[Colloquium] Talk by Manuel Blum on Tuesday, May 13, 2003

[Margery Ishmael]

-------------------------------------------------------------------------

DEPARTMENT OF COMPUTER SCIENCE
& TOYOTA TECHNOLOGICAL INSTITUTE

Tuesday, May 13, 2003 at 3:00 pm in Ryerson 251

--------------------------------------------------------------------------

Speaker: Manuel Blum

From: Carnegie Mellon University

Title: Can Humans Out-Think Computers? Can Computers Out-Compute Humans? 
Fundamentals of the CAPTCHA and HumanAUT projects.

Abstract:

1. Can a human out-think a computer? If so, it may be possible to construct 
a kind of sentry called a CAPTCHA that allows humans but not computers to 
enter the Garden of Eden.

*CAPTCHA = Completely Automatic Public Turing Test to tell Computers Humans 
Apart.

Applications: a CAPTCHA can help keep (ro)bots from sending spam mail, from 
getting computer accounts, from entering chat rooms, etc.

2. Can computers out-compute humans? Certainly! Computers can memorize huge 
data sets and do all sorts of arithmetic quickly and correctly. On these 
tasks, they leave humans in the dust.

Still, the question arises whether humans having no computational aids, no 
gadgets, no paper and pencil, can use their very meager computational 
abilities to authenticate themselves securely...in the face of (possibly 
computerized) eavesdroppers who observe and record every move they make. 
This is the goal of the HumanAUT project. **HumanAUT = Human AUThentication.

This talk will describe a working solution for CAPTCHA (see captcha.net) 
and the current thinking behind HumanAUT.

Speaker's Homepage: http://www-2.cs.cmu.edu/~mblum/

Hosts: Stephen Smale & Partha Niyogi

*The talk will be followed by refreshments in Ryerson 255*