A University of Arizona · BIO5 Institute
KEYS Research Internship · GPT 101
A One-Hour Module · No Coding Required

Generative AI
for life-sciences
research.

Forty minutes on what AI is actually doing in BIO5 labs in 2026 — then twenty minutes of hands-on practice on the U of A's own platform. You'll leave with one prompt you'll use in your lab this week.

Tyson Swetnam, PhD
BIO5 Institute · genai.arizona.edu
BIO5 KEYS Research Internship
Summer 2026 · Tucson, AZ
AKEYS · GPT 101
02 / Orientation
Who This Is For

KEYS interns in BIO5 labs.

Bioscience, biomedical engineering, biotechnology, statistics, and computational biology. No coding background required. Some high-school biology helps — the examples translate to whatever your lab does.

The point
Not to make you an expert in an hour — to give you a working mental model you can use in the moment.
What you'll walk out with
  • A working mental model of what an LLM is doing when it answers — and where that breaks
  • A short list of tasks AI is genuinely useful for in a bench-science or computational-biology lab
  • A clear sense of the failure modes — hallucinated citations, fabricated protocols, leaked data
  • First-hand experience comparing models and verifying AI-generated citations against primary sources
  • One reusable prompt for a task you'll actually do this week
AKEYS · GPT 101
03 / The Hour
Sixty Minutes · Five Blocks

Forty minutes of lecture, twenty minutes hands-on.

Block Duration Activity
15 min Why this lesson exists
210 min What generative AI actually is
315 min What it's useful for in your lab
410 min What it's risky for — and the policies that protect you
520 min Hands-on at genai.arizona.edu — four exercises, five minutes each
AKEYS · GPT 101
Part 01 · Why This Exists
5 min
Part 01

It changed
the workflow,
not the job.

AI didn't replace researchers in 2024–2026. It changed the daily work of nearly every working scientist — including, almost certainly, your KEYS mentor.

AKEYS · GPT 101
05 / Three Questions
The Whole Goal Of This Hour

In the moment, be able to answer three questions.

01 · WORTH IT?

Will this save me real time?

A task where reaching for an AI assistant genuinely speeds you up — drafting, summarizing, debugging — without costing you accuracy.

02 · TRAP?

Will it quietly produce a wrong answer?

A task where the AI will hand you a confident, plausible answer that is wrong — and you won't catch it. Citations, protocols, statistics.

03 · ALLOWED?

Would this break a rule?

A task where pasting the data would violate your lab's policies, your mentor's trust, or federal law. Unpublished data, patient info, IP.

Context · 2024
AlphaFold won the 2024 Nobel Prize in Chemistry for predicting protein structures from sequence — work that used to take years of crystallography per structure.
AKEYS · GPT 101
Part 02 · The Mental Model
10 min
Part 02

It predicts
the next
word.

That's the whole trick — and understanding it tells you exactly where these systems are brilliant and exactly where they fall apart.

AKEYS · GPT 101
07 / What An LLM Is
A Large Language Model (LLM)

A pattern-recognition system trained on enormous amounts of text. Given everything so far, it predicts the most likely next word. Then the next. Then the next.

What it is not

Models like Claude, GPT, Gemini, Llama, and Gemma are not databases. They do not "look things up." They generate plausible continuations of your text.

Why it feels like thinking

They're so good at next-word prediction that the output feels like reasoning. Useful — as long as you remember what's actually happening underneath.

Want the deeper version — transformers, training, foundation vs. fine-tuned models? See the AI Landscape lesson. For today, "next-word predictor trained on text" is enough.

AKEYS · GPT 101
08 / Consequences
Three Consequences That Matter In A Lab

If it predicts text, then three things follow directly.

01
Hallucination is a feature, not a bug.
When the model doesn't know, it produces plausible-sounding text anyway — because plausibility is what it was trained to optimize. Confident citations to papers that don't exist, protocols with invented temperatures, statistics with invented p-values.
02
Knowledge cutoffs are real.
Each model has a date after which it knows nothing. Ask about a paper published last month and it may invent one rather than admit ignorance. Some models can search the web to compensate; many cannot.
03
The same prompt gives different answers.
LLMs sample probabilistically. Run the same question twice and you may get two different — but equally confident — answers. That's a useful test, and you'll run it in Exercise 1.
AKEYS · GPT 101
Part 03 · Where It Earns Its Keep
15 min
Part 03

What it's
useful for.

Four kinds of work you'll actually do this summer — and where an AI assistant genuinely pulls its weight in each.

AKEYS · GPT 101
10 / Useful · 1 of 2
Reading And Writing

Where AI saves the most time, day to day.

Literature work
  • Summarize a paper you can't read in full — but read the abstract and figures yourself first
  • Compare two papers side by side: "what do these have in common?"
  • Fill vocabulary gaps — "what does ChIP-seq stand for and what is it measuring?"
  • Draft a lit-review outline you then fill in with real reading
Writing
  • Tighten a paragraph — "improve clarity; do not change any factual claims"
  • Turn bench notes into a methods draft — informal to formal scientific writing
  • Catch grammar and structure before your mentor reads it
  • Outline your KEYS presentation: question → methods → results → significance
AKEYS · GPT 101
11 / Useful · 2 of 2
Code And Thinking

Even with no coding background, this is in reach.

Code & data analysis
  • Generate starter code for plots, stats tests, or data cleaning in Python or R
  • Debug an error — paste the full traceback and ask what's going wrong
  • Explain unfamiliar code your mentor wrote, line by line
  • Convert data formats — CSV → JSON, FASTA → table, and so on
Thinking
  • Brainstorm experiment alternatives — "five other ways someone might test this"
  • Walk through a hard concept at high-school level, then grad-student level
  • Generate hypotheses — then you test them in lab, not the AI
  • Steel-man objections — "what's the strongest critique of my approach?"

A chat interface like genai.arizona.edu is the right starting point. For deeper AI-assisted coding, see the Vibe Coding lesson.

AKEYS · GPT 101
12 / In The Wild
Real Life-Sciences AI You Should Be Able To Name

Not chatbots — the systems reshaping biology right now.

AF
AlphaFold

DeepMind. Protein structure prediction from sequence. The 2024 Chemistry Nobel.

ESM
ESM / ESMFold

Meta. Protein language models — structure and function straight from amino-acid sequence.

Rx
Drug discovery at scale

Recursion · Isomorphic · Insilico. AI generates and screens candidate molecules.

FDA
Cleared AI devices

Hundreds of FDA-cleared AI/ML medical devices — mostly in medical imaging.

DNA
Genomics foundation models

Evo · Borzoi. Systems that learn directly from DNA sequence at scale.

The pattern

AI assists hypothesis & pattern recognition. Experiments validate.

AKEYS · GPT 101
13 / The Loop
The Unifying Pattern

AI assists hypothesis generation and pattern recognition. Experiments validate.

A model that predicts a protein structure or proposes a candidate molecule still needs wet-lab or clinical confirmation before anyone trusts it. Your KEYS project is the experiment side of that loop.

AKEYS · GPT 101
Part 04 · What It's Risky For
10 min
Part 04

Where it
will burn
you.

Four failure modes — and the U of A policies that exist to protect you and your mentor's research from the worst of them.

AKEYS · GPT 101
15 / Risk · Hallucination
The Failure Mode To Internalize First

It will invent things, confidently, and look right doing it.

"
Citations.

Specific paper titles, authors, journals, and DOIs that look completely real — and are not.

Protocol details.

Concentrations, incubation times, temperatures, buffer recipes — invented out of thin air.

p
Statistics.

Test results, p-values, and sample sizes pulled from nowhere and stated as fact.

The rule
AI-generated facts are not facts until you have verified them against a primary source. You'll prove this to yourself in Exercise 2.
AKEYS · GPT 101
16 / Risk · Data Privacy
Where Your Data Goes When You Paste It

A commercial chatbot is a third party. Treat it like one.

Never paste into a commercial chatbot

The hard "do not" list.

  • Unpublished research data from your lab
  • Patient-identifying information. Ever. HIPAA exists
  • Proprietary protocols — confidential or sponsor IP
Use the U of A platform instead

genai.arizona.edu

FERPA-aligned. Your conversations are private and not used to train models. When you need to work with sensitive material, this is where you go.

Working with human-subjects data? Your project has IRB rules. Ask your mentor first.

AKEYS · GPT 101
17 / Risk · Trust
Two More You'll Hit Sooner Than You Think

Disclose your AI use. And don't mistake it for your mentor.

The disclosure problem

Be honest anyway.

Many journals and fellowships now require disclosure of AI use. Your KEYS deliverable may too — ask before you lean on it heavily.

Research finds that disclosing AI use can make people trust you less. Disclose anyway — undisclosed use discovered later is far worse than the honesty hit.

The replacement temptation

It's a junior collaborator.

AI is not your mentor, not the experimental method, not your scientific judgment. Treat it as a fast junior collaborator — occasionally wrong, with no stake in the outcome.

You still own the conclusions. Your name goes on the poster.

AKEYS · GPT 101
Part 05 · Hands-On
20 min
Part 05

Open your
laptop.

Four exercises, five minutes each, on genai.arizona.edu. Log in with your NetID. The point is to feel what we just covered — especially the hallucination test.

AKEYS · GPT 101
19 / The Platform
genai.arizona.edu

One login. Many frontier models. Private.

An OpenWebUI interface that gives you free access to multiple frontier models — Claude, OpenAI, Gemma, Meta Llama, Amazon Nova — through a single chat window. Conversations are private and are not used to train models.

Step 0
Open genai.arizona.edu · log in with your UA NetID · find the model selector at the top of the chat.
Model selector

Switch models from the dropdown at the top — central to Exercise 1.

Private by default

FERPA-aligned; your chats are not training data.

Free for UA

No personal API key, no subscription. Just your NetID.

AKEYS · GPT 101
20 / Exercise 1
EX 15 min · Compare two models

Ask the same research question twice — to two different models.

Pick a topic from your KEYS lab. Ask one model the prompt on the right. Then switch models (selector at the top) and ask the exact same prompt.

Reflection
  • Did the two models agree?
  • Did either say something specific you couldn't immediately verify?
  • Which response would you actually use to prep for day one — and why?
prompt · paste into genai.arizona.edu
I'm a high school student starting a research project in <your lab's area — e.g. CRISPR base editing, cancer-cell migration, microbiome data analysis>. Explain in plain language: 1. What is the main experimental approach in this field? 2. What are the most common pitfalls a new student makes? 3. What should I ask my mentor about before I touch anything in the lab?
AKEYS · GPT 101
21 / Exercise 2 · The Important One
EX 25 min · Test for hallucination

Ask for three citations. Then check whether they exist.

Ask the model for three papers supporting a claim from your field. Then paste each DOI into doi.org. Some resolve to the paper described. Some redirect elsewhere. Some will not exist at all.

Reflection
  • How many of the three were real?
  • Could you tell which were fake just from how they looked?
  • If you'd put these in a poster unchecked — what would have happened?
prompt · then verify every DOI
List three peer-reviewed papers from the last five years that show <a specific claim relevant to your KEYS project — e.g. "off-target effects of CRISPR-Cas9 are reduced by high-fidelity variants">. For each: authors, year, journal, DOI. # Now paste each DOI into doi.org and see what actually resolves.
AKEYS · GPT 101
22 / Exercise 3
EX 35 min · Protocol explainer

Have the AI walk you through a real protocol from your lab.

Paste a short excerpt of a protocol you're learning this week — nothing your mentor flagged as confidential. Ask it to flag risks and silent-failure points.

Reflection
  • Did the explanation match what your mentor told you?
  • Where did it differ — and which version do you trust?
  • Anything it flagged as risky that your mentor hadn't? Bring it back to them.
prompt · paste a non-confidential excerpt
I'm a high school intern who has never run this protocol before. Walk me through it step by step, flagging: - Any safety concerns I should ask my PI about - Steps that could fail silently (give a wrong answer without an obvious error) - Reagents or instruments where a mistake would be expensive
AKEYS · GPT 101
23 / Exercise 4 · Take This Home
EX 45 min · Write a CRAFT prompt

Build one structured prompt for a real task this week.

Context · Role · Action · Format · Tone. Run it, then change one piece and re-run — notice how much output quality shifts from small changes.

Take it with you
  • Save the version that worked best.
  • You'll use variations of it all summer.
  • This is the one deliverable from today you keep.
skeleton · the CRAFT framework
[CONTEXT] I am a KEYS intern in <lab / area>. My project involves <one sentence>. This week I'm working on <task>. [ROLE] Act as a <senior grad student / staff scientist> who is patient with the basics. [ACTION] <draft an outline / explain this / suggest 3 plot types> [FORMAT] <numbered list / table / paragraph> [TONE] Clear and direct. Flag anything you're uncertain about rather than inventing details.
AKEYS · GPT 101
24 / Keep Going
When You Have Time This Summer

Where to go deeper after today.

Prompt Engineering
CRAFT, few-shot learning, chain-of-thought, prompt chaining
AI in Research
Deep-research workflows, literature synthesis, hypothesis generation
Vibe Coding
AI-assisted coding for data analysis and computational projects
Ethics & Bias
Nelson's "civic grammar," Magnifica Humanitas, and bias in medical & biological AI
Hands-on labs
Public Health AI Lab · Vibe-Coding a Public Health Map (full agentic build)
UA resources
responsibleai.arizona.edu · libguides.library.arizona.edu/genAI
AKEYS · GPT 101
End · Go Build
One hour. One habit that matters most.

Verify
everything.

AI is a fast junior collaborator — useful, occasionally wrong, no stake in your results. Let it draft, summarize, and explain. Then you read the abstract, check the DOI, run the experiment, and put your name on the poster.

genai.arizona.edu
Log in with your NetID · keep the prompt that worked
BIO5 KEYS Research Internship
University of Arizona · Summer 2026