A.I (Artificial Intelligence)
for the rest of us
Prosper Otemuyiwa | Google DevFest Nairobi 2023
Slide 2
Slide 2 text
PROSPER OTEMUYIWA
- Jollof Rice Advocate
- Open Source Engineer
- Core Team, Novu
- Co-founder, Eden Life
- Serial startup Investor
Slide 3
Slide 3 text
WHOA!
NLP
Backward
Chaining
Deep
Learning
LangChain
LLM
Cognitive
Computing
Forward
Chaining
Generative AI
Neural
network
AGI
ANI
Transformer
ChatGPT
RLHF
Zero-shot
learning
Tensorflow
Gradient
Descent
CNN
Diffusion RAG
Keras
llama
PaLM
diffusion
Slide 4
Slide 4 text
Why Me, Lord? Sielewi!
😓
Slide 5
Slide 5 text
TO BE AN A.I/M.L ENGINEER
…there are some general assumptions
Slide 6
Slide 6 text
01
Statistician PHD
Mathematician
02 03
Be an advanced numbers guru /
highly credentialed.
Slide 7
Slide 7 text
Be incredibly smart / genius
Slide 8
Slide 8 text
What About The
Rest of Us?
Slide 9
Slide 9 text
WE JUST WANT TO BUILD,
SHIP & ENHANCE SOFTWARE
not genius /
incredibly smart
not
AI-specialized
not highly
credentialed
Slide 10
Slide 10 text
There’s Hope for Us!
…I promise
Slide 11
Slide 11 text
There’s Hope for Us!
Let’s get to work
Slide 12
Slide 12 text
The AI API Approach
❖ Frontend think in Components
❖ Backend think in Services
❖ DevOps think in configs & orchestration
Slide 13
Slide 13 text
KNOW THE
VOCAB & BASICS
01 Learn the glossary of A.I,
understand the terms so you can
leverage the most of it!
Slide 14
Slide 14 text
Glossary & Vocabulary
Generative AI - AI focused on generating new and
original content, images, text, music etc based on
patterns from existing data.
LLM - A type of AI model that can generate
human-like text and is trained on a broad dataset.
Natural Language Processing - AI focused on
interaction between humans and computers
through natural language.
Slide 15
Slide 15 text
Leverage Generative AI &
Large Language Models
via Cloud PLATFORMS
02
Discover and choose the models best suited for
your needs from any of these platforms!
Slide 16
Slide 16 text
REPLICATE.COM
Run Models in the Cloud at Scale
Slide 17
Slide 17 text
REPLICATE.COM
Explore thousands of models at scale
Slide 18
Slide 18 text
REPLICATE.COM
Use them at scale via simple API calls
Slide 19
Slide 19 text
OPENAI.COM
Explore Models at Scale via APIs
GPT-3.5 & 4 - Understand &
generate natural language or code
DALL-E - Generate & edit images
with a natural language prompt.
Whisper - Convert audio to text.
Embeddings - Text search, Code search,
recommendations
Slide 20
Slide 20 text
OPENAI.COM
Explore Models at Scale via APIs
Slide 21
Slide 21 text
OPENAI.COM
Explore Models at Scale via APIs
labs.openai.com
ANTHROPIC CLAUDE - ANTHROPIC.COM
Explore Models at Scale via APIs
CLAUDE - State of the art
high performance model
CLAUDE INSTANT -
Lighter, less expensive
and much faster
Slide 24
Slide 24 text
ANTHROPIC CLAUDE - ANTHROPIC.COM
Explore Models at Scale via APIs - https://api.anthropic.com/v1/
Completion - /complete
JS
Slide 25
Slide 25 text
ANTHROPIC CLAUDE - ANTHROPIC.COM
Explore Models at Scale via APIs - https://api.anthropic.com/v1/
Completion - /complete
Python
Slide 26
Slide 26 text
GOOGLE AI
Explore Google’s PaLM 2 LLM
Slide 27
Slide 27 text
GOOGLE AI
Generative AI via the Vertex PaLM API
ImageGen - Generate & customize images
- imagegeneration
Chirp - Speech to Text. Transcribe audio easily
Codey - Generate and suggest code
(completion)
- code-bison - code generation
- codechat-bison - code chat
- code-gecko - code completion
Slide 28
Slide 28 text
GOOGLE AI
Generative AI via the Vertex PaLM API
- Grab an API Key from Google Maker Suite
- https://makersuite.google.com/app/apikey
Slide 29
Slide 29 text
GOOGLE AI
Generative AI via the Vertex PaLM API
- Text Generation - npm install @google-ai/generativelanguage
Slide 30
Slide 30 text
GOOGLE AI
Generative AI via the Vertex PaLM API
- Text API example and source code
Source code: https://github.com/google/generative-ai-docs/tree/main/demos/palm/web/list-it
Slide 31
Slide 31 text
GOOGLE AI
Generative AI via the Vertex PaLM API
- Chat API example and source code
Source code: https://github.com/google/generative-ai-docs/tree/main/demos/palm/web/quick-prompt
Slide 32
Slide 32 text
Google Model Garden
Explore all sort of Foundational and pre-trained models
https://console.cloud.google.com/vertex-ai/model-garden
Slide 33
Slide 33 text
The Everyday A.I Tools!
03 Be acquainted with the everyday A.I tools. Know
them, bask in them & supercharge your learning
& everyday task!
Slide 34
Slide 34 text
PHIND.COM - A.I Search Engine &
Pair Programmer
Slide 35
Slide 35 text
MIDJOURNEY.COM - Discord bot to create
unique imagery from short text descriptions
/imagine african engineer in coding paradise
Slide 36
Slide 36 text
CHAT.OPENAI.COM - ChatGPT, A large scale AI
Language model that generates human-like
text through conversations.
Slide 37
Slide 37 text
GitHub Copilot - Very awesome AI pair
programmer. Works like a charm! So good!
Slide 38
Slide 38 text
Angry Pumpkins made with GPT-4 (for coding)
and Midjourney / DALLE for the graphics
https://x.com/javilopen/status/171936358735174071
1?s=46&t=xc4I0Z6yfWMu3mwDrPEmNQ
Slide 39
Slide 39 text
HUGGINGFACE.CO - A.I community : Hundreds of
members, models & datasets. Enjoy collaboration
Slide 40
Slide 40 text
Like Bob the Builder,
Build & Supercharge
your Web App!
04
Learn to speedily supercharge your web apps
with AI features. Build. Ship. Deploy
Slide 41
Slide 41 text
Ship in 24, 48, 72 hours? Let’s go!
Your Project
DAY 1
Integrate the
appropriate
model via API
DAY 1
Choose an AI
Cloud
Provider
DAY 1
Deploy & Tell
the World
about it
DAY 1
Next.js, Python,
React, Node, PHP,
Ruby
Slide 42
Slide 42 text
Be like @Nutlope - The Weekend AI Engineer
URL: roomgpt.io
Source code
(GitHub):
https://github.com
/Nutlope/roomGPT
Slide 43
Slide 43 text
Be like @Nutlope - The Weekend AI Engineer
URL:
restorephotos.io
Source code
(GitHub):
https://github.com
/Nutlope/restorePh
otos
Slide 44
Slide 44 text
Be like @Nutlope - The Weekend AI Engineer
URL: qrgpt.io
Source code
(GitHub):
https://github.com
/Nutlope/qrGPT
Slide 45
Slide 45 text
“A.I gives you great power.
And with great power
comes great responsibility.”
Slide 46
Slide 46 text
THANK YOU VERY MUCH!
Twitter & GitHub - @unicodeveloper