The AI revolution waits for no one. That’s why mortgage industry veterans David Lykken and Pavan Agarwal are launching weekly AI updates to keep you ahead of the curve in this rapidly evolving technological landscape. OpenAI’s Sam Altman recently made headlines with a stunning admission that his company “totally screwed up” the ChatGPT launch, acknowledging the loss of personalization and warmth that made the AI assistant appealing to its 700 million weekly users. While demonstrating impressive capabilities like generating a Netflix-style interface from a simple prompt, deeper testing revealed competitors like Amazon’s AI system still outperform OpenAI in critical areas like code generation. The conversation takes a fascinating turn as Pavan explains why AI development fundamentally differs from traditional software engineering. “You can’t scale by adding more engineers,” he explains, drawing parallels to how Newton’s simple F=ma equation revolutionized physics. In AI, mathematical breakthroughs from genius-level insights drive progress—not just throwing more resources at the problem. This mathematical foundation helps explain why Microsoft’s CEO admits struggling to monetize AI while Oracle appears to be doubling down on the technology. Trust remains the critical challenge for AI adoption. When even Sam Altman warns users not to fully trust ChatGPT’s outputs, how can businesses build sustainable models around the technology? The answer may lie in specialized applications, as Pavan reveals with Angel AI’s latest update that introduces predictive conversation capabilities, enabling the system to anticipate the next best question in a customer interaction—much like an experienced loan officer would naturally guide a conversation. Want to experience these cutting-edge AI capabilities yourself? Try Angel AI’s full version for free at askangelai.com and discover how artificial intelligence is transforming the mortgage industry. Subscribe now to stay informed with our weekly AI updates and never miss a beat in the rapidly evolving world of mortgage technology.
David Lykken: 0:31
Listeners. There is so much going on in the world of AI that I’ve decided we need to start releasing updates weekly, maybe multiple times a week, and I’ll be doing these with Pavan Agarwal of Angel AI, of of Celligence. It’s so important, it’s so important that you guys are staying up on it, and so we’re committed to teaming up, uh, to bring this information to you, pavan, good to have you here, friend, amazing, amazing, yeah, it’s. It’s exciting that we’re going to get to share these updates every week because, again, so much happening. You sent me a headline, an article that was published by fortune in yahoo finance, and the way the headline says Sam Altman admits Open AI. “totally screwed up”. It’s chat gpt launch and says the company will spend trillions of dollars on data centers. What happened here, Pavan? Let us know where? Where did they screw up? What happened?
Pavan Agarwal: 1:23
well, I don’t work inside Open AI. I have no idea what they did, but I can only guess what’s going on. So they did a big deal in their event where they showcased ChatGPT-5, and they gave it a simple prompt that says Organize my contacts in a Netflix-type interface, and it automatically generated the code and created a full application, a custom application for you that organized your CRM contacts to make it look like Netscape or Netflix. And people looked at that and were screaming like wow, that’s so cool, that’s amazing. And I thought, oh, that’s really interesting. So I had my team go in.
Pavan Agarwal: 2:10
I said, okay, go and really beat the bleep out of it, right and see Good. Does it do all that it’s promised to do? Right, and they, you know, a day later they came back and said we tried, you know, we thoroughly tested it, and they concluded that Amazon’s AI system is still far superior to open AI when it comes to code generation and software engineering.
David Lykken: 2:41
There’s a code generation really Interesting Okay.
Pavan Agarwal: 2:43
Right.
David Lykken: 2:44
And what is the name of that system?
Pavan Agarwal: 2:46
Well, Amazon remember, amazon bought a big chunk of a company called Anthropic.
David Lykken: 2:51
Yes, that’s right.
Pavan Agarwal: 2:52
So Anthropic has a system called Cloud. So Amazon owns a chunk of Cloud and then they have their own. I think they integrate it into their own coding platform. I think they call it Q or something like that.
David Lykken: 3:05
Interesting.
Pavan Agarwal: 3:06
Okay, I don’t use it at all.
David Lykken: 3:09
I, I myself, but my engineers use it so well but the biggest issue that they talk about in this is that, um, that when they release this if I’m reading this article correctly and listeners have a link to this article, so I encourage you you to go watch this or read this they said that it lost its personalization, it lost its warmth, it lost a lot of what was really appealing to users. 700 million users use it every week. That’s a staggering number.
David Lykken: 3:42
You’re having explosive growth in Agile AI, but not quite to 700 million yet we’ll be there soon yeah give us a minute, we’ll be there yeah, you’ll be there, but the point of it is is that these releases are so critical and this one they messed up on because it they lost some of what made chat gpt special. Right when. When you’re so, give us some insights from an ai. You’re one of the top AI engineers in the world. Already You’ve been recognized that your product is extraordinarily successful around the age. What should we be paying to? What are the lessons that we should be learning as users of these AI tools?
Pavan Agarwal: 4:17
I think the lessons that Chachi PT is learning is that and the irony of it is Sam Altman himself said it a couple of years ago that in the AI world, what used to take hundreds or thousands of people, a company with two or three people, can do the same thing Wow. And he’s learning that you can’t simply scale by adding a lot of people at it. And then his best people got recruited over by Meta. They took some of his best scientists out by writing big checks, oh that whole talent acquisition thing.
Pavan Agarwal: 4:52
Yeah, right, so he’s got a problem. So let me put it in another way right Is traditional software engineering, traditional software products. You could scale by adding more engineers. If I wanted more capabilities, if I wanted to do the latest, coolest thing, I’d throw more engineers at it and I’d get a better product. It’s not quite that simple but yeah.
Pavan Agarwal: 5:21
And you can get engineers with. If you it was a machine like you. You go to college or you go to a training program and you do XYZ and you come out and then you can write quality software, okay, and you literally almost anyone could do that. That was the old paradigm. It’s great, it’s a great process. Everything that we use today, even this application that we’re using for recording right now, everything is built on that kind of paradigm. However, ai is different. Ai is built on math, okay, so let me explain how fundamentally different it is. So Newton’s law is so simple. One little equation Force is equal to mass times, acceleration F is equal to ma.
David Lykken: 6:14
Simple. For you maybe, but for some of us that may not be so simple. All right, got it Simple, but one little equation?
Pavan Agarwal: 6:22
right, it takes a genius to come up with that one little equation. One little equation right, it takes a genius to come up with that one little equation. But everything in the modern world, nothing in the modern world, would run without that equation right? It’s a simple multiplication, but everything depends on that. One equation, right, and Maxwell’s equations on electromagnetism, where everything, this web camera that we’re using, you know, the phone on your, oh, on your iPhone.
Pavan Agarwal: 6:50
Right, yeah, all the wireless communication is four little equations, four multiplications, right. But it takes a genius like Maxwell to come up with those four equations, right? So AI is like that it’s a couple of math equations and then you just apply the math equations. Okay, so in order to scale ai, you need you need more of these sort of fundamental um. Uh, you know, you can’t scale it by just throwing more people at it and just yeah and that that’s important for people to understand, because well, well, we’ll catch up.
David Lykken: 7:25
We may have gotten late to the game, but we’re going to so much, throw so many resources at this, and I’m thinking of some industry leaders that you and I talk about, which we’re not going to mention, who are, you know, making bold statements out there on the internet and they don’t have a snowball’s chance of hell of getting into getting caught up with, like what you’ve built in the mortgage industry, because you’ve been at this for 24 years, 22, 24 years.
Pavan Agarwal: 7:48
We’ve been at it for a while.
David Lykken: 7:49
We’ve been at it for a while, a long, long time yeah.
Pavan Agarwal: 7:53
And look, I wish everyone the best, and I don’t you know, I’m all for it.
David Lykken: 8:00
Go innovate and do it All companies yeah.
Pavan Agarwal: 8:02
Yeah, and create stuff. I’m not bothered, right. But the point is, my advice is is anyone going down this path? The old model doesn’t work. You can’t just go and pick up a couple of really good you can take really good engineers, right, really good traditional software engineers, and you can’t put them on AI just like that and expect them to build something that’s going to solve, build something that’s going to solve your. It’s going to maybe solve some problems, but to solve big business problems it’s not going to happen, right? So, like the other thing Sam Altman said and also the CEO of Microsoft, nadella, said something similar along these lines is he says you know I think last week he was, sam was saying it over and over again don’t trust, trust ChatGPT, don’t trust ChatGPT, don’t trust the answers, don’t rely on the answers. It’s not professional advice. So on and so on. Right, because he was making sure he doesn’t get sued, right?
David Lykken: 8:57
Right, that was all Right. His attorneys are telling him to do that.
Pavan Agarwal: 9:01
Yes, yeah, trust it. Nadella of Microsoft says I can’t figure out how to make money with AI. Okay, and it’s?
David Lykken: 9:12
That was a profound statement, yeah.
Pavan Agarwal: 9:14
Right, and he literally said I’ve cut down on AI investments. And then so what happened is, the irony of it is Microsoft was early investor into OpenAI and now OpenAI just kind of deal with Oracle I don’t know how many billions of dollars they kind of deal with Oracle. They’re shifting to Oracle data centers. So just stop and think about this for a minute, right. So Microsoft in essence, is kind of shorting AI and Oracle is going long AI.
David Lykken: 9:47
Yeah, very interesting.
Pavan Agarwal: 9:49
Right, and the reason it all boils down to one thing is how do we make money If you can’t trust the answers? How do you run real businesses on it? It could be sort of, as Elon Musk said, superficial. It could do some superficial stuff for you, like helping you draft an email or helping you write a headline for your next flyer or stuff like that, but to do heavy lifting you have to have reliable AI Right, and that’s why Elon Musk told CNBC that look all these other AIs. They’re superficial, whereas the Tesla self-driving AI is very different. You can’t have a car hallucinate a green light when there’s a red light that’s so good Right.
Pavan Agarwal: 10:38
Yeah.
David Lykken: 10:44
So it’s the level of engineering that’s required, the level of science that’s required, this is going off a topic a little bit, but have you had a chance? To look at grok. Uh, yes, xai, yeah, yeah, xai. It is that one is really intriguing because it’s going in, it’s a has a different, there’s a different, there’s a shift there, pretty significant shift uh well, I I don’t know if there’s a shift.
Pavan Agarwal: 11:02
I think the fundamental still a transformer.
David Lykken: 11:04
As a question, is it a shift or how would you describe it?
Pavan Agarwal: 11:08
I think it’s just a cleaner implementation. It’s still math, right. It’s basically OpenAI is the first mover, so they got a lot of legacy stuff in there and then, on top of it, they just lost some of their best engineers to meta. Okay, and XAI started clean on a clean slate and is able to look at what everyone else is doing right and do it with a lot of money and perhaps do it in a neater, tighter package. So, you know, it’s kind of what they call the second mover advantage. Right, that’s what XAI has. So, but the fundamental architecture is still the same. It’s still a transformer. So the father of LLMs his name is Jan Lakan, I think, and he’s now working for Meta and he’s moved on. He says I moved on from LLMs. He said that problem is solved. There’s a limit to what they can do, and now I’m working on new types of AIs that think more like that, are modeled more closer to the human brain and the human brain learners.
David Lykken: 12:25
Yeah, so the last thing I want to wrap. So we wrap this up because I want to keep right. I want our listeners to keep coming back for fresh new information, so I don’t want this to go too long. One quick thing. So sam altman said in here he wants to rewire the internet, build brain computer interfaces and maybe even buy google chrome.
Pavan Agarwal: 12:44
Now that is a bold statement well, I mean, look, he’s just copying, build brain computer interfaces. Isn’t that Elon Musk’s neural link?
David Lykken: 12:54
Yeah, that’s his neural net.
Pavan Agarwal: 12:58
I think Elon’s way ahead of him on that, and that’s a whole different problem. So if I’m Sam Altman and I just had a screwed up release, what do I do? I talk. What would anyone else do in his position? Talk it big.
Speaker 1: 13:15
Yeah, talk big.
Pavan Agarwal: 13:16
Talk as big as possible. So now he wants to take on Neuralink. Next thing, you know, he’s going to want to build space rockets too.
David Lykken: 13:24
Yeah, right, right. So you know I mean what it really gets down to’s stay in your swim lane what you’ve done and what you’ve built. You’re staying in your swim lane. You’re building and expanding some things in that swim lane that are really exciting, pavon.
Pavan Agarwal: 13:38
So you got a quick update for us, an angel ai update, as we crap this up oh, yeah, uh yeah, we look for a press release is going to be out the next day or so. We just recently released a new version, a major new version, of the AI, which has been in the lab for years and I’ve been looking at it. I’m so excited. When I see the lab I’m like get it out to production right. Going from science, r&d you know, prototypes in the lab to production is two different things right and it’s, you know, prototypes of lab to production is two different things, right, and it’s.
Pavan Agarwal: 14:10
You know it’s thinking right, it’s for lack of a better word and I don’t want to over.
David Lykken: 14:18
Over-represent it.
Pavan Agarwal: 14:19
Over-represent it, whatever, but it feels like it’s thinking. It’s not really thinking. You know my feeling on that AI is not conscious, not even close to being conscious, but it really starts to feel like it’s thinking. And now the next we already released this, and then a next minor upgrade that’s coming out in the next week or so where it’s going to be thinking ahead for you. So when you have a conversation with it, like let’s say, for example, If you talk to a normal loan officer and you say, hey, what is the? How much down payment do I need for an FHA loan, right?
Pavan Agarwal: 14:49
So let’s say I’m a borrower and you’re the loan officer right, and David, you would respond and say, oh, that’s 3.5% and, by the way, where are you looking to buy right? As a good loan officer, you want to lead the customer, right, Right. So what we did, what we’ve done now, is the AI will predict right, Like if you use ChatGPT or XAI or something. You ask it a question, give it a prompt, it gives you a response and it stops right. But now we have enough data. We have years of conversational data, but we can predict what is the best.
David Lykken: 15:26
Next question to the user oh, that’s going to really help some of the newer loan officers that are coming up in the line. So that’s going to give some lift to the new ones that are in the market and using this tool Right, which is really interesting. Those that are listening to this and have not gone in and tried angel ai you can do so by going to ask angelai and actually work with the full version for free. And a lot of people, like I’ve told, I told you the other day I was at a party in the reno area with one of my clients and they go like, don’t tell pavon, but we’re using his ai constantly, we’re not paying for it, and they and I told them don’t worry, pavon is happy about that because he wants more users. So, listeners, go sign up for Angel AI, do so. You can check it out at angelaicom or just sign up and start using it at askangelai Pavan.
David Lykken: 16:22
Thanks so much. I’m looking forward to doing these every week on these headlines. There’s so much developing. So appreciate you being here, friend. Thank you, david Cheers, you bet. Thank you, listeners. I want to say a special thank you to Angel AI, who is the sponsor of this podcast. Like and share it and be sure to go to our Lickin’ and Lending website where you can see all of our sponsors for all of our podcasts. Thank you so much Like share this podcast with your peers. A lot of good information here, thank you.
Speaker 1: 16:49
You’ve been listening to Lickin’ on Lending. Our passion is to bring you the latest insights, trends and updates from the world of mortgages. So on this show, if you’re a seasoned industry veteran or just getting started, it’s your go-to source for all things mortgage. We hope you’ve enjoyed the show. We’ll be back soon. Take care and we’ll see you next time on Lickin’ on Lending.
[David] Look forward to seeing you in November and, seeing all that’s coming up. Thank you.
Important Links
- Article about Sam Altman admitting Open AI screwed up
- SunWest Mortgage website
- Angel AI website
- LinkedIn – Pavan Agarwal
Pavan Agarwal is a renowned leader in the mortgage lending industry and a pioneer in bringing artificial intelligence to the financial markets. Agarwal serves as the President and CEO of Sun West Mortgage Company and Celligence International, LLC.