Architecture
Organizations
Training
Datasets
People
100

An animal famous for its proximity to the prevailing architecture of today’s NLP research.

What is Llama?

100

Pan-Continental grassroots organization representing NLP research communities from large landmass whose population is separated by a large desert.

What is Masakhane?

100

Teaching a model that one option is better than the other

What is Direct Preference Optimization/DPO?

100

To test how well a model can understand and provide a solution to programming issues specific to an online repository.

What is SWEBench?

100

Keeps saying AI has hit a wall

Who is Gary Marcus?

200

Possibly a woman’s name with a single lowercase letter. For the GPU-Poor.

What is LoRA?

200

The organization attributed to the creation of a dataset named after what you get after stacking many things of similar but varying properties.

What is EleutherAI?

200

We (LTI community) can’t live without it but it keeps going down every now and then.

What is Babel?

200

A staple to pretraining that’s really just the open internet

What is Common Crawl?

200

Claimed 9 years ago that a medical profession would be gone 4 years ago because of deep learning.

Who is Geoff Hinton?

300

Model component often mistaken as N-models in a trenchcoat

What are Mixture-of-Experts?

300

Site named after a situation that is better than being wrong but not quite right

What is Less Wrong?

300

Not in the beginning but also not the end of training

What is Mid Training?

300

Data derived from the very thing we sought to train

What is Synthetic Data?

300

For the concept of training a model by telling it "Bad" vs "Good", Winner of the award named after the person who argued intelligence could be tested by whether a human subject could not accurately distinguish between computer or human interlocutors.

Who are Andrew Barto and Richard Sutton?

400

Component invented by a researcher whose last name can also be read as an English Pronoun

What are Residual Connections?

400

An organization that, about 9 years ago, launched a “Conversational Understanding” bot online that quickly learned to hurl slurs and insults.

What is Microsoft?

400

Imagines as the top of a black forest cake which a certain NYU professor noted as part of a model’s training

What is Reinforcement Learning/RL?

400

Multilingual, Portuguese Named, also name of an Island in Indonesia

What is FLORES?

400

Athlete, winning 1 of 5 matches in a historic game of man vs machine that would convince a company of buying another company.

Who is Lee Sedol?

500

The first word of a foundational paper also noted for having authors whom are no longer affiliated with the organization that published it.

What is Attention?

500

Non-Profit, An Institution Associated with Intelligence

What is AI2?

500

Reward Increment = Nonnegative Factor x Offset Reinforcement x Characteristic Eligibility

What is REINFORCE?

500

An open dataset from a diverse mix of web content, academic publications, code, books, and encyclopedic materials. Made by folks on the west coast.

What is Dolma?

500

"I tend to think that most fears about A.I. are best understood as fears about capitalism"

Who is Ted Chiang?