“Oh Sh$!%”
- Anyone who has tried ChatGPT ever
Every day there's another story out about how ChatGPT and AI will replace hard working Americans. Elon Musk wants to pause the training of any system more powerful than GPT-4. Goldman Sachs estimates generative AI can lift world GDP by 7% (yay) and disrupt 300 million jobs (boo). To be honest, there's good reason for alarm. ChatGPT is AMAZING stuff. In fact, can you really tell that this post isn't written by ChatGPT? I promise you it isn't. But it’s not that I didn't try. I had this hope that I could just type in the title of this post, sit back and relax for 25 seconds, and voila! this post would be done. And though ChatGPT did write a grammatically correct post that addressed the topic at hand, it was a bit too... unbiased for my taste.
There was no spark, no loud and edgy viewpoint, and no "me" in the article. So AI can’t actually BE me, but can it replace my work? Can it put me, you, and everyone you know out of a job?
Without doubt, large language models such as Chat GPT will soon be able to replicate much of the body of writing out there, and possibly a lot of what we get paid to write. But us humans have a long history of fighting tooth and nail and never saying die. So how can you, me, all of us beat ChatGPT?
Don't try to beat AI. Hire it. Be its Boss.
I learned a long time ago, that in white collar work, you don't get paid for doing work, you get paid for work being done. In a capitalist society, your compensation is proportional to the amount of capital you are accountable for. It’s not HOW the work is done, but the quality of the work that is delivered. Is this article better if it was written in a prison cell in New York or in a New Age coffeeshop in Bali?
For many people, this is already the case. Contract companies get paid for results and often charge by project rather than hourly. With LLM’s (Large Language Models of which chatGPT is only one instance) quickly becoming a commodity service, contract companies stand to make a killing as their costs drop while the value their products create drop much more slowly (competition will eventually bring down prices).
You can do this too. Experiment with chatGPT and see how you can hire it (works for free or $20 a month at most) to help you with your job. But don’t just “play” around with it to see how good it is at sounding human, actually try giving it tasks to do. Here are some areas (that chatGPT came up with):
Research: ChatGPT can conduct research on a specific topic related to your job or industry. You can ask for insights on current market trends, customer behavior, or competitor analysis, among others.
Writing: ChatGPT can assist you in writing emails, memos, reports, and other documents. You can provide ChatGPT with specific guidelines and details, and it will produce grammatically correct and well-written text.
Proofreading and editing: ChatGPT can proofread and edit your documents, ensuring that they are free of grammar, spelling, and punctuation errors.
Data analysis: ChatGPT can help you analyze and visualize data, making it easier for you to interpret and understand complex data sets.
Personalized recommendations: Based on your job role and responsibilities, ChatGPT can provide you with customized suggestions and recommendations that can help you be more productive and efficient.
Translation: ChatGPT can translate documents, emails, or texts from one language to another.
Task automation: ChatGPT can automate repetitive and time-consuming tasks such as scheduling meetings, sending reminders, or updating reports, allowing you to focus on other important tasks.
But if anyone can do this, where do YOU add value? The good news is that chatGPT is not a very good employee and requires a good manager. I’ve found that Mr. GPT (mine identifies male, but yours may identify differently) is a great improviser, fun to talk to, and sounds super confident all the time. He’s a “go-getter”. But he’s also a bit of a people-pleaser and finishes tasks even when he is really just making stuff up. That’s where you come in. As the boss, you do the fact checking and guarantee its work.
And chatGPT isn’t just an office worker either. Properly managed, it can write code, building an entire smartphone app. If you’re not a coder, now you can hire one, for really cheap!
There’s a new science emerging for managing LLM’s called prompt engineering. This will be a critical skill for you to develop. Simply put, prompts are the “instructions” you give and LLM. Changing how you structure prompts can result and vastly different outputs. Here’s some crazy things about prompting to get you started:
Role prompting: Giving the LLM a role to play changes how it responds. For instance, saying “You are a math professor” will make its responses to math problems more likely to be correct. Telling it “you’re a comedian” or “your a comedy writer” will make its writing better and more humorous. Don’t you wish this trick worked on humans?
Chain of Thought prompting: Asking the LLM to explain their thought process, break things down into steps, or “show your work” will also yield much better results. Why? LLM’s are predictive models. So if you give it a problem, it will directly predict the answer with varying results. If instead it’s instructed to break the problem into steps, then it will solve the problem differently, with each step being much more predictable, with the end result being much more accurate.
Example prompting: If you want your chatbot to act a certain way, you can give it an example in the prompt. For instance:
Please write a stirring speech about the future of Dinocompany, which is a company that builds widgets to improve the lives of dinosaurs. Here is an example of a stirring speech:
"Four score and seven years ago our fathers brought forth on this continent a new nation, conceived in liberty, and dedicated to the proposition that all men are created equal. Now we are engaged in a great civil war…
You can learn all about it for free at learnprompting.org.
In this changing world, technology and innovation is constantly requiring us to adapt. This can feel scary, but true innovators can see beyond the fear and glimpse the opportunity that lies beyond it. It’s time to up our game, take responsibility for our destiny and become the masters of AI. We must learn prompt engineering (luckily, the new coding language is English) to leverage the great power of LLM’s as tools to enhance, not replace our work.
As many have said, “I welcome our new AI Overlords Worker Bees”