Welcome to On Tech: AI. This is a pop-up newsletter explaining artificial intelligence, how it works, and how to use it.
A few months ago, my colleagues Cade Metz and Kevin Roose described the inner workings of AI, including chatbots like OpenAI’s ChatGPT, Microsoft’s Bing, and Google’s Bard. Now we are back with a new mission. Helping them learn how to maximize the potential of AI.
People from all walks of life—students, programmers, artists, and accountants—are experimenting with how to use AI tools. Employers are posting jobs for people who are proficient in using job listings. Soon, if not yet, we will have the opportunity to use AI to streamline and improve our work and personal lives.
As a personal technology columnist for The Times, I want to help people find ways to use these tools safely and responsibly to improve many areas of their lives.
In today’s newsletter, I’d like to talk about two general approaches that are useful in a variety of situations.
Then, in the coming weeks, we’ll provide more specific tips on different aspects of life, including parenting and family life, work, personal organization, learning/education, creativity, and shopping.
First, here are some common sense caveats.
If you’re concerned about privacy, omit personal information such as your name and place of work. The tech company says your data is used to train its systems, which probably means others may view your information.
don’t share sensitive data. Employers may have specific guidelines and restrictions, but in general, entering trade secrets and confidential information is a very bad idea.
Hallucinations: Chatbots rely on a technology called large scale language models (LLMs). LLM gets its power by analyzing vast amounts of digital text culled from the Internet. A lot of things on the web are wrong and chatbots can repeat those lies. Sometimes you make up something while trying to predict patterns from a huge amount of training data.
Chat GPT, Bing and bard is one of the most popular AI chatbots. (To use ChatGPT, you need to create an OpenAI account and have a subscription to its latest version. Bing requires using Microsoft’s Edge web browser. Bard requires a Google account.)
It looks simple to use, just type something in the box and the answer will appear. — ask the question the wrong way and you’ll get generic, unhelpful, and sometimes completely inaccurate answers.
I’ve found that there is a technique for typing and constructing exact words to generate the most helpful answers. I call these golden prompts.
People who make the most of chatbots use a variation of the following strategy.
“Act as if it were.” Start your prompt with these magic words and you’ll tell the bot to emulate an expert. For example, type “act like I’m an SAT tutor” or “act like I’m a personal trainer” and the bot will model itself after people in those professions.
These prompts provide additional context for AI to generate responses. In fact, AI doesn’t understand what it means to be a tutor or personal trainer. Instead, this prompt helps the AI take advantage of certain statistical patterns in the training data.
Weak prompts without guidance produce less useful results. Suppose you just type “What should I eat this week?” The chatbot creates a general meal list for a balanced diet, such as having fried turkey and colorful veggies for dinner (this sounds very “okay” to me). ).
“Tell me what else I need to do to do this.” For more personalized results (such as health advice tailored to your specific body type or medical condition), invite the bot to request more information.
In the Personal Trainer example, the prompt looks like this: “Act as if you were my personal trainer. Create a weekly workout plan and meal plan. , the bot asks your age, height, weight, dietary restrictions and health goals to tailor your weekly meal plan and fitness routine.
If you don’t get a good answer on your first try, don’t give up too soon. Even better is the words of Ethan Morrick, a professor at the Wharton School of the University of Pennsylvania. Treat bots as if they were human interns: “If you make a mistake, point it out and ask to improve.” Be tolerant and patient and you’ll get better results.
Thread chatbot conversations
Once you get the hang of prompts, you can make your chatbot even more useful over time. The key here is to treat the chatbot as his web search and avoid starting a new query each time. Instead, keep a few conversation threads open and add to them over time.
This strategy is easiest using ChatGPT. Bing requires you to reset conversations periodically, while Bard doesn’t make it so easy to navigate between conversation threads.
Natalie Choprasert, an entrepreneur living in Sydney, Australia, advises companies on how to use AI and uses ChatGPT as a business coach and executive assistant. She runs separate dialogues for each of these roles in parallel.
In the Business Coach thread, she shares insights about her professional background and company goals and problems. Executive She shares scheduling information, such as Assistant She Threads, clients she’s meeting with.
“It builds up and gets trained properly, so if I ask a question later, it will give me an answer in the right context and closer to what I’m looking for,” Choprasert said.
She shared a special golden prompt that trained her assistant to be even more helpful. It’s about applying frameworks. She recently read Clockwork, a book on entrepreneurship. She’s a businessShe’s a coachShe asked ChatGPT for advice using the “Clockwork” framework, and she said she could incorporate the principles in the book into an action plan to expand her company. She was delighted.
share the prompt
What are the golden prompts that have yielded the most impressive and beneficial results from AI? Email us your example. We may use your submission in future editions of this newsletter.