Generative AI: What a Difference Two Years Has Made!

ChatGPT, a generative artificial intelligence (AI) platform, was launched in November 2022 and had one million users in one week and 100 million in 2 months. It was the fastest tech platform ever to reach 100 million users and, by February 2023, I was hooked. Today, about 1 in 5 workers are using ChatGPT for their work, myself included.

 

A lot has happened with generative AI in the three years since its launch, so I thought I’d share some personal Barbservations:

AI Use is Not New- Even before ChatGPT came on the scene, I had used AI in the following ways: Microsoft Word and cell phone texting that anticipated (sometimes incorrectly) the next word that I would use in a sentence, pop-up ads based on online searches, shopping suggestions based on past purchases, and Alexa and other virtual assistants.

 

It Makes Suggestions- In 2023, I told ChatGPT the output that I wanted via a detailed prompt. It then did what I asked, but no more. I was done. Today, like an overly ambitious student trying to get an “A” from a professor, it asks me if it can do more for me such as creating a checklist or PDF documents.

 

AI Output Pops Up Automatically- Three years ago, I had to seek out a generative AI content by proactively logging in to ChatGPT. Today, if I use the search engine Google, content from their AI platform, Gemini, generally pops up first to answer my question, along with the information sources that were used to generate the AI answer.

 

Text to Graphics Capabilities- In its early years, ChatGPT could only be used for text prompts for text output. I tried to create images and got an error message. Today, you can give ChatGPT a text command to create a graphic image. My experience is that it takes a “conversation” with several rounds of prompts to get exactly what you want.

 

Hallucinations Still Occur- Hallucinations are authoritatively sounding AI-generated statements that are wrong. They were a problem in 2022 and continue today. Two prompts that frequently cause problems are math calculations and reference citations. Thus, AI-generated content must always be considered a “first draft” subject to review and revisions, if needed.

 

Content Controls- Since hallucinations are an issue, some organizations have created their own data sources, called large learning models, for AI to generate content from. LLMs are the underlying technology that powers generative AI applications that are trained on a massive amount of text to understand, generate, and process human language. I recently experimented with this and asked ChatGPT to create content only from my website and archived blog posts.

 

My best advice for using AI is to play around with it and have fun. It will probably surprise you and may help you accomplish everyday tasks such as planning a meal and trip planning.

This post provides general personal finance or consumer decision-making information and does not address all the variables that apply to an individual’s unique situation. It does not endorse specific products or services and should not be construed as legal or financial advice. If professional assistance is required, the services of a competent professional should be sought.