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
OpenUI AI Application Journey
Search
Ryo Kanazawa
July 24, 2024
110
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
OpenUI AI Application Journey
Ryo Kanazawa
July 24, 2024
More Decks by Ryo Kanazawa
See All by Ryo Kanazawa
メタバース研究所による機械学習研究(W&Bミートアップ #15)
ryokanazawa1227
0
190
AIキャラクター向けLLMの開発
ryokanazawa1227
0
270
Featured
See All Featured
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.3k
Scaling GitHub
holman
464
140k
AI: The stuff that nobody shows you
jnunemaker
PRO
9
930
Organizational Design Perspectives: An Ontology of Organizational Design Elements
kimpetersen
PRO
1
800
How To Stay Up To Date on Web Technology
chriscoyier
790
250k
Measuring Dark Social's Impact On Conversion and Attribution
stephenakadiri
2
260
A Soul's Torment
seathinner
6
3.5k
How to Create Impact in a Changing Tech Landscape [PerfNow 2023]
tammyeverts
56
3.4k
GraphQLとの向き合い方2022年版
quramy
50
15k
Performance Is Good for Brains [We Love Speed 2024]
tammyeverts
12
1.8k
Efficient Content Optimization with Google Search Console & Apps Script
katarinadahlin
PRO
1
800
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
659
62k
Transcript
Building an AI Application with Weave and the future of
software development
2 Introduction Chris Van Pelt (CVP) Co-founder / web developer
/ AI & space enthusiast
3 What if I could just ask an LLM to
generate UI components for me? 🤔 CVP
Tailwind CSS 💞 LLMʼs • Compact • Inline • Popular
4 Replace box with image
5 wandb.me/openui
What does the future of software development look like? 6
7 cursor.com
8 github.com
How do you make LLMs work in applications? 9
10 Recipe for making LLMs work 1. Experiment 2. Track
everything 3. Inspect individual examples 4. Measure progress
11 Recipe for making LLMs work 1. Experiment 2. Track
everything 3. Inspect individual examples 4. Measure progress
12 Recipe for making LLMs work 1. Experiment 2. Track
everything 3. Inspect individual examples 4. Measure progress
13 Recipe for making LLMs work 1. Experiment 2. Track
everything 3. Inspect individual examples 4. Measure progress
14 332,688,821 Weave
Weave Traces Powerful, data-oriented tracing for AI apps with a
single line of code
16 Demo Time 🥳
Weave Evaluations Full fidelity measurements for your models so you
can go fast
How would I evaluate a use case like this which
has infinite possible “right” answers? 18 👀
Browser Automation • Generate HTML with a model, prompt &
temperature • Render the HTML in a browser and take a screenshot • Send the screenshot to an LLM with 👀 capabilities and ask for ratings 19 Replace box with image
20
21 Demo Time 🚀
22 332,688,821 wandb.me/weave