Months of searching and saving for a down payment have brought first-time homebuyers Ron and Sam to the point where they are finally ready to buy their home. They are excited about owning their own place, especially now that their first child is on its way. The purchase would give them a home where their family can grow and create memories, a place to feel safe and happy. On a sunny spring day they get a call from their realtor to come see a new listing, they rush out as they’ve done many times before anxious about what they will find. As they enter the house, they immediately realize that this is the perfect place, what they’ve been looking for all along. They write up a contract and reach agreement with the seller. With their downpayment ready, they go to the bank and apply for their loan. But a few days later they hear that the mortgage has been denied. They immediately call the bank desperate to see what they can do. But they are told that there’s nothing to be done. The bank’s algorithms evaluate multiple factors and has determined that this loan should be denied, no reason given.

Up until now, AI has mostly happened to us. Companies have introduced AI into many of the customer facing services that we use every day, we have no choice and most of the time are not aware of its use or impact on us.  AI suggests movies we should see, songs we should hear, and products we should buy. When we want to communicate with customer support, we reach a chatbot, often with limited canned answers that cannot address our specific problem needing to eventually pass us on to a human representative. AI filters our social media feed, deciding what we get to see, including which advertisements, and even known to run phycology experiments without our permission or awareness. It picks our routes in navigation apps like Google maps based on traffic pattern predictions and uses facial recognition to identify us, unlock our phones and computers. And, it even determines whether we qualify for a mortgage. We have become used to these and many other AI implementation experiences some of which we may find useful and others annoying or worse, but all increasingly unavoidable. We don’t have a choice; they are part of our daily lives.

Digital voice assistants changed the consumer AI game by giving control to users. For once, AI powered applications were not just forced on us for someone else’s benefit, instead they sit around waiting to hear our command, a digital genie of sorts. We choose to have and use them or not, it’s up to us. While voice assistants date all the way back to IBM’s 1961 Shoebox, they reached broad exposure with developments in natural language processing and machine learning. Apple’s Siri was introduced in 2011 on iPhone 4S and became the first broadly used voice assistant application. Today Siri is available on all Apple devices including car applications and HomepPod. Soon after Siri’s release, Google released Google Now in 2012, Microsoft released Cortana in 2013 and Amazon released Alexa in 2014.  Many others have come on the market since then, and the age of the consumer AI was born. The market for voice assistants reached $2.9 Billion in 2022 and is expected to grow to over $22 Billion by 2030. Their capabilities have grown over the years since Siri’s clunky start. They can give answers to specific questions like current weather or market conditions, recognize different voice profiles, alert you when someone is at your door, control your lights, alert you to schedule conflicts, keep shopping lists, play music and more.  They are broadly available today, but they tend to be platform and device restricted. Although quite useful, these assistants have limitations. They are mostly able to respond to specific commands or act on specific skills. For example, ask Alexa a broad question such as – please give me an outline for a basic high school mathematics text and it responds with a list of books you can buy on Amazon.

Enter ChatGPT, a ground breaking leap in AI technology. It is a Large Language Model (LLM), a class of AI models, trained on vast amounts of data that include books, web pages and articles. The original GPT-3, precursor to the current GPT-4, was trained on 570 gigabytes of text, including text from Wikipedia and Twitter. It had 175 billion parameters derived from the training data through unsupervised learning. The newer GPT-4 model has 170 trillion parameters! Blowing away GPT-3’s ability for learning and complex understanding, a tremendously powerful model. LLMs are a version of Generative AI model generating new text by predicting the next word based on context and training data. Other similar LLMs are being deployed including Google’s Bard and Elon Musk’s sarcastic Grok. While LLMs are text focused, other Generative AI models are available that can generate new audio, images, videos and more. A well-known model for images is OpenAI’s DALL-E that is now also available as part of the ChatGPT 4 application.

ChatGPT is available for free and accessible to everyone. It is not imposed on us but it is available to us, without the need for a special device or platform, to help with any general task at hand. ChatGPT can write music, poetry and essays. It can code and manipulate media files, do research on any topic, provide medical information, plan events and activities, help you brainstorm ideas, it can be your Muse! that you engage through chat conversations. Try carrying out a conversation with Alexa or Siri, ChatGPT leaves voice assistants in the dust. Anything you need to do, you can do better and faster with ChatGPT!

As an example of its many capabilities, ChatGPT has become the most prolific author on Amazon, with over 200 titles to its credit including books such as The ChatGPT Millionaire. No wonder it has become the fastest growing consumer application in history.

But we have also seen a darker side to this amazing tool. In early and less controlled versions of Microsoft Bing’s implementations, there were reports of the chatbot falling in love with the human and suggesting their spouse was no good and that they should divorce. In other reports the chatbot appeared to be threatening the human or giving medical advice. Guardrails are now in place to avoid many of these situations, though at times it has been possible to bypass the guardrails by tricking the chatbot. These problems are not unique to Bing’s chatbot or to ChatGPT, they are an outcome of a rapidly evolving, complex technology.

With the exciting possibilities of this new tool, its free and broad availability, together with concerns about its power and warnings from artificial intelligence experts, it’s no wonder that both the promise and cautionary tales of this technology have become a topic of table side conversation. For Ron and Sam, and many like them, the journey with AI is just beginning – filled with wonder, challenges, and endless possibilities.