prompting the 
ai revolution: exploring the world of large language models

>

_

bearcat days '24

https://slides.com/

lucascordova/

bearcat-days

professor lucas cordova

software engineering, cs education, artificial intelligence, social media

high-level research interests

school of computing and information sciences

teaching

programming, data structures, mobile app development, algorithms

contact me or stop by!

lpcordova@willamette.edu

|

ford hall room 210

☕️

setting the stage

ai and large language models
in realms of code and light, where ai minds take flight, language models vast, with wisdom's tome amassed.
they speak in tongues so wide, 
a bridge 'twixt worlds inside,
a ghost within the wires, fulfilling our desires. yet ponder we the role of these digital souls, servants or masters, in future's unfolding scrolls. with ethics in our grasp, a future we can clasp,
Where man and machine, in harmony, can last.
prompt: i am a cs professor. write me a poem in shakesperean about ai and llms.

setting the stage

which of these are real people?

what is a large language model?

why now?

spell checkers

familiar...

LLM is like a superpowered translator that can understand really complex text.

💪🏼

LLMs are trained on massive amounts of text, data (books, articles, code), and can respond to complex language.

🤖

so what? 👀

what can I use them for?

ways

to use LLMs

ethical

academic advice

Prompt: “As a [Major/Year] student, I need guidance on [Specific Course/Topic]. Provide me with a comprehensive overview, recommended study resources, and strategies to excel in this area.”

time mgt help

Prompt: “I’m a [Year] student overwhelmed with balancing classes, assignments, and a part-time job. Create a personalized weekly schedule that optimizes my time, considering my class timings, study preferences, and work hours.”

career path exploration

Prompt: “I’m an [Major] student unsure about my career options. Generate a list of potential career paths within my field, including job descriptions, required skills, and insights from professionals in those roles.”

research paper support

Prompt: “Assist me in developing a research question for my [Course/Topic] paper. Provide a refined question, key subtopics, and suggestions for credible sources to kickstart my research.”

exam prep toolkit

Prompt: “Assist me in developing a research question for my [Course/Topic] paper. Provide a refined question, key subtopics, and suggestions for credible sources to kickstart my research.”

presentations

Prompt: “I have a [Course] presentation coming up. Assist me in refining my slides, structuring my content, and providing key talking points to deliver a compelling and informative presentation.”

activity 1

scenario: you have applied for a position on campus and are going to be interviewed. the position is for a bearcat days tour guide. use a LLM to prepare for the interview.

sample prompt

I am applying for a position on-campus as a tour guide. help me with interview questions. ask me a question and then tell me how i did before asking me the next question.

activity 2

problem: incoming freshmen and you don't know what subjects would be good to take or that you would be good at.

come up with a prompt.

what is artificial intelligence?

why now?

An entity that performs behaviors that a person might reasonably call intelligent if a human were to do something similar.

artificial intelligence

 👀

A means by which to create behavior by taking in data, forming a model, and then executing the model.

machine learning

 👀

a model is a simplification of some complex phenomenon, e.g. a model car.

model

 👀

there are many ways to learn a model from data; one such way is a neural network which are modeled to work like the brain.

learning from data

self-driving cars

self-driving cars

self-driving cars

self-driving cars

“Once upon a ____”

language models

gpt: generative pre-trained transformer

gpt

generative: model that is capable of generating continuations to the provided input.

pre-trained: model is trained on a very large amount of data.

transformer: encoder/decoder to predict a word.

process of refining prompts that a person can input into a generative ai service to create text or images.

prompt engineering

  1. persona pattern
    
  2. audience personal pattern
    
  3. game-play pattern
  4. question-refinement pattern
    
  5. few-shot

prompt types

I need you to be a talent scout career advisor and help me identify a few professional organizations which I can join to burnish my Data Science credentials. 

Also, take some time to generate the response, do let me know if they've any volunteer opportunities where I can use my skills from MS in Data Science to contribute to ongoing projects. 

Concurrently, provide me the links so that I may do my own due diligence.

persona pattern

Explain Large-Language Models to me. Assume I am Abraham Lincoln.

Explain Jujutsu Kaisen to me. Assume I am Shakespeare.

Explain the significance of Transformers for Large-Language Models and NLP in general. Assume I am a fashion designer.

audience persona pattern

Going to play a game involving prompt engineering. The underlying goal of the game is to improve my prompt engineering skills. 

You will ask me a question or give me a task, and I have to come up with a suitable prompt to solve the said task. 

The tasks or questions should have a logical/critical reasoning or programming aspect. However, we're not writing a full-fledged source code for task involving a coding requirement. 

After my prompt, you are to evaluate my prompt and assign it a score from 0-10. Concurrently, you'll also suggest improvements to improve my prompting skills. 
You will continue asking me questions until I explicitly tell you to stop.

If everything is clear, proceed to ask/assign me the first question/task.

game-play pattern

I am going to ask you a question. Before answering the question, suggest a better version of the question and ask me if it looks good. 

Continue this process until I sign-off on the question. 
After receiving the sign-off, proceed to comprehensively answer the question.

question-refinement pattern

Consider the following examples:
Claim: When it comes to modern day anime, Jujustsu Kaisen has no equal.
Verdict: Subjective
Claim: 2+2 =5
Verdict: False
Claim: Anything multiplied by zero becomes zero.
Verdict: True
My next prompt will follow the pattern mentioned above by providing you with a claim and you're going to come up with a verdict. Is that clear?

 

Consider the following example:
Action: Applying for a job.
Check Requirements: Do you have a Resume? Do you have a cover letter?
After check: If yes, proceed to application. If no, prepare materials mentioned in "Check Requirements"
Additional steps: Enhance the materials to increase the probability of success.
Result: Successful placement
In the next prompt, I'll provide either an action or a scenario, and you'll respond accordingly.

few-shot pattern

  1. misinformation and deepfakes
  2. bias and discrimination
  3. copyright and intellectual property
  4. privacy and data security

ethical implications

pair with one or two people (family and/or friends)

 

try the prompt exercises in the handout.

you try!

professor lucas cordova

school of computing and information sciences

contact me or stop by!

lpcordova@willamette.edu

|

ford hall room 210

☕️

thank you for attending!

Bearcat Days

By Lucas Cordova

Bearcat Days

  • 39