intelligence training, robert daniel
DESCRIPTION
Robert Daniel will present the first place winner of the AI Concept Exploration category of the Federal Virtual Worlds Challenge. Intelligence training concepts incorporate AI solutions into gathering HUMINT, GEOINT, SIGINT and OSINT. HUMINT uses text to speech / speech to text and Google translate to simulate an intelligence agent looking for information on a topic by interacting with several persons of interest and questioning multilingual intelligent avatar bots. GEOINT uses swarm concepts to control the in-world environment, such as wind, rain and sand storms, including wildlife tracking and generation, such as flocks of birds, packs of dogs and schools of fish. Also includes swarms of UAVs and ground robots as geospatially aware entities. SIGINT includes voice analysis of in-world voice chat and conferences for fraud detection. It is a 3D visualization front end for AI-based plug-ins for time and frequency analyses of in-world voice chat. OSINT uses web bots to extract real world data for in-world user experience enhancement pulling from sources such as wikipedia, YouTube, Twitter and other social networking sites. Other cool functionality presented includes in-world/out-world calling using landlines and cell phones using; conference calling between worlds; and ground robot random walk to find and diffuse IDEs.TRANSCRIPT
Train for Success TalkIntelligence Training
Robert DanielSL: Quincey Dagger
May 26, 2011
Introduction
Adjunct Professor at GWU• 10 Teaching Telecommunications• SecondLife 4 years• OpenSim for the past 3 years• FreeSwitch for the past 2 year
Intelligence Training • Using AI for training• 1st winner at the FVWC 2011 for
AI Concepts
FVWC 2011
Theme: Artificial Intelligence • Must be an AI related entry• Entries must be in a virtual environment• Examples may include
– Adaptive learning systems– Intelligent conversational bots– Adaptive behavior
• Objects• processes
http://fvwc.army.mil/index.php
Intelligence Training
Our approach to the contest• Apply as many existing AI concepts to a
Virtual World– Neural Networks– Swarm Intelligence– Natural Language processing– Speech recognition– Weak AI– Strong AI– Expert systems – etc
Intelligence Training
Our approach to the contest• We picked 6 AI Concepts
– 1. Weak AI• AIML Artificial Intelligence Mark
Language– 2. Natural Language process
• Text To Speech– 3. Statistical Machine Translation
• Google Translate with German AIM– 4. Swarm Intelligence
• Biods flying birds– 5. Speech recognition
• c– 6. Speech Processing
• Signal process with MATLAB/Octave– Neural Networks– Fuzzy Logic
Intelligence Training
Theme: Intelligence Training• HUMINT: Human Intelligence Training• OSINT: Open Source Intelligence Training• GEOINT: Geospatial Intelligence Training • SIGINT: Signal Intelligence Training
Intelligence Training
Theme: Mapped INTs to AI• Station 1
– HUMINT: Intelligent Chat Bots• Weak AI with AIML
• Station 2 – HUMINT: Text To Speech Chat Bots
• TTS using Cepstrum Voices• Station 3
– OSINT: German Text To Speech Chat Bots• Google Translate
• Station 4– GEOINT: UAV Swarms
• Boids (artificial life program)• Station 5
– SIGINT: Speech Recognition• CMU Sphinx
• Station 6– SIGINT: Speech Processing
• Using Matlab and Octave
Station 1: HUMIT Avatar
AI: AI Markup Language• In-world Intelligent Chat Bots
– Using AIML– OpenMetaverse
Station 2: HUMIT Avatar
AI: Text To Speech• In-world Intelligent Chat Bots
– Using AIML– FreeSWITCH– Cepstral Voice
Station 3: OSINT Avatar
AI: Text To Speech• In-world Intelligent German Chat Bots
– Using German AIML– FreeSWITCH– Cepstral German Voice– Google Translate
Station 4: GEOSINT Objects
AI: Boids (artificial life Program)• In-world UAV Swarms
Station 5: SIGNT Avatar
AI: Speech Recognition• In-world order a pizza
– CMU Sphinux– FreeSWITCH
Station 6: SIGINT Workstation
AI: AI MATLAB Toolkit• In-world Speech Processing
– Using MATLAB– FreeSWITCH
What’s Next
External Interfaces• Xbox Kinect
What’s Next
External Interfaces• Emotiv