0% found this document useful (0 votes)
605 views15 pages

Steamship Harvard Cs 150 Slides

The document discusses building AI applications with large language models like GPT. It introduces GPT and how people are using it to build chatbots, question answering systems, companionship applications, and more. Common applications include question answering, companionship, and creativity.

Uploaded by

Daman Brar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
605 views15 pages

Steamship Harvard Cs 150 Slides

The document discusses building AI applications with large language models like GPT. It introduces GPT and how people are using it to build chatbots, question answering systems, companionship applications, and more. Common applications include question answering, companionship, and creativity.

Uploaded by

Daman Brar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Building

AI Apps
CS50 @ Harvard
Ted Benson | @GetSteamship | steamship.com/discord
Sil Hamilton | @srhm_ca | srhm.ca

Aside: Yesterday’s Hackathon: steamship.com/hackathon/harvard

Build Build and deploy an AI


bot with personality,
Your skills, and memory.

Bot

Agenda

• What is GPT?
• How are people building with it?

What is GPT?
What is GPT?

Large Language Model Universal approximator Generative AI

Neural Network Arti icial intelligence Simulator

Text predictor
Writing assistant Content generator

Agent
Chatbot Transformer
f

Language Models

Produce a probability distribution


over some vocabulary
• What follows “I am…”?
• Vocabulary = all words
• Can sample predicted word
repeatedly to generate text

Instruction Tuning

Natural language as an interface


• Dependable input/output is hard
• OpenAI InstructGPT (2022)
• Lead to ChatGPT, GPT-4
• Solid foundation to build on

How are people


building AI Apps?
Common GPT apps we’re seeing

In production today Experimental


• Companionship • Multi-step task agents
• Question Answering • Hyper-personalization
• Utility
• Creativity

Companionship

Personality
Prompt
In Prompt Instructions
Wrapper
In Prompt “Database”

GPT

Demo: Mandarin Idiom Coach Replit


Question Answering “You are an expert at answering
questions.

Please answer <question>


Vector DB
Using source documents <docs>”

Query Document Prompt


Fragment Search
Wrapper

GPT
Demo: Ask my Course
Loaded with CS50 Syllabus GitHub

Utility functions
Low hanging fruits — automate tasks that require
basic language understanding to perform.

• Generate a unit test for this function


• Scrape & summarize this website
• Tell me which Tweets to read

Incredible startup & personal project potential


Creativity
DB
Article, Article,
Idea, GPT Idea,
Etc Etc
Candidate
Output
Python
Domain
Knowledge

Demo: Reading Suggestions GitHub


Multi-step Planning (AutoGPT, BabyAGI)

GPT GPT Result


“Generate
“Build a steps to Perform Output of
<steps>
company” build a each step inal step
company”

Code
f
Building
AI Apps
CS50 @ Harvard
Ted Benson | @GetSteamship | steamship.com/discord
Sil Hamilton | @srhm_ca | srhm.ca

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy