Trends 2024 "The Promise -- and Limitations -- of Generative AI"
Illustration by Lanette Behiry/Real Estate News

Trends 2024: The explosion of generative AI 

The public release of ChatGPT just over a year ago put a powerful AI tool in the hands of everyday users. Where will it go from here?

January 8, 2024
4 minutes

Editor's note: Since 2006, the Swanepoel Trends Report has provided in-depth research and analysis to help leaders understand the forces shaping residential real estate. This exclusive series of excerpts highlights each trend featured in the 2024 report, which was released in November 2023.

The Promise — and Limitations — of Generative AI: Before ChatGPT burst onto the scene in late 2022, few people outside of the tech world were familiar with the concept of generative AI, which can create new content in ways that are simple and more "human" than ever. Now that anyone can use and adapt this technology, how will it impact real estate and other industries? Is it the wave of the future — or a passing fad? 

This excerpt outlines the key differences between traditional and generative AI with a focus on the flexibility of generative AI technology.

Generative AI in context

Generative AI — an extension of standard AI that powers tools with expansive, broad applications — has been around for years, but its popularity exploded when AI startup OpenAI released ChatGPT on November 30, 2022. The tool immediately captured the world's attention and inspired companies to accelerate the development of their own generative AI tools and for others to begin fleshing out existing AI tools with generative features.

Traditional vs. generative AI

The modern AI systems that power much of the technology consumers use every day make estimations or guesses based on data they ingest and process using machine learning, a feature in which a system can perform complex tasks that can augment or supplant human judgment.

This capability of AI has two primary applications:

  • Skill approximation: In which an AI system performs tasks for users or tries to emulate human skill (e.g., chat with a prospective homebuyer)

  • Prediction: In which an AI tool makes an estimate about the future (e.g., predict the sales price of a house)

Generative AI represents an extension of traditional AI systems and is trained specifically to make predictions. The key differences between traditional and generative AI systems cover:

  • Task Scope

  • Training Data Generality

  • Accessibility

Task scope

A distinctive quality of traditional AI systems is that they are trained and optimized for a single task, such as face recognition or automated valuations. They do not have the capability to perform any task other than the one they were trained and designed for. 

To change what they want the AI system to do, even slightly, developers must retool the system by restarting the training process from scratch with different data, and, often, a different algorithm.

Training data generality

Generative AI systems, on the other hand, are often trained on large, diverse, general knowledge bases (e.g., a snapshot of the Web, all of Wikipedia) without a specific task in mind. This gives them a knowledge and skill base rooted in a countless number of domains. 

This generality is one of generative AI's distinctive features. Tools leveraging generative AI can be used to answer questions about general knowledge, create images of spaceships, translate text into other languages, write computer code, and many, many other things — all from a single AI model, no new data or retraining needed.


Another important distinction between standard AI and generative AI systems is ease of use. Because standard AI systems are trained for a specific task, they usually require a specific format and quality of input (e.g., purchase records from a consumer for fraud detection). 

However, generative AI systems typically accept a free-form prompt as input and have an unbounded output, free of conditions. In this easy-to-use interface, users can simply type their request in non-technical language and receive a reasonable (but not necessarily accurate) response. 

This combination of general foundation, extreme versatility and high ease of use, give the impression that generative AI systems can do almost anything.

Read the full chapter: Digital and printed copies of the 2024 Swanepoel Trends Report are available for purchase at T3 Trends.

Note: T3 Sixty and Real Estate News share a founder, Stefan Swanepoel.

Get the latest real estate news delivered to your inbox.