could be 99.99999% sure it's going to give you the answer that you expect. We are rapidly going into a world where that's not the case. It was just a couple of years ago I was working with a big airline and we were talking about developing a chatbot for them based on this deterministic process. But it still had an LLM front end. And they said every single phrase that that chatbot outputs has to be approved by corporate before we can ever approve this. That was just a couple of years ago.
Dave Vellante
>> Well, okay, so it sounds like we're going in the wrong direction right now.
David Ferrucci
>> Well, I don't think we are. I don't think we understand the speed at which we're moving and what the implications are for that.
Dave Vellante
>> Dave, thanks so much for coming on. Appreciate it. It was a pleasure to meet you.
David Ferrucci
>> Thank you.
Dave Vellante
>> All right. Thank you for watching. This is Dave Vellante for the NYSE Wired theCUBE coverage. We'll be right back after this short break.
Dave Vellante