Bytes & Bricks—Tech-Trend Updates on Artificial Intelligence

Hello LCAR Real Estate Adventurers! This winter & spring I’ve had the distinct pleasure of seeing many of you in the new two-part CE course on Artificial Intelligence and Real Estate. I’ll tell you a secret, from the moment I agreed to write these AI courses for LCAR I knew they faced one huge weakness – they would be out of date by the time the class aired.

The first round was being edited right up until a few nights before class due to everything changing, then the advanced class got some last-minute changes.  Not only did the chatbot formerly known as Bard get rebranded as Gemini, but we had a big release - with the CoPilot AI integration for all users on the Microsoft platforms. Users are being given free previews before being offered subscriptions.  

Suffice it to say the updates and changes keep coming! Since we can’t have a fresh AI course and bring everyone into LCAR every month, here’s your quick rundown on what’s new working with AI in the Real Estate world and what’s gaining traction ~

On the regulatory front, The Appraisal Foundation has released an official Q&A that answers “What is an appraiser’s USPAP obligations when using artificial intelligence (AI) in an appraisal assignment?” The Foundation offers guidance on the reliance of credible data and the quality of the conclusions that are drawn from this data. The Foundation also cautions with the Competency rule on its use, and the Ethics rule for its output. The full PDF is available from their website.

Keeping up with the pace of progress these past few months ChatGPT has remained the #1chatbot, and at the latest count the website is getting close to 2 billion web visits monthly, which is about 5 x that of the second most popular chatbot Gemini (formerly known as Bard). Quill Bot, whom I introduced in the Real Estate LCAR class has remained in the top 5, which is no easy feat when you are up against research copilots like Claude and some new “uncensored” AI bots which are rising in popularity.

One of the bigger changes to ChatGPT is that plug-ins are being phased out. As the developer library of what we needed the bot to do, these “extras” grew to over 800 and had grown difficult to find the exact bot upskill you needed - and then there were so many good upskills that it had just become overwhelming. In class I spoke about the Redfin and Zillow plugins that had been used to pull property listings, but those specific plug-ins have been canceled directly, and recently the past conversations that included them have been wiped. Some of you have reached out to me directly, and I wanted to share with everyone that these specific real estate plugins (Redfin and Zillow) were removed by OpenAI due to difficulties of being able to control against guardrails of fair housing discrimination laws.  Basically, what that means is to do data analysis right now, you need to upload your own data from an excel file.

As the plugs-ins are phased out there has been a tidal wave of customGPTs replacing them. Many of them have built-in synchronicity with other apps. One of my favorite new customGPTs on the scene is the Spotify GPT - you tell the bot what your music interests are, how long you need the playlist to be, maybe a “theme”, and the GPT will create a custom Spotify playlist just for you. Another one that has been helpful for me is a customGPT called MyVisionBoardArtist that makes images with the DALL-e image generator which are suited for a mood board. A mood or vision board is one of the top 2024 trends for business management. Where do you see yourself going this year? What specific goals have you set for yourself? Create a visual of it and add it to your 2024 Vision Board.

For those of you who attended the introductory AI course and were introduced to Claude, this bot has now been upgraded to read images, and the company Anthropic has released a massive prompt library with several popular categories to inspire people and get them going with the bot. This of course comes with a tiered subscription package for that value added bump over the free version. The free version will still be there to analyze documents as large as a harry potter book – and answer questions, summarize, analyze the documents all without a subscription.

While Gemini (the Google bot formerly known as Bard) has tried to join the image generation game and had a failed start, Meta (formerly known as Facebook) still has a low-level image model turnout, CoPilot has kept their image tool running, but it’s poor quality and doesn’t handle text all that well. As mentioned, Claude now at least reads images, Perplexity will return images to you in a way that feels like a very tailored Google image search – but no one is matching Midjourney’s image generation. The image at the top of this article is made with Midjourney with the following prompt: