Summary of Loose Lips Sink Ships: Asking Questions in Battleship with Language-informed Program Sampling, by Gabriel Grand et al.
Loose LIPS Sink Ships: Asking Questions in Battleship with Language-Informed Program Sampling
by Gabriel Grand, Valerio Pepe, Jacob Andreas, Joshua B. Tenenbaum
First submitted to arxiv on: 29 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper investigates how people navigate large hypothesis spaces to ask informative questions given limited cognitive resources. A classic question-asking task based on the board game Battleship is used to study these tradeoffs. The language-informed program sampling (LIPS) model generates natural language questions, translates them into symbolic programs, and evaluates their expected information gain using large language models (LLMs). The results show that a simple Monte Carlo optimization strategy yields informative questions mirroring human performance across varied board scenarios with a modest resource budget. In contrast, LLM-only baselines struggle to ground questions in the board state, with GPT-4V providing no improvement over non-visual baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper explores how we think about uncertainty when asking questions. It looks at how people make decisions about what to ask given limited brain power. The study uses a popular board game called Battleship as a test case. A special computer program, LIPS, is developed to mimic human question-asking abilities. LIPS uses big language models to create questions and evaluate how good they are at finding answers. The results show that with just a little bit of effort, the program can come up with great questions that are similar to what humans would ask. In contrast, other programs based only on language models struggle to understand the game and make smart decisions. |
Keywords
» Artificial intelligence » Gpt » Optimization
