Meta-problem of ML projects: very hard to know what's actually going to be hard. Progress (or lack thereof) on ML problems can be unpredictable, and human intuitions about "hardness" may be misleading.

Back to Lukas Biewald's garage robots: a known pitfall of vision ML in robotics where the "target" object is not perfectly centered in the camera field of view (whereas they usually are big training corpora), causing degraded performance.

"Training data becomes your biggest expense in many serious machine learning projects."

Thought-provoking blog post by Dr. Alex Wissner-Gross about the whether datasets or algorithms have a bigger impact on major AI/ML breakthroughs: "Datasets over Algorithms" spacemachine.net/views/2016/3/d…

Save 20% on upcoming O'Reilly conferences in San Jose with code "UGSFBAML": #OReillySACon oreil.ly/2ETt5HW #VelocityConf oreil.ly/2EVdEPt

Thanks everyone for coming out to the meetup earlier this week, and especially Amit Lakhani and Lukas Biewald for the great talks, and Nextdoor for hosting, food/drink, and video! Will post here and on the meetup page when videos are up meetup.com/SF-Bayarea-Mac…

We are happy to announce the keynote speakers for SBTB 2019: Helena Edelson Joe Beda Heather Miller The CFP opens tomorrow and runs until May 31. Submit your best talks early on — functional programming — service architectures — data pipelines (including for ML/AI) — &more!

Event of interest: O'Reilly #VelocityConf in San Jose (June 10-13) - save 20% w/ code "UGSFBAML", early price ends May 3 oreil.ly/2VpOcf3

Discount code "UGSFBAML" for 20% off #OReillyAI in San Jose Sept 9-12 (early price ends June 14): oreil.ly/2VXjRFO

Determinant Point Processes are a simple way to model probabilities on subsets (combinatorial objects), e.g playlists, shopping baskets, Ad banners, etc. Here are 2 preprints on the subject: (Non symmetric DPPs) arxiv.org/pdf/1905.12962… (Deep DPPs) arxiv.org/abs/1811.07245 Plse RT

We've revealed the speaker line-up! Join us at #ScaleByTheBay in San Francisco this November and hear from the top minds in functional programming, service architectures, data pipelines, and AI/ML at scale. scale.bythebay.io/speakers

Highly recommend the excellent “Projects To Know” newsletter from Sarah Catanzaro & Amplify Partners: short, well-curated digests of interesting recent ML research papers, code repos, & blogs mailchi.mp/amplifypartner…

Excited to moderate the panel on AI Products with Prof. Anima Anandkumar Peter Bailis Thanh Nguyen at @scalebythebay 2019 sched.co/RtTU #scaleSF

A couple of days left till Late Bird tickets go on sale. Do not miss out and book your ticket now. buff.ly/2NX42LG

Exciting summer series of online ML talks from Anyscale kicking off May 13th w/ Profs Michael Jordan & Ion Stoica, free registration here: eventbrite.com/e/the-future-o…

Sign up for #RaySummit—A FREE virtual event about Ray, the open-source Python framework for building distributed applications that run at any scale. You'll hear from experts such as Wes McKinney, founder of Ursa Labs and the creator of the pandas project. buff.ly/31ipI9J

And we are live! Scale By the Bay 2020 program is live and registration is open! Early bird through September 30. Become a Patron By the Bay so we run the best online event this year and return by the bay next year. Instantly sponsor with your logo. scale.bythebay.io

Early Bird is almost gone -- hurry up to get yours soon! scale.bythebay.io/register

Tons of great ML / pipelines / systems talks at this year's (online) AI by the Bay conference - 20% off registration with code "SFBAML20"