Today, we’re continuing our foray from grammar into artificial intelligence and diving into the remarkable capabilities of ChatGPT-like programs. How the heck do they work? What can they do for YOU? What are some safety concerns? There isn’t a single human who couldn’t benefit from a gen AI resource like this. So, here are some AI hacks and an easy-to-understand breakdown of how ChatGPT4 can help different areas of your life.
What is generative AI?
In the last episode, we discussed the ways AI is already built into the things we use every day. Now, we are focusing on AI that provides requested information or does specific tasks for which you actively ask. Models like ChatGPT and Google Gemini fall under the umbrella of Generative AI. “Gen-AI” is like a super creative artist who uses a computer to make new things. Or a super nerdy researcher or analyst who wants to help you organize info in an easy-to-comprehend way.
Gen-AI can write stories, create art, compose music, and even generate human-like conversations by learning from limitless examples of online material. It’s like having a tool that can develop original ideas or content inspired by what it’s learned from all the existing stuff.
How does generative AI work?
ChatGPT learns by processing massive amounts of data (text, books, music, sounds, images, etc.) and recognizing patterns in how language is used. It continuously refines its understanding through feedback, allowing it to talk to us like humans and provide useful information on a wide range of topics. Using its training, prompting, and chat history, it’s basically the world’s most ingenious and sophisticated predictive keyboard.
What makes a good prompt?
Some people say there are six elements to a great prompt, but I am going to condense them to three.
- Give it context. Tell it who it is, what it knows about, where you are, and what the scenario is. Set the scene! Be specific.
- Give it a task. Tell it what you need it to do using words like explain, list, translate, describe, or summarize. Be clear.
- Give it some parameters for a valuable response. Tell it the tone of voice (technical or casual?), response length (a few sentences or a few paragraphs?), formatting (bullet points or paragraphs), etc.
Let’s “fix” these three prompts
Lazy prompt: Help me renovate my guest bathroom.
GREAT prompt: You’re a home improvement expert. I’m a 33 year old female wanting to aesthetically remodel my full guest bathroom (CONTEXT). I want to replace the shower, tub, tiles, vanity, sink, faucet, two light fixtures, and toilet. Can you make a list (TASK) of 1. Ranges of estimated costs. 2. Tips to get started. 3. Online resources I could use for the design. 4. Nashville stores to visit for inspiration (PARAMETERS).
Lazy prompt: Help! My allergies are horrible.
GREAT prompt: You’re an expert in allergy relief. I live in Nashville, Tennessee, and springtime is horrible for my allergies (CONTEXT). Can you list some tips and tricks that might help with congestion, itchy eyes, and runny nose (TASK)? I have already tried local honey, so don’t include that in your response. Please give me both over-the-counter options and natural options (PARAMETERS).
Lazy prompt: Tell me about AI.
GREAT prompt: You’re a professional in AI for productivity. I am new to AI but interested in learning how AI can help me with everyday tasks as a busy mom of three (CONTEXT). Can you explain how AI can help with schedule management, meal prep, and back-to-school shopping (PARAMETERS)?
What are ChatGPT’s limitations?
- There’s a cut-off date for its information. This varies by platform, and you can easily find the cut-off date by simply asking the bot. It doesn’t pull in real-time, so it can’t tell you the weather or current stock prices.
- It doesn’t understand meaning like humans do and has no human experience. Therefore, it cannot give advice based on personal life experiences or preferences.
- It may produce incorrect information, and it can’t verify the accuracy of what it tells you. Always fact-check!
- It doesn’t know when it’s lying. It can “hallucinate,” which means it will produce an answer even if it doesn’t know the correct one.
- It’s dependent on the quality of the input it receives. This is the main takeaway of this episode! The better at prompting you get, the better these platforms are at helping give you accurate and helpful responses.
What are some safety and ethical concerns?
These models undergo extensive training before they are released to the public. A considerable amount of energy and human power are invested into training the models to mitigate biases, comb for unethical/harmful content, and refuse inappropriate requests.
For example, ChatGPT definitely knows how to rob a bank or cheat on taxes, but it won’t provide a step-by-step explanation. It won’t help you commit a crime or spit out explicit language or sensitive images. It uses built-in filters and ongoing human feedback to avoid generating anything inappropriate, harmful, or sensitive. It’s good to have your own set of best practices. Here are some basic tips for using the models safely and ethically:
- Avoid sharing personal information. Don’t input anything you wouldn’t willingly post on Facebook. Just kidding. But definitely no full names, phone numbers, addresses, or other sensitive personal information.
- Don’t just copy and paste. To avoid legal murkiness, it’s best to use generative AI as a framework to help you form your own ideas. Let it suggest things, then you make it your own. For example, if you ask it to write an email template to entice donors to your charity dinner, rework the email template to make it personalized and in your own voice. If you do use an image or a poem verbatim, cite the model that made it.
- Ask your workplaces, organizations, and schools for their AI policies. Young people should be especially cognizant of what they are allowed and not allowed to do with AI in a school setting. If your business or company doesn’t have an AI policy in place, nudge them to do so!
We are going to dedicate an entire episode to AI as it relates to children. This will include how to have open conversations with kids about AI, the kid-friendly platforms out there, and what parents should look out for. So, stay tuned!
I hope this episode has equipped you to use generative AI more comfortably in your daily life. Send burning AI questions to [email protected], and I’ll see you next month!
**********
For life hacks, travel itineraries, style inspiration, and more — subscribe to our daily emails!
