Artificial intelligence may no longer be the most provocative frontier in computing. In this episode of the Weekly AI Update, David Lykken and Pavan Agarwal explore the emerging field of biocomputing, where researchers are integrating living human neurons with electronic systems to perform computational tasks. From the remarkable energy efficiency and potential capabilities of biological computing to much deeper questions surrounding consciousness, ethics, regulation, and the boundaries of scientific innovation, this conversation asks whether we are witnessing the beginning of an extraordinary new era of computing—or approaching a line technology should not cross. David and Pavan also look ahead to developments at Angel AI, including a newly awarded patent involving the management of AI context windows.
[David] Folks, welcome back for another AI update with my good friend, Pavan Agarwal. Pavan, good to have you here.
[Pavan] Thanks, thanks for having me back.
[David] It’s always fun to do these. Okay, folks, so we’re wearing the same outfits as what we were the last one. Yes, we’re recording these back to back, but we’re gonna break them out. Pavan, I want to talk today about the latest Angel AI research. You have just published another white paper. Tell us about this. What’s so significant about this yet another really informative white paper?
[Pavan] Yeah, so this white paper is about fair lending and the impact, if fair lending is done correctly, the impact it has on the national economy. So the bottom line is as much as a quarter percent in GDP increase by simply delivering…
[David] Wow, did you say a quarter GDP?
[Pavan] A quarter in GDP. I mean that’s massive. That’s massive. I mean we’re, you know, folks, we’re thirty trillion dollar plus economy and that means just by doing the right thing and not spending a dollar more of the government’s money, of public money, just by using the current programs correctly and by using the banking system correctly, we can increase our national GDP by a quarter point. I mean
[David] Yeah, I should read the title of this. Operationalizing fair and accessible lending with AI from systemic loan declines to national recovery, evidence from Angel AI and HMDA. So it looks like you’ve taken a lot of a HMDA data, and you’ve gone through, you and your team have gone through and started looking at this. When you mean from systemic loan declines, are you talking about volume?
[Pavan] That that is correct. So what we did was we looked at the HMDA data for the past three years from 23, 24, 25. And we looked at the how loans are declined, and we broke it down by race across the whole spectrum. And then and then we compared against Angel AI’s decline rates, okay, for the same time period across the across the same spectrum. Okay, and the result is that You know, depending and if you could put that chart that graph up, I know those of you who listen to the podcast want me see it, but you can you can on page five of the white paper, you can find it, and you can find the white paper on angelai.com/white paper. And you could it is publicly available, anyone can download it and go to page five, and there’s an there’s a chart that compares the national decline rates with Angel AI decline rates. Okay. And the delta is significant. and interestingly enough, the broadest difference is in Asian borrowers where from four percent to something like twenty-nine percent or something, like eightfold, eight or ninefold difference in decline rates. And you know, and can you can you put the chart up?
[David] We’re now have page five, where you actually have the graph. And by the way, when we’re talking about decline, we’re not talking about a decline in volume. We’re talking about declination, a turndown. When we say in declines, the turndowns, the rate of turndowns. Okay, so we have that clarified. And it’s the decline percentage by ethnicity. Let’s run what are you what are we looking at here, Pavan?
[Pavan] Okay, so and on this chart you have two bars next to each other, the orange and the blue.
[David] Right?
[Pavan] And the orange is the national based national declines based on HMDA. Okay, and the blue is the for the same time period declines by angel AI. So app
[David] Yes, it says the blue says in the in the legend up your son was mortgage, but it is being run by the angel AI technology, is what you’re saying, correct?
[Pavan] Correct. Correct. Right. So
[David] Look at the significant difference. So in Asians, we’re looking at a decline rate of 32.98, almost 33 percent. And if it’s run through those same loans run through Angel AI, it’s only four point one percent. That
[Pavan] Correct, but but we c we can’t say same loans because we have to we don’t have the data that the other that the rest of the industry in
[David] I see, yeah, of course, yeah.
[Pavan] Industry is doing. What we have is we have our sample, which we can say is a statistically relevant sample, is pretty you know, we’re getting we’re
[David] Right. Okay, that’s a that’s good clarification. Thank you.
[Pavan] Yeah and there’s no adverse selection. I mean we’re getting it from retail and broker channels like everyone else is. And so we’re getting this flood of loans, and that would be a representative sample of the national loans that originated. Okay. And the and also remember, this is not just one year, it’s not an anomaly. We’ve done this for the purpose of this white paper, is going from 23, 24, 25. Internally we’ve done this going all the way back to 2020. Okay. So it’s and the results are consistent year after year. in fact, if anything, the results we’ve gotten, you know, Angel AI has performed better and better every year. So what that means is like four percent versus what is that, 29%? I can’t see from right?
[David] 29%, yeah. No, 30, so it’s in the Asian population is 32.98 and 4%, 4.01 in the in the Sun West sampling of documents. Again, representative for blacks it’s 44.32 percent declination rate versus what Angel AI is 18.55%. I mean, that’s significant.
[Pavan] Yes, right. So if you if you think about it, like for Asians, that’s you’re eight times more likely to be declined if you if your application did not go through angel AI. Right? And and in and for black population, I don’t remember the numbers, but it’s about four times more likely to be declined than if you hadn’t gone through angel AI. And what’s interesting when you when you look at this graph from a distance, you’ll see that, you know, black, white, Hispanic, Asian, under Angel AI they’re they’re almost about the same now.
[David] Fascinating. It’s really interesting, Pavan, what this does. How why is that? Because is it because we removed the human factor in the decision?
[Pavan] Yeah. and before I go into that, can you read the numbers for white? Because it this improved everyone. Not so this isn’t yeah.
[David] yeah. Yeah. So for whites it’s thirty-three point three one percent is the Nash 2025 national HMDA data, and angel AI is nine point two four percent, a twenty-four point seven gap. Now that’s the smallest gap, but whereas Asians were at twenty-eight point ninety-seven percent. So when you look at that disparity that’s there, I mean they’re all pretty close. That is significant gap.
[Pavan] Yes, exactly. So my point here is that because there’s still a gap on it’s not like okay, there’s no gap with the white cohort, but there’s a gap with the others. But there’s a similar gap across the board. Okay, right? And obviously for certain races it’s even higher than other races. But the point is that this problem affects everybody. And I think even more so today than it did say twenty years ago, because so many more people are falling into the gig economy, non-traditional credit, non-traditional income, world reality, right? it’s just, the times are not so cut and dry anymore. Life is gonna work way more harder, way more complicated than it used to be. And less lot life is a lot less predictable than it used to be. Okay, this is the reality today of you know income and credit. Okay, so that means is because life is less predictable, that means your paperwork is more complicated. And if your paperwork is more complicated, then there’s a greater human bias towards just not wanting to deal with it. Okay. and AI has no bias on that. It’ll it’s gonna deal with it the same way every time. Right? Whether it’s one bank statement or twenty bank statements, whether it’s whether it’s one tax return or twenty K1s. Right? it doesn’t care. And what I’m seeing these days is very common is we’ll see people having lower income people having complex tax returns that would normally you would see with higher income people like multiple K1s and various deductions and so forth. Okay and if you’re a bank and you’re lending somebody you know five million dollars and to lend the five million dollars you have to deal with 10 K1s and rental properties and all this kind of stuff. You’re like, okay, that’s worth the investment. I’ll put in the time to read through all those documents and analyze those tax returns. But are you gonna spend the same amount of time for somebody who needs a hundred and fifty thousand dollar loan? Right? Where
[David] Answer is probably no.
[Pavan] Yeah, where the numbers are smaller, just you know, a guy could have two or three different part-time gigs and different, you know, partnerships and whatever, and he’s making a legitimate living. Okay, and he’s earning real money and is working really hard, but he’s not making the millions of dollars that the high-end borrower is making. So, but that means the loan amounts are small, and as you know, smaller loan amounts means less income per loan, right? And less income per loan means less money to spend on underwriting development. Right. So you’re more likely to decline that borrower.
[David] Yep. That makes total sense. Not that it goes full circle out back out to the economy, what people are making, what they’re doing, and whether or not they have hope for a brighter future. What other things in this paper do you cover, Pavan?
[Pavan] What’s very interesting if you skip down to page 10. so the first half of the paper goes into a lot of detail of and the support behind this thesis and how and explains that in in detail the methodology
[David] This is here’s page ten.
[Pavan] Yeah, there you go, first ten pages really goes into supporting this thesis and the deep research that we did to come to this conclusion. On page 10, then we say, okay, if the whole country used angel AI and if the decline rates actually fell to the angel AI decline rates across the board, what impact would that have on the economy? Okay, and so then from page 10 on we do the analysis of that based on publicly available data, based on the impact that You know, we’re we’re taking, you know, from thirty percent, thirty five percent to, you know, four percent, ten percent, right? That’s a lot. That’s hundreds of thousands of borrowers. You know, I think I can I think the calculation is one point one, one point two million borrowers that were declined that should have been approved. Right. Think about it. It’s another like
[David] Right. Think about what that does in the front of billions of dollars that could get released.
[Pavan] Exactly. And as loan originators and real estate agents watching this, think about just another million borrowers that could have been homeowners. What does that do to housing prices? What does that do to your income, personal income? Right? and all the builders that are watching this, how many how many borrowers, if only one out of eight Asian borrowers that are declined should have been declined. Think about how many more houses you could have sold.
[David] So, Pavan, if people want to get this, first of all, we’re going to make it available with your permission. We’ll make this available to our listeners on our website. so that’s number one listeners. He’s nodding, so that’s affirmative. We’re gonna have this available, this website. Are there other white papers that you would encourage? I mean, all your white papers are very good. Some are more technical in nature, but the reality is there’s other really good white papers out there on your website.
[Pavan] Yes. Yes. So two things I want to say about that. Number one, the bottom line on the GDP analysis that we did is as much as a quarter point in national GDP.
[David] Quarter point.
[Pavan] Yeah, that’s massive. That’s massive. I mean that that you just you just think about how much employment that that creates and what it does. Quarter point increases national GDP, what it actually reduces our debt and so forth. So massive. Since we’re running at about 105% debt to GDP ratio right now, quarter point would change everything. So that’s number one. Number two is to answer your question, where to find the white paper? Just go to angelai.com/white paper. And you’ll see this white paper is on will be on top. It’s the most recent. It’s published in chronological order. Most recent white papers on top and all our previous white papers and academic research papers are all there.
[David] If you’re listening to this and you’re going to it by the computer, you’ll find it there. If you’re clicking on a link to listen to this, you will also find it in the pro in our podcast description with a link to this article. Very good stuff, Pavan. Again, I applaud you for what you’re doing to bring about really healthy lending and leveling the playing field, getting rid of this disparity that is so prevalent in our nation, which is You know, you get a bunch of people, they get tired, especially a month in, there’s decisions being made. They’re not the right decisions. And I mean, I granted I give everyone a hall pass from a standpoint we’re all human, make mistakes. But when it’s this consistently off by this much, then I’m I get really concerned, and I’m going up here scrolling off to these percentages right here. When we go from thirty three point three percent for whites down to nine point or twenty-four, nine point two four versus the most shocking one for me is the what’s going on with the Asians. And I thought by and large they were getting pretty much run a rub rubber stamp through. But we’re at a thirty-two point nine eight percent declination rate, declining rate being turned down, versus a four percent if they were in that same loan with the same data in the file. So when you start looking at the HMDA data for your and you’re not happy with the results, you’re looking at the turndowns that are going on, consider Angel AI. Angel AI is a solution to a higher declination rate, and that’s found money for you, folks. This is it’s found money.
[Pavan] Yeah. That’s found money. That that’s true. Imagine basically, you know, on average, if you look at this, out of every four declines, right? you know, well you know, three of those should have been approved. That’s what this data is saying, on roughly between, you know, two or three of those should have been approved.
[David] It is so important, Pavan, that people learn more about this, start taking more advantage of it. How can they learn more? Angelai.com. I know that’s where this is. We’ll also have some of this. What’s the white paper in the show notes? But they could go there, but also get some lot of testimonials. And if people want to learn more by talking to the appropriate people within Angel AI, what’s the best place to do so? Again, the website, I’m assuming.
[Pavan] Yeah, just go to the website and then if you really if you want to have a conversation with one of our experts, there’s a link right on top or you can request a Zoom meeting.
[David] Perfect.
[Pavan] So you click that link and you request a Zoom meeting anytime you want. It’s on your schedule. You say I want a Zoom meeting today at two o’clock, you’ll get it today at two o’clock. And one of our experts will show up over Zoom and talk you through it.
[David] Amazing. I couldn’t believe it when I first ran across this system, the claims, but I have now spent enough time listening to the firsthand at these meetings, these mastermind meetings, mindset meetings, and just hanging around Pavan when we’re at some of these events and I’m hearing it consistently. I encourage you to check it out.
[Pavan] Yeah, and for the lenders that are watching this, I mean, think about it yourself. Is is you know, times are tough right now, right? Rates are higher, production volumes are down. And if you were to say, if you declined a hundred loans last month, potentially 75 of those could have been approved. Right?
[David] Yes. Yeah.
[Pavan] Anywhere, anywhere from, you know, 60 to 75 of those could have been approved, think about what that does to your bottom line. That could literally take you from being in the red to being in the black.
[David] Right. It would for many lenders. Yeah. Good stuff, Pavan. Thanks so much for coming on with another AI update. Appreciate you and what you’re doing with Angel AI. Appreciate
[Pavan] Cheers. Thanks, David.
[David] Appreciate it.
Important Links
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.