Before we open a chatbot
What AI is, plainly.
A lot of people hear "AI" and think ChatGPT. That's one kind of AI. There are others you probably already use every single day without calling them AI. A quick tour before we pick the piece that's most useful today.
The big picture
"Artificial intelligence" is a very large umbrella. The simplest way to think about it: any software that learns patterns from data to make predictions or produce something new. That covers a lot of territory.
Here's a map of what fits under the umbrella. You already use most of it.
AI you use without thinking about it
Recommendation systems
The reason your TikTok feed feels like it knows you. The reason Netflix suggests the next show. The reason your "For You" page on any platform feels personal. Those are AI systems trained on what millions of people watch, like, and skip. They predict what will hold your attention. They are good at it.
Spam filters and fraud detection
When Gmail puts an obvious junk email in the Spam folder, AI did that. When your bank texts you because a charge "doesn't look like you," AI flagged it. Both are older than ChatGPT and both quietly work well.
Voice assistants and autocomplete
Siri, Alexa, Google Assistant. The phrase your phone suggests when you're typing a text. All AI. The text autocomplete in particular is a close cousin of the chatbots we'll focus on later today.
Translation and captions
The auto-generated captions on YouTube. Google Translate. Live translation in a video call. All AI, all surprisingly recent.
Image classification
When your phone's photo app groups all pictures of your dog together, or lets you search for "beach" and pulls up every beach photo you've ever taken. AI trained to recognize what's in an image.
Route finding and maps
Google Maps predicting traffic and suggesting the faster route. The ETA that updates mid-drive because an accident happened. AI plus a lot of real-time data.
The newer kinds of AI
Large language models (LLMs)
ChatGPT, Claude, Gemini. A chatbot you can ask to write, summarize, explain, translate, or brainstorm with. This is what we'll spend most of today on because it's the most useful kind of AI for the career tasks on our agenda: resumes, cover letters, interview prep, planning.
What an LLM really is: a system trained on a huge amount of text to predict what words come next. When you ask it a question, it's producing a likely continuation based on patterns it learned. It is good at sounding fluent. It is not good at knowing whether what it says is true.
Image generation
Midjourney, DALL-E, Stable Diffusion. You type a description and get a picture. Useful for mockups, illustrations, social media graphics, and professional headshots. There are free options (Bing Image Creator, HuggingFace Spaces demos, Google's AI Studio).
Speech to text and text to speech
Whisper transcribes an audio recording into text. ElevenLabs generates a voice reading a script. Useful for turning a meeting recording into notes, or a written script into a voiceover. Free tiers exist for both.
Vision models
Upload a picture and ask questions about it. Most of the big chatbots can do this now. Useful for reading a receipt, understanding a chart in a screenshot, or explaining a diagram.
Agents
AI that doesn't just answer a question but takes an action: books a meeting, fills out a form, browses a website, writes and runs code. This is the newest area and the one changing fastest. Today's examples are uneven. They work beautifully sometimes and embarrassingly other times.
Why we focus on LLMs today
Out of everything above, LLMs are the one kind of AI where a plain-text conversation with no technical setup can help you land a job this week. The resume, cover letter, and interview prep work all lives in writing. LLMs are specifically good at writing.
The other kinds matter, and a few of them we'll touch on later in the day (headshots in the bonus guide, for example). But the core four hours are about learning to write prompts that make LLMs useful for the jobs and roles you want.
Three honest things about AI
- It will be wrong sometimes. Confidently. Your job is to know enough about the topic to catch it. If you ask an LLM about something you know nothing about, you will not know when it's wrong.
- It does not know what is private. Unless you use an enterprise-approved tool, assume anything you paste could be used to train future models. Don't paste anything you wouldn't email to yourself at home.
- It gets better fast, and it costs money. The frontier models improve every few months. The best ones are usually paid, but the free tiers are plenty for the work we're doing today.
Now, to the work. Open the first activity, or if you want to know more about the tools we'll use, see the tools page.