Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Fergal Reid - Building products in the age of Ai
Search
Turing Fest
PRO
July 05, 2023
Technology
0
81
Fergal Reid - Building products in the age of Ai
Turing Fest
PRO
July 05, 2023
Tweet
Share
More Decks by Turing Fest
See All by Turing Fest
Anika Zubair - Customer Success for Revenue Growth: People, Process, Tools, and Repeatability
turingfest
PRO
0
60
Anna Shipman - Analysing, Deciding, Doing
turingfest
PRO
0
77
Annjana Ramesh - Inclusion as a Product
turingfest
PRO
0
74
April Dunford - Using Strategic Positioning to Unlock Growth in Noisy Markets
turingfest
PRO
0
230
Bastian Grimm - The Rise of AI
turingfest
PRO
0
180
Bob Moesta - Should I Stay or Should I Go?
turingfest
PRO
0
140
Chris Savage - Buying Out Your Own Company
turingfest
PRO
0
69
Christine Itwaru - Embracing Product Operations
turingfest
PRO
0
110
Emily Grossman - How to Grow
turingfest
PRO
0
190
Other Decks in Technology
See All in Technology
Handling focus in 2024
tahia910
0
230
Rustで「プリズモイダル法」を利用して「土量計算」をガチでやる
nokonoko1203
1
310
一生覚えておきたい「システム開発=コミュニケーション」〜初めての実務案件振り返りLT〜
maimyyym
2
330
よく聞くけど使ったことないソフトウェアNo.1 KafkaとSnowflake
foursue
4
520
プロンプトエンジニアリングでがんばらない-Agentic Workflow へ-近藤憲児
kenjikondobai
6
1.2k
AWS学習者向けにAzureの解説スライドを作成した話
handy
3
200
Tellus の衛星データを見てみよう #mf_fukuoka
kongmingstrap
0
280
Building Dashboards as a Hobby
egmc
0
380
Google Cloud Next '24 Recap(Cloud Run/k8s)
mokocm
0
340
Python と Snowflake はズッ友だょ!~ Snowflake の Python 関連機能をふりかえる ~
__allllllllez__
2
140
Microsoft Intune 勉強会 第 2 回目
tamaiyutaro
2
460
ルーターでプレゼンする
puhitaku
1
3.3k
Featured
See All Featured
GitHub's CSS Performance
jonrohan
1025
450k
The World Runs on Bad Software
bkeepers
PRO
61
6.7k
Typedesign – Prime Four
hannesfritz
36
2.1k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
66
14k
Ruby is Unlike a Banana
tanoku
96
10k
The Invisible Customer
myddelton
114
12k
Music & Morning Musume
bryan
41
5.6k
Optimizing for Happiness
mojombo
370
69k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
358
22k
A Modern Web Designer's Workflow
chriscoyier
689
190k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
155
14k
The Cost Of JavaScript in 2023
addyosmani
21
3.9k
Transcript
Building products in the Age of AI @fergal_reid
GPT / LLMs • Internet sized change • Change in
capability • Change in how we build and use AI
None
None
None
Level 1: GPTs are incredible! Level 2: GPTs make things
up and aren’t trustworthy. Level 3: GPTs can be incredible when used right
See them as engineering components Separate out aspects accidentally bundled
What is GPT?
Training objective: token prediction
Training objective: token prediction
None
None
• A sequence model • That uses ‘attention’ • Gradient
descent
• A sequence model • That uses ‘attention’ • Gradient
descent
Not a useful model • Human = genes and evolution
? • Distrust: ‘It de fi nitely can’t do X because its just trained to predict the next word’
Model: Database + Reasoning Engine • The reasoning engine is
key • Often, the database is a liability
Reasoning capabilities
None
None
Model: ‘Interpolative’ vs ‘Extrapolative’ tasks
None
• Less reliable at extrapolation • Favour interpolation • Perform
a task, given a context • ‘Retrieval Augmented Generation’
Model: Human intuition Ask a human to answer a historical
question vs Give them a history book and ask them the question
Note: Context window limited • Thousands of words • Can’t
put a whole KB, or context, in it • Synergizes well with Vector Search
How we build with GPTs
None
None
None
None
30 November 2022: ChatGPT
First features we built • Summarisation • Edit tone of
voice • Expand from shorthand
None
None
• 5th Dec: Rolling • 20th Dec: Internal use •
~13th Jan: Customer beta • 31st Jan: Launch with testimonials Timeline
Model: Easy vs Hard AI features
• ‘Easy’: • Out-of-box accuracy high • Cost of error
low • E.g. ‘Draft me a summary’
• ‘Hard’: • Out-of-box accuracy low • Cost of error
high
Development Tactics
• Fast customer contact • Assume you can build v1
of most ML with powerful LLM • Make cheap later • “LLMs aren’t all of AI” • How we build software has changed
Hard feature: Fin • GPT-powered question answering Bot
• An LLM can seem inert • However, can easily
be turned into an agent
My key points • Internet sized change • Good model:
DB+Reasoning • Changes how we build ML • Feature dif fi culty varies
Guessing what’s next
• V1: text tools, working around clunky interfaces • V2:
features reasoning can enhance • V?: End to end problems where intelligence can help • Don’t underestimate the reasoning capability, very sophisticated
• Breakneck progress • Smaller models, open? • Exciting but
overhyped today • Productisation • Larger models
Thank you! @fergal_reid