AI - The Prompt

Using AI in the real world today

For the past year or so AI has been making the news more and more. I'm not going to bother talking about that, if you are here, you already know it.

There has also been a fair amount of news about people losing their jobs, only we don't call it losing a job we call it job displacement, reminds me of Blade Runner where the replicants are "retired". You don't get fired anymore, you get displaced. I'll used fired in this post, it has fewer characters.

The truth is AI is going to have a profound impact on society for sure. Being optimistic I believe that it will benefit society rather than turning the world into some dystopian terminator nightmare. Let's hope I'm right ;-) The reality is that a whole lot of people are going to get fired. Let's see if we can prepare ourselves for the up-and-coming retirement apocalypse.

In these posts, I am going to give you what I think are the main things most people need to know right now to start getting to know AI.

For many people, AI means chatGPT, Dalle-3, Bing and Bard (recently renamed Gemini), and not much else. Developers also have Copilot a little tool to help us all write code. Microsoft is also rolling out a version of Copilot that will be in Word, Excel, PowerPoint and other Office tools.

First things first, one for the IT project managers, despite what you may have read, Copilot does not 10x or 100x your developer's code. They are not able to write a week's worth of code in half a day, it is not happening. Sure you can get it to speed things along, it adds most of my comments, and you can get it to generate code for you for real. But at the current time, you would have to be insane to take code generated by copilot and roll it into a production environment without someone checking it all first.

Same goes for SQL generated from natural language. While it may not matter much on a tiny database on someone's PC, you try and run a full table scan on a database table with a billion entries and the DBA will probably come looking for you with an axe.

Maybe this will be different a few years down the line, but not 2024, at the moment AI can help your developers, but it is not going to "retire" them just yet.

OK, so how do we make a start? chatGPT and Dalle-3, these are cool. Even if you are using the free version of chatGPT, it is a glimpse at the future. There are user-made GPTs on chatGPT that can be accessed through the chatGPT website. There are now 10s of thousands of these already and cover a wide range of subjects. Some are better than others. For most people though these don't tend to help you with your day job. At least not at the moment they don't.

For many people, you try chatGPT and it's OK, and sometimes it's pretty good, and sometimes it is really crap and a waste of your time.

First question, how much use is it in your day job now? For a lot of employees the answer is "Not much". Part of the reason for this is that people don't know how to use it properly. You have to learn how to ask the right question. The better the question, the better the answer. This is not as easy as you would think though.

In real life, when someone asks you a question, you might not understand and so you will ask for clarity. The person asking the question may be struggling to formulate the question and you might help them. In many discussions with project managers, it has sometimes taken a fair bit of back and forth just to get us on the same page. It does not work like this with chatGPT. You ask it a question and it will try to answer. Sometimes if it can't find a real answer, it will invent something. This is called an AI hallucination, if it doesn't have an answer, it will make some shit up. What is weird is that people do this sometimes, though we call it lying or bullshitting. I'll let the philosophers argue over that one.

Todays take away. If you want your people to start using chatGPT, or Microsoft copilot, do yourself and your staff a favour and get them some training on how to ask questions. Don't just expect people to be good at asking questions because historically we are not very good. We are often piss-poor. Like everything else, it is a skill that can be learned, practiced and improved upon.

The better you are at asking questions the more you will get out of the AI.

For any company out there who is hoping their staff is going to start making real use of AI and LLMs to 2x, 5x, or 10x their productivity. Teach them how to use the AI properly. If you don't they won't.

Bryan

For those interested GPT is a Generative Pre-trained Transformer, LLM - large language model.