Hammond Pearce

Research Assistant Professor @ NYU Center for Cybersecurity

[Hammond Pearce @kiwihammond]

Absolutely stoked to report that I won the @efabless inaugural AI Generated Silicon Design Challenge: https://t.co/TIkZxvv9gL

[NYU Tandon @nyutandon]

#NYU #Tandon in the #News:

#Tandon researchers @kiwihammond Ramesh Karri, @sg1753 & PhD student Jason Blocklove have fabricated a #microprocessingchip using plain English "conversations" with an #AImodel & more
#NYUTandonMade
https://t.co/IcaXUgy3KK

[NYU Tandon @nyutandon]

#Tandon researchers @kiwihammond Ramesh Karri, @sg1753 & PhD student Jason Blocklove have fabricated a #microprocessingchip using plain English "conversations" with an #AImodel.
#NYUTandonMade
https://t.co/EnQHe3eWYf

[Hammond Pearce @kiwihammond]

@sg1753 @psduggirala Exactly right, the SOTA performance is limited, esp. w.r.t. the size of the design you are working on (context window), still, an advantage of RISC-V is that the LLMs will have seen plenty of relevant training data!

[efabless.com @efabless]

Automating the translation of natural language specifications to hardware description languages could reduce sources of human error in the engineering process - https://t.co/YsfXcidIjk

[Siddharth Garg 🌈 @sg1753]

Can developers design chips using plain English and no Verilog background? We tried two fully conversational approaches to chip design: "Scripted Chat" (constrained to follow rules) and "Free Chat." (w/ Jason Blocklove, Ramesh Karri and Hammond Pearce) https://t.co/OBWDwvls5l

[Hammond Pearce @kiwihammond]

Excited to be sharing our latest #LLM work at @IEEESSP next week, where we looked at applications for automated security bug repair! Thanks to my co-authors @ichthys101 Baleegh Ahmad @moyix & Ramesh Karri! Teaser: https://t.co/S98r3ltZ55

[NYU Tandon @nyutandon]

#Tandon #cybersecurity professor @kiwihammond interviewed for @ITINewsBreaks about using #ai for content creation. #NYUTandonMade
https://t.co/tcGvgMtiAK

[NYU Tandon @nyutandon]

#Tandon professor @moyix discusses his paper on recounting how he and other @nyuniversity researchers @kiwihammond @sg1753 put the security of source code created with the help of large language models to the test with @TheRegister #NYUTandonMade
https://t.co/28T14H25oD

[Hammond Pearce @kiwihammond]

@MoodyHikmet At my last apartment rental in NZ, we had a broken window which let in cold air. The landlord didn't have it fixed for 4 months.
I hate renting.

[Hammond Pearce @kiwihammond]

@MoodyHikmet @AucklandUni I mean, @AucklandUni makes us do cybersecurity awareness training to identify suspicious emails...

[Hammond Pearce @kiwihammond]

Just received this email. Surely a scam of some kind? Asked @AucklandUni security team, nope, it's actually just a minimum-effort real email notification (confirming I'm no longer a student! Over 2 years late...) from the health team :) Am I being too paranoid?

[Hammond Pearce @kiwihammond]

Very pleased to share that our paper 'Examining Zero-Shot Vulnerability Repair with Large Language Models' has been accepted by IEEE S&P 2023 and is now up on arXiv: https://t.co/RTO0nTUSQQ - Special thanks to the co-authors @ichthys101 @moyix and Baleegh and Ramesh!

[Pwnie Awards @PwnieAwards]

Huge huge shoutouts to the speakers who volunteered their time to help present the awards this year! 🎤

Thank you: @moyix @kiwihammond @juanandres_gs @SupriyaMaz and @steventseeley !!!

[Hammond Pearce @kiwihammond]

Stoked today to have presented our research into #github #copilot with @ichthys101 at #BlackHat today! Special thanks also to our co-authors on the work @moyix and Baleegh and Ramesh!

[Nucleus Security @nucleussec]

One of our favorite #BlackHat sessions of the afternoon was “In Need of ‘Pair’ Review: #Vulnerable Code Contributions by #GitHubCopilot” — and not just because of their top class memes 👌🏼 @kiwihammond @ichthys101 @moyix #blackhat2022 #livefromblackhat

[Hammond Pearce @kiwihammond]

@moyix The worst part was getting my bag over the fence first

[Brendan Dolan-Gavitt @moyix]

Woo! Our paper "Examining Zero-Shot Vulnerability Repair with Large Language Models" has been accepted at @IEEESSP 2023! This paper evaluates using off-the-shelf code models like Codex, AI21's Jurassic J1, and our own GPT-CSRC for fixing vulnerabilities. Updated preprint soon!

[NYU Tandon @nyutandon]

#AI -generated content is the holy grail of efficiency...or is it? Tandon #cybersecurity researcher @kiwihammond shares thoughts on the pros & pitfalls of AI-generated content engines like GPT-3, CoPilot, DALL-E, Codex, Jarvis, & more...https://t.co/nOjDyL6TTu