Suggestions
Matthew Gould
I help people own their digital property with NFT domains.
Matthew Gould is the Founder and CEO of Unstoppable Domains, a company that specializes in providing Web3 domain names. He is based in Reno, Nevada, although his company operates with a presence in San Francisco, California.14 Here are some key points about Matthew Gould:
Background
- Education: Matthew holds a Bachelor of Science in Finance from the Georgia Institute of Technology and pursued graduate studies in Economics at the same institution.14
- Early Career: Before founding Unstoppable Domains, he founded multiple software companies across various industries. He also held roles at Talkable as Director of Operations and Director of Analytics.15
Professional Achievements
- Unstoppable Domains: Founded in 2018, this company aims to onboard users onto the decentralized web by offering blockchain-based domain names. These domains allow users to replace cryptocurrency addresses with human-readable names and host decentralized websites.23
- Web3 Domain Alliance: Unstoppable Domains is a founding member of this alliance, which seeks to advance web3 domain registries and prevent malicious activities like phishing attacks.3
Personal Journey
- In his early twenties, Matthew faced legal issues related to selling marijuana during his graduate studies. He served time but has since built a successful career as an entrepreneur.5
Matthew Gould's LinkedIn profile can be found under matthew-gould-7877361.4
Highlights
Many English verbs can be passivized, but not all of them; we can't say "seven dollars were cost by the book" to mean the book cost seven dollars. In a new paper in Journal of Memory and Language, @craaaa shows that language models are sensitive to these exceptions, and tests two hypotheses from linguistics about the sources of information that can be used to learn them - frequency and affectedness.
We do this by manipulating the training corpus - either removing or adding evidence for the passivizability of a verb, or adding a novel verb to the corpus - and retraining the models. We find independent support for both hypotheses.
This is a new and significantly expanded version of an old-ish preprint, and a cool proof-of-concept of the LLM corpus manipulation methodology for linguistics, if I should say so myself. So please take a look even if you've read the previous version!
I'll be giving a keynote talk at Architectures and Mechanisms for Language Processing (AMLaP) this year, please submit an abstract by May 1 if you would like to have a beer with me in beautiful Saarbrücken https://t.co/7C2mCmQjMp


