What is AI?

AI has different definitions, and the one that is relevant for AI today is when technology enables machines to perform human tasks. An example I like to use is the self-driving car.

While we are driving a car we are watching the traffic lights, reading the road signs, tracking the cars around us and looking at the road ahead of us just to name a few tasks. This is all happening in the background while we listen to music. How does a car drive itself; how does technology perform these same tasks well enough to get us from point A to point B autonomously and safely? Two components that make it possible are Machine Learning Models or Models and AI.

A Model can classify or predict. Take the traffic light. If we put a camera on the car and send the video to a Model that can classify whether the light is red, yellow or green, the model might return the traffic light is 85% red, 5% yellow and 10% green. Our car could then know the traffic light is likely red. The key word is likely. Another Model can use a camera to determine if there are obstacles in front of the car by returning "obstacle” or “no obstacle”. Another Model uses a vibration sensor to predict if the car is traveling too fast. You can break down all the actions of a driver and make many Models that align with their decision making. These Models run continuously classifying and attempting to predict the world around it. Each model on its own is useful, but the real value comes when they are all used together in a meaningful way.

That is AI. AI uses the information from these models to drive the car. AI can determine that the light is green, and we have gas in the car, but there is an obstacle in the road that is just in front of us so we should brake hard to not crash. Models use data to create information. AI uses information to take action.

What Changed?

Around 2022/2023, two types of models made giant leaps. Models started to create text and images really well. This creation of data is known as generation. These models are referred to as Generative Models. Before this time Models were mostly used to classify and forecast. Now a Model could take a bunch of words and turn it into something new. If Models can generate text, then Models can write emails or software code. Even though the debate about how well Models can do this continues, there is consensus they are improving and will continue to improve at a rapid rate.

Generative AI, also known as GenAI, is when we use Models to Generate Data, and its adoption is faster than anything seen before, including the Internet and Computers. As GenAI is used by individuals, corporations, and governments at such a rapid rate every day, we are learning what works well and what fails.

Chatbots and Agents

As GenAI became mainstream the average person refers to GenAI simply as AI. The use of this AI was Chatbots starting with ChatGPT. ChatGPT was able to reach one million users in 5 days compared to months for Instagram or Spotify. People from around the world were enthralled by AI's ability to answer questions, write poems, summarize emails, and what seemed like having a conversation. Unlike prior Chatbots with fixed pre-determined responses, the Chatbots were varied and nuanced.

A Chatbot is turned based between users, humans and the assistant, the AI. We say hello, and the AI predicts "hi" as the best response. Our input is known as the prompt and the AIs response to the completion. This is a great tool for service representatives and combined with text to speech call centers. One obstacle that appeared as these new AI driven Chatbots started to be used to solve real problems for organizations was their lack of current data and private information. The industry created new technologies known as tools and retrieval augmented generation to solve these. Now a new field of prompt engineering is emerging.

Agents are software programs that use AI to complete tasks to reach a goal. A travel agent is a good example we can relate to. If you are using Chatbot, you might say I want to go on vacation this year. Chatbot asks where would you like to go what do you want to do when you get there. And taking turns with the Chatbot you ask questions and get answers going back and forth making your vacation plans. When you are ready and have your plans then you would go and book your travel. With an Agent you would state your vacation plans and then the agent would not only research all of the travel plans but also book the trip. Where Chatbots are great tools to get information, Agents take action on your behalf. Unlike traditional software applications that are very predictable, an Agent's reliance on AI gives it unpredictability which is a risk that needs to be assessed.

The Value

Some say the value of Generative AI is found in the action its knowledge enables. As incredible commoditization of knowledge by AI is, it lacks action. As an organization successfully enabling AI to use your knowledge to take actions that result in getting more done in less time and less resources is where the value it. This can be as simple as an asset manager using a chatbot to finally create that excel function for the complex calculation that used to take them days every month. If can be the developments team ability to take that quarterly process and write new code and refactor existing code so it runs with less human involvement in less time and less errors. It is the future state where organizations are exchanging information and transacting with each other on the Internet's new agentic layer.

Smart Coins AI helps you with the right strategy to navigate this space and implement the tactics so you can show the measurable results of AI adoption in your organization. That is the actual value 1. the vision to adopt AI 2. the ability to execute on that vision and 3. measurable results showing outcomes over activities delivering meaningful results.

Let's Get Started

At Smart Coins AI we want to hear your story and help your organization adopt AI easily and in the right way. Get in touch to start the conversation.