AI Agents to Automate Scientific Discoveries [11/3]

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Email *

In the Virtual Lab's COVID-19 case study, why did the AI agents specifically decide to design nanobodies as binders for the new variants?

1 point

(A) Because nanobodies are the only binders that AlphaFold is able to model

(B) Because agents reasoned that nanobodies are smaller and computationally easier to model, which played to their strengths.

(D) Because the human scientist explicitly instructed the agents to use nanobodies.

According to the lecture, what is the key technical mechanism that allows agents in the "Agent School" to acquire and integrate new, specialized knowledge? *

1 point

(A) After generating a curriculum and reading papers, they perform supervised fine-tuning on their own model parameters.

(B) They add all the papers they read to a large vector database, which they query during inference.

(D) They re-run the entire pre-training process from scratch using the new scientific papers as a data source.

What is the specific technical framework that the "Paper2Agent" project uses to convert a static paper and its code into an interactive agent? *

1 point

(A) It creates a complex, multi-step prompt that instructs a base model on how to run the paper's code.

(B) It converts the paper's methods and codebase into a "Model Context Protocol" (MCP) that the agent can interact with.

(D) It relies on a human expert to manually write a new API for every paper that is added to the system.

In the example of autonomous agent-to-agent collaboration, what novel scientific insight did the "AlphaGenome" agent and the "ADHD study" agent discover together? *

1 point

(A) They identified several hallucinated references in each other's papers, improving the reliability of both.

(B) They designed a new nanobody that could bind to both COVID and the biomarkers for ADHD.

(D) They found that Claude Sonnet 4 was the most balanced reviewer for their respective domains.

How did Professor Zou characterize the primary cognitive limitation of current AI scientist agents, based on his group's work? *

1 point

(A) They are unable to perform interdisciplinary research and can only focus on a single, narrow domain.

(B) They are better at creating new computational tools than they are at using existing tools like AlphaFold.

(D) They excel at "breadth-first" (interdisciplinary) search but lack the "depth-first" expertise of human researchers.

A copy of your responses will be emailed to the address you provided.

Submit

Clear form

reCAPTCHA

Privacy Terms

This form was created inside of UC Berkeley.

Does this form look suspicious? Report

Forms