RCA
Elite Member
- Joined
- Jun 27, 2017
- Professional Status
- Certified General Appraiser
- State
- California
Well, you are overdoing it perhaps.
First, we need to stop using the term "AI" all by itself. It has to be qualified by what in "AI" you are talking about. If you don't define what you mean by AI, then you are asking for problems.
1. "AI" could just mean "Large Language Models" or "LLM's". These are programs that parse informations such as text, photos, video and audio to develop the ability to predict "what comes next" through training. Ilya Sutskever was one of the first to notice that traditional LLMs based on transformers have more or less reached their peak in intelligence. They are plateuing. And that is confirmed by Grok-4. The newer models are adding (a) Reasoning Engines, (b) Calculators, (c) Coding Engines, and so on to improve their intelligence. These additions create what are sometimes referred to as LLM-hybrids.
2. So "AI" could mean LLM-Hybrids such as Grok-4.
3. But it could also refer to other tools which use other kinds of neural networks than transformers, such as Convolutional Neural Networks (CNNs), such as ResNet developed by Microsoft, which forms the backbone of many other AI systems such as TensorFlow, Caffe, YOLOv8, OpenCV and PyTorch. Indeed, in newer CNNs we now see hybrids that also include Transformers in the form of Vision Transformers.
4. And in addition, there are other systems that somewhat similar methods and methodologies as neural networks, such as MARS regression - and are also considered AI. Even advanced reasoning systems are considered AI. Prolog is considered to be AI.
5. In fact, its hard to draw the line between computer technology in general and AI. Grok-4 gave me this definition of Aritificial Intelligence today: "Artificial Intelligence (AI) is the development of computer systems that can perform tasks typically requiring human intelligence, such as learning, problem-solving, decision-making, and perception, by processing data and adapting through algorithms and models."
====
1. I don't nor ever have "audited appraisal narratives for bias or manipulation".
2. I wouldn't say I "cross-reference market trends with real-time data", because I think that implies you are finding real-time data to support your opinion of market trends. NO. I would say I develop market trends FROM real-time data (and actually that is also not true because as Grok-4 states, real-time data is: "Real-time data refers to information that is collected, processed, and made available for use immediately or with minimal delay, typically within milliseconds to seconds, enabling timely decision-making and dynamic system responses.")
3. I don't know what you mean by "challenging regulatory assumptions using Grok's raw output." But certainly that should be very possible through a wide variety of methods.
My approach is simply fact, logic and reasoning. Sometimes I am stating my opinion without proof: For example, IMO, good appraisal protocols would largely obviate the. need for policing intention. Either you follow the protocols or not. If you believe this, then an appraiser should be able to create a certified appraisal report for himself, relatives or the richest person in the world, - and an AI system should be able to verify its accuracy with an audit, aside from the physical measurements by some person or robot.
