20 Minute AI Appraisals Are Coming

[George Hatch]

All I can say is my personal experience with AI so far is that it is wildly over hyped - it is a product and the developers want to make $. They've integrated it by now into many areas, and I have seen no improvement and, in some cases, adverse effects.

AI comes out with the first search result in a google or other online search. It sums up things in a bland way. Big freaking dwal. One or two clicks would bring up the same info, and to find anything in-depth or anything controversial, you have to keep searching anyway.

It does compose a short email well, I have heard, and it can auto-respond to an email - fine, but again, is that earth-shattering? I mean, like big deal.

Creative: In novels or creative writing or poetry, it plagiarizes and does not originate. The writing is stilted and predictable, and the plots are rehashed and generic.

Art: Computer art, for the most part, is boring and insipid once one gets past the brilliant colors or photo shop tricks. Not emotional or mysterois or detoth of would like authentic artt.

Customer service _ AI now answers the phone at some businesses and announces itself as the friendly AI agent. It loops in a circle, can not answer anything other than a programmed question and will not connect to a person, who is the only one who can answer anything but a rote question

Spell check: It inserts the wrong word that is spelled correctly but out of context. It will not spell check any controversial word and either deleted it or puts in a substitute ..

I have heard it can be utilized well in science and medicine. I wish we could restrict it to those fields and let everything else alone

I have heard AI can sync to run a factory in China. Good for China, I suppose.. Am sure it will develop further in its applications for business. Which won't make our lives better, just more anxiety filled as people will have existential dread about whether their job will be there tomorrow and those who have jobs having to work at warp speed and max "efficiency" I am starting to hate the word Effiency - it means a reduction in pay and a reduction in quality of life - the only ones it benefits are the top tier owners or large share holders.

Forget where they are now because their development is going to continue. Over time they're going to become more refined.

Consider instead the underlying fundamentals -

• Who (what) is doing what?
• What is the utility of that what in the marketplace?
• How close to what an appraiser is doing does an AVM have to get to be considered by the users to be sufficient to purpose?

refer to tagline below:

 

[J Grant]

If that is truly your experience, then: (1) you haven't really much experience with it, or (2) you're intentionally lying.

I am not lying.
Those are my experiences with it. It spell checks from my Microsoft whether I want it to or not - the weird poss I make lately are due to its crazy spell checking and word sutistou.

Besides reading about it, I have tested it in creative writing, and that is my opinion on computer-generated art. AI answers the phone at some businesses I use and that is my experience with it as customer service. I do find it helpful with email, as explained. What experience do you have with it?

I am aware it can beat people at chess since it analyzes thousands of moves and learned from it. But again, who cares ? The drama is watching people play and innovate moves and responses, not some data-crunching robot - or maybe that will be the future, data-crunching robots and all human heart, soul, history, mistakes,, and triumphs just more data for it to spew back into a void. Wonderful!

 

[J Grant]

Forget where they are now because their development is going to continue. Over time they're going to become more refined.

Consider instead the underlying fundamentals -

• Who (what) is doing what?
• What is the utility of that what in the marketplace?
• How close to what an appraiser is doing does an AVM have to get to be considered by the users to be sufficient to purpose?

refer to tagline below:

Idk the answer.

Right now an AVM and an appraisal are two different products, IMO they are both suitable for either different or for complimentary uses and the ideal product might contain both.


Any AVM as a computer product is subject to the advantages (speed, consistency, large data sets and accuracy wrt math or spelling ) and, at the same time, vulnerable to the weakness ( no oversight, or agenda-driven oversight, hacking, unpredictable results, "dumb" mistakes in concept or context )

 

[George Hatch]

The way I would characterize it is that we obviously see the differences but the consumers (our target market) will be making their comparisons based on the utility-to-them. Not upon what is/isn't an appraisal.

 

[Mark K]

In the final analysis and when the next real estate crash occurs, does it really matter if an appraisal was completed with natural or artificial intelligence?

There was no AI during the last crash and the banks still lost hundreds of $ BILLIONS. Appraisals completed by humans did nothing to prevent the losses. I'm doubtful if robot generated appraisals can do worse.

 

[CGinMN]

A 20-minute AI-assisted appraisal only has to be good enough to pass underwriting. Right now, there are humans pumping 6/day without AI that pass underwriting. Soo it only has to be as effectvie as the worst appraisers. This is a very low bar, which it will pass and likely far exceed.

Think about how long it actually takes you to get within 5% of the estimated market value by flipping through some comps. That's probably how long an appraisal will take in the future with the AI doing all the administrative work. I'm not saying that this hypothetical "appraisal" will be credible or UPSPS compliant, but it'll be good enough for them.

 

[djd09]

but but but....what about the highest and best use analysis...no need now

 

[CGinMN]

Users don’t care, checkbox and boilerplate is enough.

 

[J Grant]

A 20-minute AI-assisted appraisal only has to be good enough to pass underwriting. Right now, there are humans pumping 6/day without AI that pass underwriting. Soo it only has to be as effectvie as the worst appraisers. This is a very low bar, which it will pass and likely far exceed.

Think about how long it actually takes you to get within 5% of the estimated market value by flipping through some comps. That's probably how long an appraisal will take in the future with the AI doing all the administrative work. I'm not saying that this hypothetical "appraisal" will be credible or UPSPS compliant, but it'll be good enough for them.

Will it be good enough for them?

IDK.
The emphasis on speed is comical - it is not a timed Olympic footrace. So what if the values are spit out fast (which an AVM does now ). There is no human with a license on the line and accountable, and the bar can go way below the worst appraisal ( and the worst appraisals and appraisers should not be used anyway) . These loans are for 15 years or 30 years and the occupancy for years, and a wrong lending decision or purchase decision can wipe out equity and savings and ruin finances - to save a few days on a valuation delivery - humans are pretty dumb after all .

It will not increase profits ot lenders, there are a finite number of loans given in any year no matter what.

 

[J Grant]

Users don’t care, checkbox and boilerplate is enough.

How do you know that to make such an assumption? Are you CEO of a bank or a regulator?