핵심 요약
AI는 뛰어난 수학 연산 능력을 갖추고 있지만, 보고서나 제안서 작성 같은 큰 작업 안에서는 텍스트 생성에 집중하느라 숫자를 틀리기 쉽다. 이는 AI가 단어 예측 방식으로 숫자를 '추측'하기 때문에 발생하며, 정확성을 위해선 연산 전용 경로로 유도해야 한다. 중요한 숫자를 다룰 때는 AI가 글 작성을 멈추고 코드나 도구를 실행해 직접 계산하도록 전환하는 과정이 필수적이다.
주요 포인트
- **수학 능력과 문서 내 계산 오류의 괴리**: AI가 복잡한 수학 문제를 풀 수 있어도 보고서의 합계를 틀리는 이유는 숫자가 문서 맥락 속에 묻혀 연산 모듈에 도달하지 못하기 때문이다.
- **AI의 두 가지 숫자 처리 경로**:
- **코드 실행 경로 (정확함)**: 글 작성을 멈추고 코드를 작성·실행하여 오차 없이 정확한 값을 산출하는 방식.
- **단어 예측 경로 (오류 발생)**: 텍스트 흐름에 따라 다음 단어를 예측하듯 숫자를 어림짐작해 출력하는 방식.
- **오류의 근본 원인**: AI가 전체 문맥 작성에 초점을 분산하면서 멈추지 않고 텍스트 예측 방식으로 숫자를 끼워 맞추기 때문에 계산 실수가 발생한다.
- **해결 원칙**: 중요한 수치 작업 시 AI가 단순 텍스트 생성을 중단하고 코드 작성 및 도구 활용을 통해 연산하도록 프롬프트와 작업 흐름을 설정해야 한다.
AI can win a gold medal in math and still give you the wrong total in a report. That sounds impossible, but it's exactly what's happening. The problem isn't that AI can't calculate. It's that most of your numbers are buried inside of proposals, spreadsheets, and summaries. So, they never reach the part of the AI that does the actual mathematics. If you're new here, I'm Dylan. I run an AI consultancy, and I see this constantly in the real businesses I work with day in and day out. So, in this video I want to show you the two paths every AI number takes, the one that calculates and the one that guesses, and the simple 2-minute check that makes sure your important numbers end up in the right one. It only takes a couple of minutes to set up, and once it's there, every report, proposal, and summary gets the same protection automatically. So, let's get into it. Now, the question we want to ask ourselves is why can such a smart AI make such a simple mistake? Well, often what happens is the AI is trying to do the math within a larger task. So, it's likely working on something like a proposal or some form of market research or a summary, and there are a series of numbers inside of that document or whatever that artifact it is creating for you. So, its focus is kind of distributed amongst different things, and it's not focused on the math itself. And that's the [snorts] thing that we want to try to solve for is altering that focus when we need to to make sure the numbers are correct every single time. Because when we think about numbers and how AI interacts with numbers, there are two paths numbers can take in an AI's head. This one here, the first one, is one we want it to take, where the AI stops writing cuz when the AI writes, it's predicting basically the next word. That's how AI tends to function. We want the AI to pause, take a step back, and instead of continually predicting words and thinking about how to guess the next thing, we want it to stop and write code, and then use that code to then calculate the number for us, ensuring that it's correct every single time. So, we're using that brain power towards writing code to make sure the numbers are accurate. The second path, and likely the path you're taking every single time AI interacts with numbers inside of a larger task, is the AI just keeps writing. It doesn't pause and write code. It just keeps trying to predict and then tries to predict your number based on whatever work it's been doing historically. And that then feeds into a guest answer. Sometimes it's right, but sometimes it's wrong. And oftentimes when it's wrong, that impacts us in different ways. So, the name of the game here is refocusing the AI in the moment that it matters to write code or use tools to make sure the numbers are correct. And the primary rule I want you to think about going forward is when you have an AI doing a task that has numbers inside of it that matter, that need to be accurate, you want to make sure that math itself has its own step. And there are two different problems we're trying to solve when it comes to numbers because the numbers come from two sources. One source is AI simply copying those numbers from one place to the next. So, it's extracting information from PowerPoints, Excel sheets, whatever else. It's pulling it out from there and putting it into another system for you. Another thing is calculations. So, the AI could be doing a calculations and analysis on the way from pulling it from one place and putting it into another. So, we need to solve for both of these issues and how AI interacts with those numbers. Quick pause. If you're enjoying this, you're going to enjoy two other things. First off, below is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, below are a series of offerings to see if there's a good fit between the two of us. Now, let's get back into the video. And there are three ways we're going to solve it. It's either going to be something that's a simple one-off, something that's reoccurring, or something that's high stakes. Each of these scenarios have different solutions. So, our first one is one-offs. So, this here is a prompt that we would use for a one-off task we're asking AI to do something. So, I recommend copying and pasting this prompt at the bottom or top of whatever you're doing to make sure the numbers are accurate. Now, what is this prompt actually doing? Well, we're telling the AI, "Before you give me a final version of whatever you're trying to create for me, so that's a report, presentation, proposal, whatever else, I want you to list out every single number inside of that output. After you've listed the numbers, I need you to recalculate any numbers that had calculations involved using code. Now, this here is a trigger word. We're triggering the AI in its head to take a different path. We're