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
0
49
OpenUI AI Application Journey
Ryo Kanazawa
July 24, 2024
Tweet
Share
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
140
AIキャラクター向けLLMの開発
ryokanazawa1227
0
180
Featured
See All Featured
Six Lessons from altMBA
skipperchong
27
3.6k
Testing 201, or: Great Expectations
jmmastey
42
7.2k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
251
21k
The Invisible Side of Design
smashingmag
299
50k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
32
2.1k
Fight the Zombie Pattern Library - RWD Summit 2016
marcelosomers
233
17k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
27
1.9k
The World Runs on Bad Software
bkeepers
PRO
67
11k
Save Time (by Creating Custom Rails Generators)
garrettdimon
PRO
29
1k
Building Applications with DynamoDB
mza
93
6.2k
Sharpening the Axe: The Primacy of Toolmaking
bcantrill
40
2k
Fashionably flexible responsive web design (full day workshop)
malarkey
406
66k
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