Avinash @avi3presto

Great discussion on United States v. Google case by @patient_capital. Always surprised me that it's not "United States v. Apple": “the U.S. government appears to be targeting a product that people are generally satisfied with and attempting to drive them towards using the…

Being able to interpret an #ML model’s hidden representations is key to understanding its behavior. Today we introduce Patchscopes, an approach that trains #LLMs to provide natural language explanations of their own hidden representations. Learn more → goo.gle/4aS5epd

Graphs, structures that describe connections between objects, are everywhere — imagine the tools in a kitchen, parts of a bike, or a group of friends. Learn about our latest work that explores how to encode graphs in a format that an LLM can understand: → goo.gle/3TbowPr

one of the most important things I know about deep learning I learned from this paper: "Pretraining Without Attention" this what I found so surprising: these people developed an architecture very different from Transformers called BiGS, spent months and months optimizing it and…

unpopular opinion: first off this whole idea of using analog circuits to accelerate ai workloads aint exactly new. ibm intel and others have been experimenting with that shit for years tryna leverage the noise tolerance of neural nets to boost efficiency heres the thing analog…

keyboard_arrow_left Previous keyboard_arrow_right Next

Jason Carman @jasonjoyride

3 days ago

Exclusive first look inside @Extropic_AI, pioneers of thermodynamic AI compute, on episode 41 of S³

223 575 4K 2.3M 2K

Download Video

Good episode on how utilities can be a bottleneck for AI revolution. "I don’t think people quite realize the degree to which Microsoft is really just an energy company. We need power and we need silicon, we need chips...If we don't have one of those two things, we're in a lot of…

keyboard_arrow_left Previous keyboard_arrow_right Next

Diversification is a Negatively Priced Lunch with @markku_kurtti In this episode I chat with Markku Kurtti, author of the blog Outcast Beta. Markku is classically trained as an electrical engineer and works on receiver algorithms for mobile phones. A passion for investing,…

A Distributed Systems Reading List 👌 ferd.ca/a-distributed-…

New blog post, looking at the great new paper from the Amazon MemoryDB folks: brooker.co.za/blog/2024/04/2… Much of the discussion on distributed systems is about scalability, but this paper shows how availability, durability, cost, and performance are equally important.

The Semis layer makes ~90% of all Gen AI profits today! Where does future value accrue and how do we get there? I explore where the $$ in Gen AI is today (semis), where it's headed (apps), and how it compares to prior tech breakthroughs.

Just read Apple's "OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework". Similar to the OLMo, it's refreshing to see an LLM paper that shares details discussing the architecture, training methods, and training data. Let's start with the…

Mauboussin and Damodaran are truly exhausting. Reminds me everytime of Munger criticism “The craving for that physics-style precision". Dude, 20x PE is just the present value of earnings growing at 5% (GDP+) discounted at 10% (your return requirement). Or 3% and 8%. Whatever.

Some comments from the follow-up call: "we’re building this out in a way where we can repurpose the capacity for other things. So if it turns out that we do not need as much training capacity as we expect, we will have the flexibility to deploy it towards inference capacity,…

Mostly Borrowed Ideas @borrowed_ideas

5 days ago

Of all the post-earnings drop that I have experienced following $META since 2018, this one probably made me nervous the least. Here's my highlights from tonight's call mbi-deepdives.com/meta1q24/

10 17 190 78K 77

Of all the post-earnings drop that I have experienced following $META since 2018, this one probably made me nervous the least. Here's my highlights from tonight's call mbi-deepdives.com/meta1q24/

For Folks seeing higher debt in Hospitals, QSR Companies, Retailers like Titan etc. Its purely due to IND AS 116, where lease liabilities are considered a part of debt now. Leases were earlier part of Other expenses in Income statement. As a result👇👇

META put up a decent qtr (beating slightly) and guided more or less in-line but highlighted they will enter an AI investment cycle. Seems to me story is getting better (core engines chugging along w/ potential ban of TikTok reducing competition, AI under development) while…

In my humble opinion the recent Stream of Search paper (arxiv.org/abs/2404.03683) is truly outstanding. Everyone should give it a thorough read.

Feeling super generous so, somewhat reluctantly, here are 5 industry secret holy grail books that have immensely assisted me in my quantitative trading endeavours (as an ex manual HFT mm) that I highly recommend new and experienced traders pick up if they can. No specific order.

I've written up my study group lectures on implementing Transformers in PyTorch into a blog series: Creating Transformers from Scratch: - Part 1: The Attention Mechanism benjaminwarner.dev/2023/07/01/att… - Part 2: The Rest of the Transformer benjaminwarner.dev/2023/07/28/res…

I just read the 43-page International Monetary Fund paper on cross border flows in bitcoin. Thought I'd share three interesting takeaways to crystallize my thoughts. A quick thread. imf.org/en/Publication…