Kabir Walia (@kabir__walia) 's Twitter Profile
Kabir Walia

@kabir__walia

ID: 1144383697816346624

calendar_today27-06-2019 23:15:09

22 Tweet

98 Takipçi

1,1K Takip Edilen

Kabir Walia (@kabir__walia) 's Twitter Profile Photo

Came across Netflix’s “Chaos Monkey” tool netflixtechblog.com/the-netflix-si…. They built a tool that randomly breaks components of their cloud infra to find new weakness and build fault tolerance. My favorite streaming service for a reason 🫶

@levelsio (@levelsio) 's Twitter Profile Photo

GPT4 being free for everyone could pump all AI startups in a massive way As if billions of users are suddenly onboarded onto AI and "get it" This is kinda related to the Starbucks effect: When a Starbucks opens up in a neighborhood, local indie coffee shops were expected to

GPT4 being free for everyone could pump all AI startups in a massive way

As if billions of users are suddenly onboarded onto AI and "get it" 

This is kinda related to the Starbucks effect:

When a Starbucks opens up in a neighborhood, local indie coffee shops were expected to
Kabir Walia (@kabir__walia) 's Twitter Profile Photo

Autopilots (autonomous agents) are a big part of the future of AI-based human computer interaction. But in the short/medium term, Copilots will continue to win customers. Customers are most excited and ready to commit when the AI is auditable and controllable.

Kabir Walia (@kabir__walia) 's Twitter Profile Photo

Plenty of research around motivating and threatening LLMs to improve output adherence. Still, I inevitably chuckle when I come across such prompts :) Jokes aside, cool work on STar-GATE Chinmaya Andukuri and team!

Plenty of research around motivating and threatening LLMs to improve output adherence. Still, I inevitably chuckle when I come across such prompts :)

Jokes aside, cool work on STar-GATE <a href="/chinmaya_mohan/">Chinmaya Andukuri</a> and team!
Myles David (@tunedintotennis) 's Twitter Profile Photo

Aryna Sabalenka will receive $3,600,000 for winning the Women’s event in New York Jannick Sinner will receive $3,600,000 for winning the Men’s event in New York They played at the same venue for 2 weeks, they entertained the same crowds, and they WILL get paid the same #USOpen

Aryna Sabalenka will receive $3,600,000 for winning the Women’s event in New York

Jannick Sinner will receive $3,600,000 for winning the Men’s event in New York

They played at the same venue for 2 weeks, they entertained the same crowds, and they WILL get paid the same #USOpen
Yuchen Jin (@yuchenj_uw) 's Twitter Profile Photo

The highest-scored paper at ICLR 2025 with full scores, 10, 10, 10, 10! The first time in ICLR history? IC-Light is designed to control image lighting. They managed to collect >10 million images for training illumination editing models, with amazing results on SDXL and Flux

The highest-scored paper at ICLR 2025 with full scores, 10, 10, 10, 10! The first time in ICLR history?

IC-Light is designed to control image lighting. They managed to collect &gt;10 million images for training illumination editing models, with amazing results on SDXL and Flux
Dwarkesh Patel (@dwarkesh_sp) 's Twitter Profile Photo

We’re way more patient in training human employees than AI employees. We will spend weeks onboarding a human employee and giving slow detailed feedback. But we won’t spend just a couple of hours playing around with the prompt that might enable the LLM to do the exact same job,

Mike Knoop (@mikeknoop) 's Twitter Profile Photo

This is a more important product update than most realize. It's the first async background processing AI (read: unsupervised) from the major labs.

Britton Winterrose (@bwinterrose) 's Twitter Profile Photo

I left aws because my current life’s mission is to serve the world’s most formidable founders I met a ton of founders in YC W22 who needed access to Open AI, aws couldn’t provide, and in the words of Michael Jordan… “and I took that personally.” on Monday a founder you know

Kabir Walia (@kabir__walia) 's Twitter Profile Photo

Our paper, "Concept Distillation from Strong to Weak Models via Hypotheses-to-Theories Prompting," has been accepted at #NAACL2025! arxiv.org/pdf/2408.09365 We introduce a simple yet powerful prompt optimization method that can efficiently adapt prompts across models (essential