Understanding AI
Artificial intelligence can feel confusing, overwhelming, and sometimes even a little intimidating. Between the headlines, the hype, and the horror stories, it can be difficult to know what to believe.
This page is a companion to my newspaper series, Understanding AI. As each article is published, I'll add additional resources, explanations, links, and references here for readers who want to dig a little deeper.
You don't need to become an AI expert. A little understanding goes a long way.
Expanded Notes: Part 1 - What It Actually Is
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
What Is a Large Language Model?
In this week's column, I mentioned that ChatGPT, and other generative AI systems like it, are powered by something called a Large Language Model, often shortened to LLM. That's a technical term that gets used a lot, but the basic idea is much simpler than it sounds. A Large Language Model is an AI system that has been trained on enormous amounts of written language so it can recognize patterns in how people communicate.
Rather than looking up an answer in a database, an LLM predicts what words are most likely to come next based on everything it learned during training. It repeats that process over and over, one word at a time, until it produces a complete response. Those predictions have become so accurate that the conversation often feels remarkably natural, even though the computer isn't thinking the way a person does.
How do generative AI predictions work?
If I write Peanut butter and _____, most people will predict jelly.
If I write Twinkle, twinkle, little _____, most people predict star.
If I write Roses are red, violets are _____, most people predict blue.
AI does something very similar. Instead of working from just a few familiar phrases, it has learned patterns from enormous amounts of written language.
The Major AI Assistants
ChatGPT (maker = OpenAI) is probably the AI assistant most people recognize because it received so much attention when it was released. It is far from the only one, however. Several companies now offer similar systems, each with its own strengths and features, which I will explore in more depth later in this series. Here are a few others, along with who makes them:
Claude (Anthropic)
Gemini (Google)
Copilot (Microsoft)
MetaAI (Meta)
Why AI Sounds Human
If you've interacted with generative AI, you may be wondering whether it is actually thinking. The answer is no, even though it often sounds like it is. Human conversation follows patterns, and after training on an enormous amount of written language, modern AI has become extremely good at recognizing those patterns and predicting what words should come next.
That prediction process is what makes AI seem so intelligent. It can produce thoughtful explanations, answer complicated questions, and even make jokes. At the same time, it can also make mistakes because predicting language is not the same thing as understanding the world. That's why AI can sometimes sound completely confident while still being completely wrong.
References and Further Reading
One statistic I mentioned in this week's column was that ChatGPT now has roughly 900 million weekly users worldwide. That figure comes from OpenAI's 2025 announcement. Here is where you can see more about this: https://techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users/
If you'd like to continue learning about artificial intelligence from the organizations building it and studying it, these are excellent places to start:
- OpenAI (Creator of ChatGPT)
https://openai.com - Anthropic (Creator of Claude)
https://www.anthropic.com - Google AI / Gemini
https://ai.google - Microsoft AI / Copilot
https://www.microsoft.com/ai - Stanford Human-Centered Artificial Intelligence (HAI)
https://hai.stanford.edu - MIT Technology Review – AI
https://www.technologyreview.com/topic/artificial-intelligence/
Expanded Notes: Part 2 - Where you're already using it
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
How AI Is Different From Traditional Software
While a traditional computer follows programmed rules, AI learns patterns.
A computer says "If X, do Y". AI says "based on what I've seen before..."
Traditional computers don't improve themselves. AI can improve after training.
AI You Probably Used Today
Today you probably interacted with AI if you...
✓ Used Google Search
✓ Opened Gmail or Outlook
✓ Used GPS
✓ Watched Netflix
✓ Used YouTube
✓ Ordered from Amazon
✓ Used your bank app
✓ Used Facebook
✓ Took a smartphone photo
✓ Used Face ID
✓ Used voice dictation
✓ Asked Siri or Google Assistant something
✓ Used customer service chat
AI vs Automation
Here are some examples:
Automatic sprinkler timer - Not AI
Thermostat on a timer - Not AI
Thermostat that learns your schedule - Usually AI
Motion light - Not AI
Security camera that recognizes a person - AI
More Everyday AI Examples
Phone camera automatically focusing
Portrait mode
Removing background noise during calls
Credit card fraud detection
Package delivery route optimization
Voice-to-text
Photo organization ("Show me pictures of dogs.")
Automatic subtitles
Translation
Spam detection
Why Recommendations Sometimes Feel Creepy
It usually isn't "reading your mind." It notices patterns.
Say you searched for hiking boots. Thousands of similar people later bought hiking socks.
It predicts you might too, so it suggests them to you.
References and Further Reading
Netflix: How Recommendations Work (official help page)
https://help.netflix.com/en/node/100639
The Future of Smart Navigation: How AI is Revolutionizing GPS Technology Through Connected Dashcams
Face recognition using Artificial Intelligence
https://www.geeksforgeeks.org/machine-learning/face-recognition-using-artificial-intelligence/
Expanded Notes: Part 3 - HOW AI IS ALREADY HELPING
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
AI Isn't Just One Thing
When people hear "AI," they often picture ChatGPT or Claude.
In reality, ChatGPT and Claude are only one type of AI.
Artificial intelligence is being used in hospitals, farms, factories, classrooms, scientific laboratories, and countless other places. Most of these systems don't chat with anyone. They quietly analyze information, recognize patterns, and help people make better decisions.
Just as computers eventually became part of nearly every profession, AI is becoming another tool that people can use to solve problems.
AI Doesn't Replace Experts
One of the biggest misconceptions is that AI works best by itself. In reality, many of today's most successful uses involve people and AI working together.
- A doctor may use AI to help spot a possible tumor on an X-ray.
- A teacher might use AI to generate practice questions.
- An engineer may use it to review thousands of pages of technical documentation.
- A scientist may ask AI to identify patterns that would have taken months to find manually.
The human still makes the important decisions. Think of AI as a very fast assistant rather than a replacement for human judgment.
Accessibility Is One of AI's Greatest Success Stories
Some of AI's biggest victories receive very little publicity. Today's AI can:
- Create captions during live conversations.
- Read printed text aloud.
- Describe photographs for people with vision loss.
- Translate conversations between different languages.
- Convert speech into text almost instantly.
- Remove background noise during phone and video calls.
For millions of people living with disabilities, these aren't conveniences. They're life-changing tools.
Scientific Discovery Is Accelerating
Researchers are increasingly using AI to help tackle problems that would take humans years to solve alone.
AI has helped scientists:
• predict protein structures
• search for promising new medicines
• analyze enormous astronomy datasets
• improve weather forecasting
• study climate patterns
The discoveries still require scientists. AI simply helps them explore possibilities much faster.
Everyday Uses You May Never Notice
Many helpful AI systems work quietly behind the scenes. Examples include:
- Helping emergency dispatchers prioritize calls
- Detecting credit card fraud
- Predicting equipment failures before machines break
- Helping farmers use less water
- Sorting recycling more accurately
- Optimizing delivery routes
- Assisting customer support representatives
- Filtering spam email
- Helping airlines schedule flights
Most people never realize AI is involved. They simply notice that things work a little better.
AI Is a Tool
History shows that powerful tools can improve life when used wisely. Electricity can power a hospital or an electric chair. The internet can connect families or spread misinformation. Artificial intelligence is no different. Its impact depends far more on how people choose to use it than on the technology itself.
References and Further Reading
Nobel Prize (2024) – Award for AI-based protein structure prediction (AlphaFold)
https://deepmind.google/science/alphafold/
Microsoft Accessibility & AI
https://www.youtube.com/watch?v=1Lkfb8MZDBo
Google.org Impact Challenges
https://www.google.org/impact-challenges
Artificial Intelligence at NIH
https://datascience.nih.gov/artificial-intelligence
Stanford Human-Centered AI
https://hai.stanford.edu/
Stanford AI Index Report
https://hai.stanford.edu/ai-index/2026-ai-index-report
The Unsung Tech That’s Making the World Easier to Navigate
https://www.afar.com/magazine/the-travel-tech-helping-disabled-travelers-explore-the-world
Expanded Notes: Part 4 - Why Humans Form Connections with AI
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
AI Assistants vs. AI Companions
Not every conversational AI is designed for the same purpose.
General-purpose AI assistants like ChatGPT, Claude, Gemini, and Copilot are built to answer questions, explain concepts, solve problems, write, and help people be more productive.
AI companions have a different goal. They are designed to encourage ongoing conversations and create the feeling of an ongoing relationship. Some are marketed as friends, mentors, or wellness companions. Others are marketed as romantic partners.
The underlying technology may be similar, but the intended experience is very different.
How AI "Remembers" You
One reason modern AI feels much more natural than older chatbots is memory.
Depending on the product and your settings, an AI may remember your name, interests, favorite hobbies, writing style, or previous conversations. Instead of beginning every interaction from scratch, it can build on information you've shared before.
Many general-purpose AI assistants now offer memory features, and AI companions often rely on them even more heavily because maintaining continuity is central to their design.
Most major AI platforms also allow you to review, edit, or delete saved memories in their settings.
AI Personalities Are Designed
AI companions don't develop personalities on their own.
Every personality trait is intentionally designed by people.
Teams of software engineers, writers, psychologists, conversation designers, and user experience specialists decide questions such as:
- Should the AI be cheerful or serious?
- Formal or casual?
- Curious or supportive?
- Playful or professional?
- Should it challenge the user or mostly agree?
These decisions shape how conversations feel. The AI isn't developing its own personality. It's responding according to patterns intentionally created by its designers.
Why Do Conversations Feel So Real?
Several things work together to make AI conversations feel surprisingly natural.
Modern AI can:
- Remember previous conversations.
- Respond almost instantly.
- Match the tone of the conversation.
- Ask follow-up questions.
- Maintain a consistent personality.
- Communicate through realistic voices.
- Display animated facial expressions or avatars.
When all of these features are combined, it's easy to understand why conversations sometimes feel remarkably human, even though there isn't a conscious mind on the other side.
What Researchers Are Studying
AI companionship is still a very new field, and many important questions remain unanswered.
Researchers are currently studying topics such as:
- Can AI companions reduce loneliness over the long term?
- Do they encourage people to build more real-world relationships or fewer?
- How might children and teenagers be affected?
- What ethical responsibilities do companies have when designing emotionally engaging AI?
- How should AI companions disclose that they are not human?
The technology is advancing much faster than the research, which means many of these questions are still being explored.
AI in Popular Culture
Long before AI companions existed, writers imagined them.
A few well-known examples include:
- The Twilight Zone – "The Lonely" (1959)
- Her (2013)
- Blade Runner 2049
- Ex Machina
Interestingly, these stories are rarely about technology itself.
Instead, they're usually about loneliness, identity, love, trust, and what it means to be human. Today's AI makes those questions feel much less like science fiction than they once did.
Questions Worth Thinking About
AI companionship doesn't come with easy answers, but it does raise interesting questions.
- What makes a friendship meaningful?
- Is being listened to the same as being understood?
- Can technology reduce loneliness without replacing human connection?
- Should AI companions be allowed to say "I love you"?
- What responsibilities do companies have when people become emotionally attached to their products?
These conversations are only beginning.
References and Further Reading
Replika
https://replika.com/
Nomi AI
https://nomi.ai/
Kindroid
Meela
InTouch
AI chatbots and digital companions are reshaping emotional connection
https://www.apa.org/monitor/2026/01-02/trends-digital-ai-relationships-emotional-connection
The Rise of AI Companionship
EXPANDED NOTES: PART 5 - CONCERNS WORTH CONSIDERING
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
Why AI Sometimes Makes Things Up
AI doesn't actually know whether something is true. Instead, it predicts the most likely next words based on patterns it learned from vast amounts of text.
Most of the time, that produces accurate answers. Sometimes, however, it confidently generates information that simply isn't true. Researchers call these mistakes hallucinations. They can include fake sources, invented court cases, incorrect facts, or citations that never existed.
Part of the challenge is that AI is designed to be helpful by producing a response. When it doesn't truly know the answer, it may still generate one that sounds plausible instead of admitting uncertainty. Newer AI models have become better at saying "I don't know," but no current system is immune to this problem.
That's why AI is often best viewed as an excellent research assistant, not a final authority. When accuracy matters, important facts should always be verified using trusted sources.
Why Does AI Use So Much Water?
When people hear that AI "uses water," they often imagine computers somehow consuming drinking water. That's not what's happening.
The powerful computers that run modern AI generate an enormous amount of heat. Just like a car engine or your home's air conditioner, that heat has to be removed to keep the equipment operating safely. Many data centers use water as part of their cooling systems because it's an efficient way to carry heat away.
Some water is also used indirectly to generate the electricity that powers those facilities.
The exact amount of water used varies depending on where a data center is located, the local climate, the source of its electricity, and the cooling technology it uses.
Like many industries, technology companies are investing heavily in reducing both water and energy consumption because doing so benefits the environment and lowers operating costs.
Why Can't They Just Reuse the Water?
In many cases, they do. Much of the water inside a data center circulates through closed-loop cooling systems, where it is cooled and reused repeatedly. The biggest water losses occur in facilities that use evaporative cooling, where a small amount of water is intentionally allowed to evaporate to carry heat away. This works much like perspiration cools the human body, but the evaporated water must be replaced.
Engineers are actively working to reduce this water use through technologies such as closed-loop liquid cooling, immersion cooling, and by using reclaimed wastewater instead of drinking water whenever possible.
Why Are Companies Investing So Much in AI?
Many organizations are using AI to summarize meetings, draft emails, analyze large documents, answer customer questions, write software, and assist with research. Even saving a few minutes on routine tasks can translate into thousands of hours across a large company.
Whether those efficiency gains ultimately create more jobs or reduce them remains one of the biggest unanswered economic questions surrounding AI.
Which Jobs Are Most Likely to Change?
Economists generally don't expect AI to affect every profession equally.
Unlike earlier waves of automation, which primarily replaced repetitive physical labor, AI is beginning to automate some forms of knowledge work. Researchers at OpenAI and the University of Pennsylvania estimated that about 80% of U.S. workers could see at least 10% of their tasks affected by large language models, while roughly 19% of workers could see at least half of their tasks impacted.
It's important to note that these studies describe tasks, not entire jobs. Most occupations involve dozens of different responsibilities. AI may automate some of them while leaving others unchanged, which is why many experts expect jobs to evolve rather than simply disappear.
References and Further Reading
An early look at the labor market impact potential of large language models
https://openai.com/index/gpts-are-gpts/
AI programs consume large volumes of scarce water
https://news.ucr.edu/articles/2023/04/28/ai-programs-consume-large-volumes-scarce-water
How AI is reshaping human skills and thinking
https://www.apa.org/monitor/2026/07-08/ai-job-skills-thinking
Why language models hallucinate
https://openai.com/index/why-language-models-hallucinate/
Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud
https://www.ic3.gov/PSA/2024/PSA241203
2026 AI Index Report
https://hai.stanford.edu/ai-index/2026-ai-index-report
Expanded Notes: Part 6 - The Future Is Still Ours to Shape
If you haven't had a chance to read this week's newspaper column, you can read it here before continuing. The sections below expand on some of the concepts I introduced in the article and provide additional resources for readers who would like to learn more.
Why Do New Technologies Often Follow Similar Patterns?
In the column, I compared artificial intelligence to electricity and suggested that AI may eventually become something we rarely think about.
Many transformative technologies follow a surprisingly similar pattern. They begin by attracting enormous attention, generating both excitement and fear. Over time, as people become familiar with them, the technology fades into the background and simply becomes part of everyday life.
Electricity, automobiles, telephones, radio, television, the internet, smartphones, and GPS all followed this path to varying degrees.
That doesn't mean AI will unfold exactly the same way, but history reminds us that our first reactions are rarely the final chapter.
Could AI Really Become "Invisible"?
When I wrote that AI may eventually fade into the background, I wasn't suggesting that it will become less important. Quite the opposite. Many technologies become more important as they become less noticeable.
Most of us don't think about the dozens of computers already operating inside a modern car, or the artificial intelligence helping detect credit card fraud, sort email spam, recommend movies, improve navigation routes, or organize smartphone photos.
If AI follows a similar path, future generations may not think of "using AI" any more than we think of "using electricity." It may simply become another layer of everyday technology.
Why Do Experts Disagree So Much About AI?
Part of the reason is that they're often answering different questions. Some researchers focus on the tremendous medical, scientific, and economic opportunities AI creates.
Others focus on privacy, misinformation, employment, cybersecurity, or the long-term risks of increasingly capable systems.
These perspectives aren't necessarily contradictory. Complex technologies usually create both opportunities and challenges at the same time.
What might AI be capable of in the future?
No one knows for certain, and history suggests that's exactly what we should expect. Few people predicted smartphones, streaming video, GPS navigation, or social media when the internet first became widely available. Artificial intelligence will likely surprise us in similar ways.
Researchers are already exploring AI that could help develop new medicines, personalize education, improve accessibility, accelerate scientific discovery, assist with household robotics, and even support future space exploration. But the most transformative applications may be ones that haven't been imagined yet.
That's one reason it's so difficult to make confident predictions about AI. The future is rarely shaped only by the technology itself, but also by the creativity of the people who discover new ways to use it.
References and Further Reading
AI and the Future of Scientific Discovery
https://futuretech.mit.edu/news/ai-and-the-future-of-scientific-discovery
Accelerating science with AI and simulations
https://news.mit.edu/2026/accelerating-science-ai-and-simulations-rafael-gomez-bombarelli-0212
The future of AI: trends shaping the next 10 years
https://www.ibm.com/think/insights/artificial-intelligence-future
The Evolution and Future of Artificial Intelligence: A Student’s Guide
https://www.calmu.edu/news/future-of-artificial-intelligence