AI is unlocking new mathematical, algorithmic, and scientific discoveries. Our team at @Oracle found a much simpler approach to LLM-driven discovery that sets a new state of the art—in under 5 hours on 8 H100s
Hill Sampling for Test-Time Scaling
https://t.co/E1XDRepBC9
Hill Sampling is remarkably simple:
• Condition the LLM on the single best verified solution found so far
• Generate and evaluate many edits in parallel (e.g., 512)
• If an edit is at least as good, make it the new context
• Repeat
That’s it.
Frustrated and venting. Charging port for my Pixel 2 was broken, so I called @Verizon. I was told I could get a free replacement as part of my plan. I sent in my old phone got a refurbished Pixel 2...and was charged $300 for damage to the phone I sent in...
Fine! I say, but am I still going to be charged the $300 for the phone I currently have? No, says the rep, it's been wiped from you account. OK, cool, thank you, I say...Then today I find out there is a charge for $428 on the account which includes the $300 plus shipping, etc..
Friend-to-friend encouragements to vote work
Here a texting campaign shows an increase of 1 to 15 percentage points
https://t.co/ZFJaP7pk0y
Data and analysis code: https://t.co/aJtTocjnPG
@AaronSchein@dhanya_sridhar@victorveitch@keyonv
At every turning point of this crisis, Trump has failed the test of leadership—but he hasn't failed alone. Republicans sat on their hands while COVID-19 spiraled out of control and took our economy along with it. This November, we need to hold them all accountable. #Debates2020
More than 200,000 Americans are dead.
That's not just a number. It's someone’s grandmother, someone’s father, someone’s best friend.
It's my brother, Don Reed.
Trump has spent this pandemic lying, pointing fingers, and playing politics—leaving a nation behind to grieve.
(Thread) How can I say no to this? My course will be made public! Stay tuned.
I was pulling your leg here. Here's my (hot? cold?) take: academics have a *moral obligation* to make as much of their material public as possible. Not just papers, but code, talks, and lectures. 1/n
Thread: Last week, a list of 100 important NLP papers (https://t.co/PUHTvKCuiI) went viral. The list is okay, but it has almost *no* papers with female first authors.
NLP is rich with amazing female researchers and mentors. Here is one paper I like for each area on the list:
Um ... Google plans to launch a new AI ethics service to offer others advice on tasks such as spotting racial bias in AI algorithms. My spidey sense is tingling, but haven't figured out why yet. After all, what could possibly go wrong? @tsimonite
https://t.co/WwiRQ0ovKb
As part of #ShutDownSTEM, I'm trying to educate myself better. In doing so, I've come across the Justice in America podcast (@Justice_Podcast), which is wonderful so far, and I highly recommend. Full of important information.
The program for FORC 2020 is now live! https://t.co/VCGFhHv8Iw Register for FREE to attend the online event on June 1+2. In addition to a terrific set of contributed talks, we have amazing keynote speakers: Adrian Weller, Rakesh Vohra, Patricia Williams, Jon Kleinberg. Retweet!