Tweet
Tweet

Slide 1

Slide 1 text

Ines Montani Explosion LARGE LANGUAGE LARGE LANGUAGE MODELS ✨ CHATGPT " ARTIFICIAL INTELLIGENCE # MACHINE LEARNING ✨ PROTOTYPE TO PRODUCTION MODELS FROM LLAMA $ NATURAL LANGUAGE PROCESSING % ✨ OPEN SOURCE & PYTHON ' PROMPT ENGINEERING ⚙ ZERO-SHOT LEARNING ) GPT-4 EVALUATION * COPILOT + GENERATIVE AI , Ines Montani - Explosion

Slide 2

Slide 2 text

SPACY SPACY.IO & @SPACY_IO ✍ SPACY.TV / GITHUB.COM/EXPLOSION/SPACY Open-source library for industrial-strength Natural Language Processing 140m+ downloads

Slide 3

Slide 3 text

SPACY SPACY.IO & @SPACY_IO ✍ SPACY.TV / GITHUB.COM/EXPLOSION/SPACY Open-source library for industrial-strength Natural Language Processing 140m+ downloads ChatGPT can write spaCy code!

Slide 4

Slide 4 text

PRODIGY Modern scriptable annotation tool for machine learning developers PRODIGY.AI & GITHUB.COM/EXPLOSION/PRODIGY-RECIPES 8k+ users 700+ companies

Slide 5

Slide 5 text

PRODIGY Modern scriptable annotation tool for machine learning developers PRODIGY.AI & GITHUB.COM/EXPLOSION/PRODIGY-RECIPES 8k+ users 700+ companies

Slide 6

Slide 6 text

THE HISTORY OF FUTURE TECHNOLOGY

Slide 7

Slide 7 text

THE HISTORY OF FUTURE TECHNOLOGY How people in 1900 imagined the year 2000

Slide 8

Slide 8 text

THE HISTORY OF FUTURE TECHNOLOGY How people in 1900 imagined the year 2000

Slide 9

Slide 9 text

THE HISTORY OF FUTURE TECHNOLOGY

Slide 10

Slide 10 text

THE HISTORY OF FUTURE TECHNOLOGY manual calculation vs. calculator

Slide 11

Slide 11 text

THE HISTORY OF FUTURE TECHNOLOGY manual calculation vs. calculator

Slide 12

Slide 12 text

THE HISTORY OF FUTURE TECHNOLOGY “knocker-uppers” vs. alarm clock manual calculation vs. calculator

Slide 13

Slide 13 text

THE HISTORY OF FUTURE TECHNOLOGY

Slide 14

Slide 14 text

THE HISTORY OF FUTURE TECHNOLOGY human assistant vs. calendar apps Calendly Fantastical

Slide 15

Slide 15 text

THE HISTORY OF FUTURE TECHNOLOGY human assistant vs. calendar apps Calendly Fantastical WHAT’S NEXT?

Slide 16

Slide 16 text

NLP IN THE AGE OF LLMS

Slide 17

Slide 17 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 %

Slide 18

Slide 18 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 "

Slide 19

Slide 19 text

VISION #1 dialogue is all you need % 2 LLM 3 user actions or information natural language input

Slide 20

Slide 20 text

VISION #1 dialogue is all you need % 2 LLM 3 user actions or information natural language input LLM is the system and needs to manage the whole interaction

Slide 21

Slide 21 text

VISION #2 prompting is all you need " 2 LLM 4 text % prompt 5 system 3 user 6 structured data

Slide 22

Slide 22 text

VISION #2 prompting is all you need " 2 LLM 4 text % prompt 5 system 3 user LLM replaces the specific ML model 6 structured data

Slide 23

Slide 23 text

VISION #3 modern practical NLP - 7 developer 8 code 2 LLM 9 training data 5 system 3 user 6 structured data ⚙ ML system

Slide 24

Slide 24 text

VISION #3 modern practical NLP - 7 developer 8 code 2 LLM 9 training data 5 system 3 user 6 structured data ⚙ ML system LLM helps with building the pipeline

Slide 25

Slide 25 text

VISION #3 modern practical NLP - 7 developer 8 code 2 LLM 9 training data 5 system 3 user 6 structured data ⚙ ML system LLM helps with building the pipeline

Slide 26

Slide 26 text

VISION #3 modern practical NLP - 7 developer 8 code 2 LLM 9 training data 5 system 3 user 6 structured data ⚙ ML system LLM helps with building the pipeline

Slide 27

Slide 27 text

COMPANY COMPANY MONEY INVESTOR “Hooli raises $5m to revolutionize search, led by ACME Ventures” 5923214 1681056 CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA Database

Slide 28

Slide 28 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA

Slide 29

Slide 29 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition

Slide 30

Slide 30 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation

Slide 31

Slide 31 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation custom database lookup

Slide 32

Slide 32 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation custom database lookup currency normalization

Slide 33

Slide 33 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation custom database lookup currency normalization entity relation extraction

Slide 34

Slide 34 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation custom database lookup currency normalization entity relation extraction 6 structured data 2 LLM 7 developer quick prototype

Slide 35

Slide 35 text

Database CLASSIC NLP PROBLEM: EXTRACT STRUCTURED DATA named entity recognition entity disambiguation custom database lookup currency normalization entity relation extraction 6 structured data 2 LLM 7 developer quick prototype ⚡ fast to develop, slow to run, hard to improve

Slide 36

Slide 36 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP -

Slide 37

Slide 37 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data

Slide 38

Slide 38 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data humans in the loop

Slide 39

Slide 39 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data fast prototyping humans in the loop

Slide 40

Slide 40 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data fast prototyping humans in the loop powered by open source

Slide 41

Slide 41 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data fast prototyping humans in the loop powered by open source conversational and graphical interfaces

Slide 42

Slide 42 text

NLP IN THE AGE OF LLMS SQL is all you need dialogue is all you need 0 % lots of humans is all you need prompting is all you need 1 " modern practical NLP - structured data fast prototyping humans in the loop powered by open source robust evaluation conversational and graphical interfaces

Slide 43

Slide 43 text

A CASE FOR LLM PRAGMATISM EXPLOSION.AI/BLOG/AGAINST-LLM-MAXIMALISM NOOO YOU CAN'T JUST MIX UP ALL THE STEPS OF YOUR TASK AND ASK AN LLM TO DO IT ALL. HOW WILL YOU EVER MAKE A RELIABLE AND EXTENSIBLE SYSTEM THAT WAY? HAHA LLM GO BRRR

Slide 44

Slide 44 text

A CASE FOR LLM PRAGMATISM EXPLOSION.AI/BLOG/AGAINST-LLM-MAXIMALISM NOOO YOU CAN'T JUST MIX UP ALL THE STEPS OF YOUR TASK AND ASK AN LLM TO DO IT ALL. HOW WILL YOU EVER MAKE A RELIABLE AND EXTENSIBLE SYSTEM THAT WAY? HAHA LLM GO BRRR avoid coupling prediction tasks to arbitrary business logic

Slide 45

Slide 45 text

A CASE FOR LLM PRAGMATISM EXPLOSION.AI/BLOG/AGAINST-LLM-MAXIMALISM NOOO YOU CAN'T JUST MIX UP ALL THE STEPS OF YOUR TASK AND ASK AN LLM TO DO IT ALL. HOW WILL YOU EVER MAKE A RELIABLE AND EXTENSIBLE SYSTEM THAT WAY? HAHA LLM GO BRRR avoid coupling prediction tasks to arbitrary business logic design modular solutions

Slide 46

Slide 46 text

A CASE FOR LLM PRAGMATISM EXPLOSION.AI/BLOG/AGAINST-LLM-MAXIMALISM NOOO YOU CAN'T JUST MIX UP ALL THE STEPS OF YOUR TASK AND ASK AN LLM TO DO IT ALL. HOW WILL YOU EVER MAKE A RELIABLE AND EXTENSIBLE SYSTEM THAT WAY? HAHA LLM GO BRRR avoid coupling prediction tasks to arbitrary business logic design modular solutions prototype modules with LLMs

Slide 47

Slide 47 text

A CASE FOR LLM PRAGMATISM EXPLOSION.AI/BLOG/AGAINST-LLM-MAXIMALISM NOOO YOU CAN'T JUST MIX UP ALL THE STEPS OF YOUR TASK AND ASK AN LLM TO DO IT ALL. HOW WILL YOU EVER MAKE A RELIABLE AND EXTENSIBLE SYSTEM THAT WAY? HAHA LLM GO BRRR avoid coupling prediction tasks to arbitrary business logic design modular solutions prototype modules with LLMs evaluate alternatives

Slide 48

Slide 48 text

TRADE-OFFS performance on the bar exam kentlaw.iit.edu

Slide 49

Slide 49 text

TRADE-OFFS performance on the bar exam kentlaw.iit.edu OpenAI API latency promptlayer.com

Slide 50

Slide 50 text

TRADE-OFFS Supervised 1 LLM 2 accuracy words/s accuracy words/s Textcat on SST2 (Stanford Sentiment Treebank) 0.9 4019 0.9 <100 Textcat on Banking77 (intent recognition) 0.9 3234 0.7 <100 NER on AnEm (anatomical entity mentions) 0.7 5146 0.1 <100 1. RoBERTa-base with spaCy, 2. text-davinci-003 zero-shot ongoing experiments comparing LLMS to task-specific models performance on the bar exam kentlaw.iit.edu OpenAI API latency promptlayer.com

Slide 51

Slide 51 text

LLM-POWERED NLP IN PRACTICE LLM-powered collaborative data development environment 7

Slide 52

Slide 52 text

LLM-POWERED NLP IN PRACTICE LLM-powered collaborative data development environment 7 Assign labeling tasks to LLMs "

Slide 53

Slide 53 text

LLM-POWERED NLP IN PRACTICE LLM-powered collaborative data development environment 7 Assign labeling tasks to LLMs " Review label decisions, correct errors ;

Slide 54

Slide 54 text

LLM-POWERED NLP IN PRACTICE LLM-powered collaborative data development environment 7 Assign labeling tasks to LLMs " Review label decisions, correct errors ; Tune prompts and compare LLMs empirically 6

Slide 55

Slide 55 text

LLM-POWERED NLP IN PRACTICE LLM-powered collaborative data development environment 7 Assign labeling tasks to LLMs " Review label decisions, correct errors ; Tune prompts and compare LLMs empirically 6 Build data sets to train and evaluate e icient, production-ready pipelines +

Slide 56

Slide 56 text

8 PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS

Slide 57

Slide 57 text

8 correct mistakes PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS

Slide 58

Slide 58 text

PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS correct mistakes

Slide 59

Slide 59 text

add correct answer to prompt to tune it PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS correct mistakes

Slide 60

Slide 60 text

PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS

Slide 61

Slide 61 text

PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS query LLM and parse response

Slide 62

Slide 62 text

PRODIGY.AI/FEATURES/LARGE-LANGUAGE-MODELS query LLM and parse response tune prompt if needed

Slide 63

Slide 63 text

GITHUB.COM/EXPLOSION/SPACY-LLM TOWARDS STRUCTURED DATA Prompt Template 2 LLM London is bigger than Berlin LOCATION: London, Berlin LOCATION

Slide 64

Slide 64 text

GITHUB.COM/EXPLOSION/SPACY-LLM TOWARDS STRUCTURED DATA Prompt Template 2 LLM London is bigger than Berlin LOCATION: London, Berlin LOCATION

Slide 65

Slide 65 text

SPECIFIC Task-specific models powered by LLMS IS BETTER.

Slide 66

Slide 66 text

SMALLER & FASTER Task-specific models powered by LLMS IS BETTER.

Slide 67

Slide 67 text

PRIVATE Task-specific models powered by LLMS IS BETTER.

Slide 68

Slide 68 text

BETTER Task-specific models powered by LLMS IS BETTER.

Slide 69

Slide 69 text

THANK YOU! - Explosion – explosion.ai < spaCy – spacy.io ✨ Prodigy – prodigy.ai = Twitter – @_inesmontani > Mastodon – @[email protected] ? LinkedIn