@lookyhere123 iPhone zoom doing some work. It’s about 4-5 miles away but I work in Santa Monica wand was just about two miles from it for most of the day.
English
Rex McArthur
872 posts
Rex McArthur
@drexmcarthur
AI research @youdotcom Technical Advisor at @us_abstract formerly led AI @heyjasperai @Apple Maps, Lyft.
Los Angeles, CA Katılım Kasım 2014
1.1K Takip Edilen326 Takipçiler
@drexmcarthur how far away is it from you? Have relatives in Manhattan
English
Taken from my roof of the palisades fire. We’re safe but it’s a scary night. If anyone in LA needs a place to stay, let me know. We have a guest room in Marina Del Rey.
English
Up and to the right. Got lots in store for 2025. You.com will gain some serious capabilities this year. 🫡
Richard Socher@RichardSocher
2024 was an incredible year for you .com. We grew revenue by 40x into the 8 digits. I want to thank our many large enterprise customers and pro subscribers. There's a lot of hype in AI but ultimately knowledge workers and companies need accurate answers and reliable workflow automation with agents. Several of our largest customers first had to fail with a Copilot or ChatGPT and then learned about our solution. They like our superior accuracy, not just of the answers but also the citations and search backend over massive, complex or fast-changing datasets. The speed of change has been incredible so having quick access to all the best LLMs, all while being connected to the most accurate web search backend and your own company-internal data is incredibly powerful for our enterprise customers. Our platform powers some of the largest companies in the world now. We provide over 20M answers on many days across our ydc product and APIs. It's exciting to see an extremely positive NDR in enterprise and our customers growing their usage. The future of knowledge work is already here but not equally distributed. If you want to skip the hype and POCs and become more productive, we're here to help. 2025 will be incredible!
English
@PhilippeLaban @BMarcusMcCann @PranavVenkit @jasonwu0731 Great paper! Thanks for sharing. Excited to take a look and would fun to get a small “leaderboard” setup for when we make large updates to citation logic.
English
How good is #SearchGPT? How does it compare to other answer engines like You.com, Perplexity, or Bing Chat?
The AnswerEngineEval benchmark we developed with @PranavVenkit helps us evaluate scientifically.
English
Sorry to interrupt your election doom scrolling but look who has the most accurate citations!
@youdotcom had made huge strides. Cool paper too.
arxiv.org/pdf/2410.22349
English
US immigration is deeply flawed. It took me years (and several visa permits) before I could stop being anxious about my residency status. Lots of people helped me in the process.
Kudos to @0xSigil for creating a network of people who went through the same struggle. Now it’s our turn to help you.
If you did notable work in AI, you can count me in for a reference letter!
Extraordinary@extraordinary
Michele Catasta (@pirroh), President of @Replit, ex AI @ GoogleX @Stanford
English
@jasonfried Insane ROI on what is effectively one fewer FTE.
English
So, we do these full-company meetups twice a year in different cities around the world.
We fly everyone in, we hang out for a week.
People are often curious how much these cost, what we budget, how different cities compare.
Here's a look at real numbers from the last 6 meetups across the US, Europe, and Canada. Company size varied a bit from Miami - Montreal, so it's not truly apples to apples overall, but we also break it down by person to equalize it.
Hope this is helpful.
English
@julianeagu @HamelHusain Metrics and measurement >> everything
English
Excellent guide from @HamelHusain👌
IMO the hardest challenge for evals is folks deciding on what their definition of good is. Unless that happens, you can’t standardize, bootstrap, and scale your evals.
Hamel Husain@HamelHusain
The most common mistakes I see teams make with LLM judges: • Too many metrics • Complex scoring systems • Ignoring domain experts • Unvalidated measurements That's why I wrote this guide, w/ detailed examples to help people avoid these issues (1/4) hamel.dev/blog/posts/llm…
English
@HamelHusain @julianeagu is great! I didn’t realize you knew each other. Her work on evals is fantastic.
English
You should meet Julia. She is a talented colleague I worked with at GitHub who has been focused on evals!
Julia Neagu@julianeagu
On my way to @githubuniverse — excited to catch up with everyone! hmu you’re in SF this week.
English
@andrew_n_carr Bar is low for v0. Use APIs. Save money improve quality second.
English
The irony is you don't need to build custom NLP classifiers for every task anymore, but there is still so much alpha in training a modern arch for those tasks
English
California usually has TONS of proposition ads, but I think I’ve missed all of them, but I’m very well informed on Nevada propositions.
English
YouTube TV for some reason thinks I live in Nevada. Having only lived in solid states (UT, CA) I am SHOCKED at the percentage ads that are political. I just got Trump, Harris, go vote, Trump, Harris.
English
us just won by importing 50% of workers from taiwan to run fabs.
manufactured in the us, by imported Taiwanese engineers.
Bojan Tunguz@tunguz
Wow. 🤯 People have no idea how big of a deal this is. The US just won. Everything.
English
@NathanpmYoung I do this with cursor and some good pre prompting and markdown. Greatest markdown editor I’ve ever used.
English
I want cursor but for writing. Give me a VScode instance with a nice white instead of black, LLMs tuned for writing text, not code, good speech to text and an LLM that sits over the top.
"Reformat the essay so that it has five distinct sections"
Anyone doing that?
English
@TheRideshareGuy I use and choose Waymo every single time. After tip it’s cheaper. I can plan ahead and wait an extra 10 mins, and in LA traffic, highways are rarely faster anyway. I mostly stick to the west side regardless. 100% Waymo user no matter the price
English
I’ve been tracking prices for Waymo vs Uber/Lyft but the big thing that stood out to me on this ride was Waymo’s in-car travel time of 29 minutes - so what’s going on?!!
🚗 Uber: $28, ETA 2 mins, in-car travel time 14 mins
🚙 Lyft: $31, ETA 4 mins, in-car travel time 22 mins.
🤖 Waymo: $28, ETA 17 mins, in-car travel time 29 mins.
Since Waymo doesn’t operate on freeways, any time the freeway is the fastest route, you lose a ton of time by opting for Waymo. No idea why Uber’s in-car estimate is so aggressive compared to Lyft but if we conservatively average the two, that is 18 minutes vs Waymo at 29 minutes.
Not to mention the absurd 17 min ETA wait time for Waymo. I’m not convinced that once the novelty of Waymo wears off that passengers will be willing to wait longer for a ride that also takes longer. What do you think?
This is just another reason why I think it makes sense for Waymo to partner with Uber. That way you can balance the ETA and domain of operation problems that Waymo is bound to face.
P.S. – I’ll dive deeper into this topic and more in this week’s ‘Driverless Digest’ newsletter - subscribe below!
English
Do you think @EduardHabsburg is ever annoyed they took this from him? Does he have to pay admission to see the ol’ family cabin?
English
@burkov This is a pedantic gotcha question. Of course attention and measurement of output distributions is deterministic, but almost everyone uses them with some sort of stochastic sampler to look at sequential outputs.
English
