Computational Model for Acquisition of Knowledge about Possession

1
Computational Model for Acquisition of
Knowledge about Possession Gaylee Saliba
2
Main Aspiration of A.I. Lab

• Attain a thorough understanding of intelligence
• Design/ develop systems exhibiting such
  intelligence

3
Solution to Problem?

• Comprehend the inner workings of the human mind
• Robotics, vision, and language

4
Genesis Group

• Who/What?
• Active group of faculty members striving to
• understand intelligence led by Professor Winston
• Mantra?
• Its all about vision, language, and motor
  faculties

5
Vision

• If we are to successfully create intelligent
  machines, it is essential to develop a system
  that grounds abstract notions such as possession
  in the physical world

6
Prior Work

• Ray Jackendoff (Linguistics)
• Trajectories
• Gary Bordchart (A.I.)
• Transition space representations (TSR)
• Key focus on change of state not time

7
Solution
8
Possession Space

• 3 Dimensions
• - Physical State
• - Mental State
• - Desire

9
Key Examples

• A ? B A ? B
• D D D D
• Exchange/Sell/Buy
  
  Swindle
• A ? B A ? B
• D D ND D
• Give
  Steal

10
Templates
11
Prominent Physical Possession Transfer Verbs

• Exchange
• Sell/Buy (represented by 1 template)
• Swindle
• Donate
• Give
• Steal
• Share
• Lend
• Borrow
• All other verbs get classified into stacks!

12
Facilities and Equipment Requirements

• Scientific papers from various MIT facilities and
  CSAIL resources
• Computer

13
Progress and Schedule

• Academic advisor Prof. Winston
• Summer/ Fall 2003 UROP work
• IAP 2004 Model for physical possession
• Spring 2004 6.170
• Model for abstract possession
• Develop learning algorithm
• Year 2004/2005 More A.I. courses
• Implementation (in Java)

14
Risks

• FUNDING!!!

15
Bridge Vision

• If we are to fully understand intelligence, we
  need to provide a substrate for abstract thinking

16
Bridge Steps

• Select some abstract domains and determine the
  information needed by agents to work in such
  domains
• Determine a method to obtain this information
  using knowledge confined within the Bridge
  Projects world