Inexpensive token generation and agentic workflows for large language models (LLMs) open up intriguing new possibilities for training LLMs on synthetic data. Pretraining an LLM on its own directly generated responses to prompts doesn't help. But if an agentic workflow implemented

My mind was blown when I heard this earlier today..

'Imagine we discovered a new continent with 100 BILLION people on it, and they're all willing to work for free!
That's what's about to happen with #AI , so you'd better factor that into your plans.'

Nat Friedman #Abundance360

Microsoft presents The Era of 1-bit LLMs

All Large Language Models are in 1.58 Bits

Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single

LLM API responses too verbose? Here's 11 tweaks that have worked for me in an AI for HR system.

1. Give a few-shot example as a (user, assistant) pair in the conversation. This works very well to guide how the assistant should respond.
2. Give an example in the system prompt.

Just had an idea! It's so neat! Gonna ask reality what its opinion is.

Its fine to be building working AI applications and prompts, but there is tons of value in building really ambitious ones that almost work.

Better brains (GPT-5, Llama 3, Gemini 2.0, etc) are coming. When they do, you can swap them into the almost applications for a fast start.

Yes! Exciting that it summarizes whatever needs to be summarized (summarize clusters of chunks -> summarize clusters of summaries).

Esp. that it then collapses the tree of chunks and summaries for mixed level retrieval. Nice simplicity.
arxiv.org/abs/2401.18059

We’re excited to announce our $20.5 million #SeriesB led by 01A with Index Ventures and Zoom Apps Fund!

At BrightHire, we’re transforming the quality of hiring—and we can't wait for what's to come!
#interviewintelligence #hrtech #techfunding

brighthire.ai/blog/brighthir…

Wow - our CEO at Periscope Data Harry Glaser is the #1 rated CEO in San Francisco! Thanks Comparably comparably.com/companies/top-… #companyculture #bekind

Book on implementing machine learning algorithms from scratch. Neural nets, regression, decision trees, ~100 LOC per. joelgrus.com/2015/04/26/dat…

Yann Lecun, Facebook // Artificial Intelligence // Data Driven #32 (Hosted by FirstMark Capital) youtu.be/AbjVdBKfkO0

Another post that started life as an interesting customer problem!

Using Self Joins To Calculate Retention Metrics: periscope.io/blog/use-self-…

I just posted: How to calculate confidence intervals in SQL periscope.io/blog/how-to-ca…

Just got home from the #html5gamejam with @mrotondo, and I'm really proud of our game, Ninja Leap: http://mikerotondo.com/ninjaleap

Airlines should have every boarding event be a contest. Fast enough, get inflight snacks.

Stuck at O'Hare. At least the airport wifi is faster than my home connection. Yay 2Mbps downlinks.

Finishing up my google application.