AI for Normal People
Hearing the terms GenAI? ChatGPT? Is your kids' school warning it won't tolerate cheating with these AI tools? Here's what you need to know - and five family-friendly GenAI prompts to experiment with!
“Siri, set the timer for 25 minutes.”
I confess, despite specializing in AI and search for almost two decades, I’ve rarely used Siri on my iPhone or computer - aside from setting timers, and maybe the occasional reminder. Not shunning Siri for any good reason - just old dogs, new tricks?
My 101-year old Aunt Molly uses her voice assistant (VA) Alexa for weather. My sister Chris uses Alexa for weather forecasts, traffic reports, restaurant reviews, and to call Diane. Adoption for voice assistants here in the US is a little more than a third of the population — with the overwhelming amount of people asking about the weather: be it for today, tomorrow, or for a neighbor’s wedding. My sons use Siri all day for a zillion things that I still put on PostIt notes. As I look at the litter of yellow notes all over my desk, I am thinking I should follow my sons’ lead.
Siri and Alexa were revolutionary because they let you use voice to interact with the computer — instead of typing. For any of these Voice or Virtual Assistants to function, they require natural language processing (NLP) - which is really just very sophisticated (and fast) pattern recognition. This is similar to how humans learn. We group things, people, and experiences together based on patterns we see or experience. Fire and Stove are things that are hot. Ice and snow are things that are cold. Computers can recognize speech in somewhat the same way.
Remember when we diagrammed sentences back in sixth grade? We learned that, in English, the adjectives are (likely) just before the noun, the subject is before the predicate, the direct object is AFTER the predicate. Turns out machines are much faster (and better) at diagramming a sentence than a 12-year old is. It also turns out that with enough memory, computers have faster recall.
Despite not having worked Siri into my life, I have made room for the next generation of AI tools that uses NLP: ChatGPT, Google Gemini, Bing’s Copilot. If you aren’t familiar with them, follow along, because I think you will really find them useful.
I don’t like ascribing human intelligence to machines so I prefer to use the technical term “recall” instead of remembers. Computers don’t remember - anything. They can find or recall a document - but they don’t know the blood, sweat, and tears you went through to create it. They don’t know if anyone else collaborated on it. So it’s not a memory. A computer finds a representation - of your memory.
A computer can “learn” that the word “dog” is a noun, and that it is often associated with puppy and canine. It also may be associated with pets, and sometimes associated with cats - although much less frequently. Computers learn by human training, which requires extensive tagging and supervising to make sure that a Tabby is not inadvertently labeled a dog. Another way of learning is to apply statistical models. A computer may recall a pattern that when people research “pets,” they often are researching “dogs” or “cats.” Those little thumbs up and thumbs down you see on some applications? That is the computer taking feedback on how well it did with the data it gave you.
Here in the US, more than a third of people use voice assistants regularly, with 70% saying that it is much faster than typing.
To date, Siri* and Alexa have been constrained by recalling from only a single category of information at a time. For example, what time is it (from your time app), play Taylor Swift (from a music app), or what is the weather like (from your weather app).
Which finally brings me to ChatGPT, Gemini, et al. They are called generative AI (GenAI) chatbots - and people call them virtual assistants. A chatbot is similar to voice assistants, in that it understands human understandable questions. But GenAI takes that to an entirely different level. It scours all sorts of data from across the entire Internet, and generates elaborate answers. It’s so mesmerizing to watch it write the answer that you are forgiven for thinking it possesses human intelligence.
By the way, these GenAI tools doesn’t generate just prose. They can create images, code, animations, music, virtually anything you can imagine. A person merely types in a question (which is called a prompt) and the chatbot goes to work. You can and should make your prompt as descriptive as possible. I used Microsoft Designer to generate the image on this page. I then brought it into another tool to refine it and get more of the look I was going for.
The reactions are mixed.
The Pros to GenAI
Foodies and cooks like my son love the speed at which it can come back with menus for any cuisine, with or without any ingredient, for any number of courses. Just type in a prompt like “Prepare a 5-course Latin menu that features lamb. I want to be able to prepare most of it the day before.”
Travelers are delighted to create itineraries with ease - for the intrepid - or the not-so intrepid voyagers. Need it to be wheelchair or stroller accessible? Just add it to your prompt.
Writers can quickly amass research and ideas on any subject, geared for any specific audience. Have an interview with former astronauts who worked at Cape Canaveral in the 1980s and want to get up to speed on topics to ask? You’ll have a list in seconds.
Programmers are thrilled to get ideas for code and built-in functions, without having to put up with the snark that they might get from the peanut gallery on public sites like Stack Overflow.
Entrepreneurs and small businesses can use it for visual brand concepts — that can be refined - and then given to a designer to finesse (or not.)
The Cons to GenAI
Educators are horrified that kids will create entire essays using these bots. Essays can be created this way - whether or not they are truthful or compelling is another story. Tell your kids not to try it, teachers have tools to see whether or not an essay was human—or machine—generated.
Media and academic outlets are furious that the intellectual property that they invested in creating is being appropriated by GenAI companies to train their chatbots. The GenAI companies will generate billions of dollars of value - and share none of it with the content creators.
CEOs, foolishly, are ecstatic that they might be able to replace their workforce with these bots. GenAI firms like OpenAI are helping fan those expectations. Its CTO recently said, “Some creative jobs maybe [sic] will go away, but maybe they shouldn’t have been there in the first place if the content that comes out of it is not very high quality.”
Researchers note that the training is done on older data. Stale data in, means dated data out.
I have so many thoughts about both the pros and the cons- and will dig into some of them in separate upcoming pieces. Bottom line, while there is a ton of intelligence in chatbots, it is not the same as human intelligence. They merely look for patterns found in what others have clicked on in the past - and regurgitate that information back to you. Yes, they can construct it in a linguistically sound way - because, as I said, they can diagram a sentence better than a 12-year old.
Chatbots also – like some people in the news – make stuff up. We call it hallucinations when machines do it. So fact-check responses accordingly - and don’t use it to create essays. Sheesh.
All of that said, do I think you, the average consumer, should familiarize yourself with this new technology? I do!!
Go to Gemini.google.com or ChatGPT.com and just play. You can and should be detailed in writing your prompt. I give some ideas below.
5 Fun GenAI Prompt Ideas:
Modify the following to meet your needs.
“Create a dinner menu for 7 people that includes _______ but not ________.”
”Help me come up with an itinerary to travel by car with 2 active boys ages 8 and 10 through the Adirondack Mountains, in August. They love history.”
”My kids [boys/girls/ages] are going to sleep away camp for the first time, what are things they are likely to forget?”
”Can you come up with 3 rainy day activities to do with my 3- and 5-year old children. The activities should only require common household items. It would be great if one of the activities has something to do with science, and one with cooking, and one that is just fun.”
”I have a car trip through the Shenandoah Valley this summer with 3 kids ages 5-11. Can you come up with a car game that would keep them entertained?
The way to get the most out of GenAI is to write exceedingly detailed prompts. But remember what I said about pattern matching: GenAI will regurgitate things it has “learned” about a given subject. So if you get an answer you don’t like — simply tell it do try again — this time without including (whatever it is you don’t like).
* It’s anticipated that Siri will evolve from being a voice assistant to a true chatbot. Keep your eyes here, I think this is going to be truly interesting.
Send Me Your Prompts!
I’d love to know your experiences with any of these tools. In the comments below, share with me some of the prompts you found particularly interesting - I’d love to incorporate them in future pieces.
Additional Articles on GenAI
