Introduction
In the first of the using AI series, let’s look at some practical uses of AI at home.
I chose to start this series at home for a couple of key reasons. It allows us to bypass some challenges you’ll face when wanting to use AI at work or at school. From policies to firewalls, many work/school provided devices aren’t allowed to access AI tools and services right now. I look forward to addressing those topics as we continue into the series, but for now, let’s stick with the home setting on personal devices.
One of the easiest and most popular ways to get into AI at home is by installing the ChatGPT mobile app or accessing ChatGPT online. ChatGPT, developed by OpenAI, has been at the forefront of AI development over the past year, and is arguably at the front of the pack in terms of capability. The “free tier” of ChatGPT currently runs GPT-3.5-Turbo, whereas a paid tier grants access to GPT-4. Once you open the app and create a free account, you can immediately begin chatting with GPT, asking it questions, and learning how it responds.
What are some good questions for a first-time user to ask AI?
When experiencing new technology, sometimes we just need a little help to know where to start. Try a few prompts like these:
Facts
- How long does it take to get to Mars?
- What is the longest continuous road in the world?
- Provide a list of the 10 fastest cars by top speed
Analysis
- Provide a summary of the US Constitution
- Summarize the first 3 Harry Potter books
- Given an annual income of $75,000, how much should I spend on housing
Fun/creative
- Create a nursery rhyme in the style of doctor Seuss about computers
- Write a haiku about a T-rex living in Tokyo
Although ChatGPT is currently class leading, this space is developing rapidly. Tomorrow could see the release of a shiny new model that surpasses the competition. I always aim to provide choice to readers in the form of competing solutions as well as open-source options. At the time of writing, here are some of the top AI tools available:
Chat tools
- ChatGPT and Microsoft Copilot / Bing Chat (OpenAI)
- Bard / Gemini (Google)
- Chat with Llama2 (Meta, Open source)
- Gpt4All (open source, self-hosted)
Image tools
- Dall-E and Microsoft Bing image creator (OpenAI)
- Midjourney (Midjourney inc.)
- Imagine with Meta (Meta)
- Stable Diffusion (open source, self-hosted)
We’ll talk more about image generation in a future artical, but for today let’s focus on chat / text generation.
How can I get started using AI?
As mentioned above, you can check out many of the tools above simply by accessing the corresponding website or app. For example, chat.openai.com and bard.google.com are a click away. AI Image generators are similarly easy to find and start using. The self-hosted options, however, must be installed and run locally; preferably on a higher end workstation or gaming computer. Future articles here at artificially-awesome will walk through setting up and running AI workloads on personal devices, so check back if that interests you!
What can I do with AI at home?
The obvious answer is “chat and generate images”, but let’s explore a little more! First, it’s important to recall that we are distinguishing between “home” and “work/school” for this article. Which is to say “help with homework / work” would be disingenuous to list here. We plan to consider those topics in follow-up articles. So, what else could AI do for you at home? Answer: tons. Let’s focus on chat tools first.
AI chat tools provide text-based responses to your questions or prompts, and they do so by soaking up huge quantities of written human knowledge. Imagine you have a friend who’s read millions of books, magazines, and every public website on the internet. This friend would absolutely slay at trivia night. And by analyzing your question and generating an answer based on its own learning, it can be better at answering questions than a simple web search. Let’s dig into a deeper example.
Say you want to know the shortest highway in your home country. Most likely, a web search will lead you to a Wikipedia page or somewhere similar with a correct answer. Now, let’s say you want to know the 10 shortest highways in your country. Suddenly, your simple task could take hours of research. However, an AI with extensive knowledge of geography, maps, history books, and more may be able to immediately respond to your query with a list. Let’s put it to the test:
Prompt: What are the 10 shortest highways by length in the USA? Include the length in your answer.
As of my last knowledge update in January 2022, here are some relatively short highways in the USA along with their approximate lengths. Keep in mind that these figures may have changed, and for the most accurate and up-to-date information, it’s recommended to check official sources:
- I-375 in Detroit, Michigan: Approximately 1.06 miles
- I-878 in Queens, New York: Approximately 1.40 miles
- I-180 in Williamsport, Pennsylvania: Approximately 1.23 miles
- I-110 in Pensacola, Florida: Approximately 6.39 miles (including the business loop)
- I-587 in Kingston, New York: Approximately 1.28 miles
- I-277 in Charlotte, North Carolina: Approximately 4.11 miles
- I-878 in Nassau County, New York: Approximately 0.70 miles
- I-395 in Miami, Florida: Approximately 4.44 miles
- I-184 in Boise, Idaho: Approximately 2.49 miles
- I-238 in San Leandro, California: Approximately 2.90 miles
After an initial response I asked it to include the length of each highway in the answer, that response is what you see above.
Suddenly, the value of AI starts to become apparent. Not only can it answer simple facts, but by ingesting huge amounts of data, it is able to perform analysis and report the results back, saving us potentially minutes or hours of manual work.
Are answers reliable?
At this time, no AI can be considered 100% reliable. Why is that? Check out our example prompts above under the fun/creative heading. Developers want AI models to have some liberty to imagine and create. And when a system is allowed to create, it opens the door on creating answers that are wrong. This is referred to as hallucinating. The AI system believes it has generated an appropriate response to your prompt, but turns out is just plain wrong.
This means that for the foreseeable future, any AI generated output should be fact checked and tested, before being accepted. However, these tools are rapidly evolving. Perhaps someday soon we will be able to have AI #1 fact check everything that AI #2 is saying. Look for a future article on how AI is helping to make AI better.
Wrap-up
That’s it for now. If you are new to using AI at home, I hope you enjoy your first foray and start exploring new ways to make use of this exciting new toolset!
Look ma, no ads! If you find this site helpful and would like to help keep the lights on, please consider ☕buying me a coffee, thanks!