Good question! We think about Artificial Intelligence (AI) differently than most companies. We don’t think it will save the world or kill us all. The people who say those things are the same people who invented them, and they make a lot of money on the idea that Artificial General Intelligence (AGI) is just over the horizon, and once it arrives all of your dreams will come true. Well, At SCL CLPTL, we think the future of AI is customized, not general.
AI is actually good at being creative. But it’s not very good at being a database.
At SCL CPTL we’re using AI to build brainstorming buddies, leadership coaches and HR specialists. Using the most advanced LLMs, we can use your data, brand voice and style guide to create assistants that help your team work smarter, not harder. Do you need to create a project plan? Your custom AI Assistant from SCL CPTL can create an outline (based on your data), help you set milestones and goals, and then schedule follow up meetings. Creating a new training for your team? You can ask your AI Assistant for feedback or to create a follow-up survey. Need to have an uncomfortable conversation with a coworker? Your AI Assistant can help you prepare. Hiring? Your AI Assistant can help you come up with interview questions and an onboarding plan.
If you read the news or watch movies, AI is always a super intelligent being that is simultaneously the best scientist in the world and also an amazing playwright. In media, AI is always anthropomorphized with human characteristics. Today’s Large Language Models (LLMs) do an amazing job of mimicking human thought and communication, even creativity. But would you ever expect one human to be good at everything? We wouldn’t. Which is why we provide customized AI Assistants that are good at what you want them to be good at because we don’t believe the hype that one AI can be good at literally everything.
So, what kind of things are AI Assistants really good at? Our Founder and CEO started SCL CPTL after almost a decade as a nonprofit leader. The first AI he created supported the nonprofit’s staff with grant writing, proposals, policies and procedures, brainstorming, etc. Tired of confronting constant challenges with staff capacity and limited resources, he set out to create a “digital employee” that the entire team could use to alleviate some of the administrative burden and focus on the nonprofit’s mission.
one-of-a-kind not one size fits all:
What is Artificial Intelligence actually good for?
We work with your team to identify your use case — what problem are you trying to solve?
Our AI Assistants are trained to write in your voice and style, so that your whole team writes with one voice and everything you publish is on brand. Integrating APIs from your favorite productivity tools allows us to create custom work-flows for your team including calendar integration, task creation, project management, etc.
Small businesses use our custom AI Assistants for: Business planning, team building, contracts, proposals, scheduling, project management, brainstorming, onboarding, etc.
Nonprofits use our custom AI Assistants for: Fundraising, grants, strategic planning, board management, staff development, proposals, brainstorming, project management, scheduling, etc.
Don’t worry, we’ll also train you and your team on AI best practices and key words for your assistant. Our subscription-based model means that we’re in it for the long haul and we continue to train and update your model based on your feedback using reinforcement learning. We also offer training options where we help you build your own AI Assistant and train you on how to host and maintain it.
So, AI isn’t good at data recall but is good at data creation? Exactly!
It’s important to remember that today’s most advanced AI’s are Large Language Models (LLMs), which means that they are trained on massive amounts of text and work by using advanced math to identify patterns in human writing and data to predict the most likely combination of words in response to a prompt. Even AI that isn’t language based — image, video, sound generators — work by predicting the most likely next step in the pattern. We see this, for example, in today’s incredible Nvidia chips that create amazing digital animation for video games using advanced AI. Even robotics and “AI vision” work through prediction.
When a machine is predicting the most likely data from the information it’s trained on, it hallucinates… trippy 😵💫. In other words, it’s creating the most likely answer, not necessarily the right answer. AI is really good at processing a lot of data and coming up with creative solutions based on that data.
Okay… But I want to access real-time data. Us too! We can integrate your AI Assistant with your favorite web tools using Application Programming Interfaces (APIs). Some people are calling this “agentic AI”, meaning that the AI can interact with its environment autonomously. But we think this is a little bit misleading. When you’re using an API integrated with an LLM, the AI basically become a User Interface (UI). So you’re using an AI LLM as a UI through an API… whoa, that’s a lot of acronyms (don’t worry, we remember them so you don’t have to). For example, your AI Assistant could access your calendar and find a time for you to meet with your colleague using a Google Calendar or Microsoft Teams API. You can create a project plan outline using the AI Assistant’s integrated LLM, and then create tasks using Asana or Slack APIs.
So, why are APIs better for accessing accurate, real-time data? Because they are accessing the external data directly, through the “interface” part of the API — one application talking to another — instead of predicting what the data will be. It’s important to understand that artificial intelligence doesn’t “know” or “remember” anything. It predicts everything based on the most likely outcome. Sometimes the most likely outcomes isn’t the most accurate. APIs don’t remember anything or “think” either. They follow a very specific set of coded commands to interface with external applications directly, no prediction involved.
The way AI processes information is similar to natural evolution: The most likely outcomes of reproduction are the most common, but the most likely outcome is also determined by the highest survival rate. Even so, there are both good (adaptive) and bad (pathologic) deviations from the most likely outcome. Similarly, AI that has seen the answer to a particular question a million times, will probably get the answer right almost always. But when asked an outlier question, it will give an outlier answer, which is more likely to be incorrect because it has less information to pull from. The more times you ask the question and then reinforce the correct answer by feeding it back to the AI (surviving longer due to adaptive attributes), the more likely the AI is to get the answer right. There is no way for the AI to be able to correctly predict every possible answer, especially if being asked about a very limited dataset, such as company financials or evaluation results.
Did we stretch the metaphor too far? Maybe…
Still have questions? Get in touch with us.