Invited Session (10:15-11:05)} - EEG-based Brain-Computer Interfaces Fabio Babiloni (Sapienza University of Rome, Italy) - Intelligent Autonomous Robots Luca Iocchi (Sapienza University of Rome, Italy) == Neural Network \& Artificial Intelligence == What is Artificial Intelligence? Which agent is more intelligent? Degree to how? == IBM's Deep Blue == Turk A legendary chess automaton in 18th century... == Napoleon vs. the Turk == It was human not the machine who played == The secret had been perfectly kept for more than 100 years == Turing's Chess Machine == How can machine intelligence be defined? == Legg & Hutter's definition Called "Universal Machine Intelligence" == Let's be specific not universal == Our mesure of intelligence - a: reward (degree of achivement) - b: how performance is improved than previous experiance - c: simplisity (entropy) - Occam's rasor - d: similarity (how different even in a same situation) == E.g. Stock market investment as a benchmark Input: Daily prices of all stocks of previous one year Run the algorithm we want to measure intelligence multiple time total reward (i.e., profit or loss) degree of improvement of result of $n$ runs complexity of series of actions in each run similarity of $n$ series of actions == Footboal bot == Simulation of brain with spiking neurons == Obama's proposal on brain science (2013) $3 billion investment to map 100 billion neurons in the human brain == Markram et al. in Lausanne Let's create a "virtual copy of the human brain == Neuroscience: rich in data & poor in theory Analyzing the brain data without a theory => A person who does not understand how a computer works got the wiring diagram of a computer and all snapshots of voltages at all the spots of the wires (high definition fMRI movie). Then he said "now we have data, and we need to analyze it! (J.~Weng) == Neural Dust A concept of human-machine interface by Seo et al. (U.C.Berkley). Thousands of tiny sensors, 100 micrometers, in a human brain $\Rightarrow$ dust. Each dust, powered by ultrasound transmissions, picks up electrical signals from nearby neurons They return data from the brain via ultrasound to a transceiver implanted under the skull == Neural Dust works like MRI and PET It would easily monitor neurological activity for research and medical monitoring. We could also see this technology as a brain-machine interfaces == Use your brain to control the world! A promise of the brain-machine interface The most direct way to record brain activity is to cut open the skull and place electrodes directly onto your grey matter, which presents other problems. So most of the research on brain machine interfaces in humans uses EEG At the same time, it enables cursor control, brain-to-brain communication etc. * Farewell to Smart-phone! (Bleecker & Nova at the South by Southwest Interactive conference === It might sound Science-Fictions.Now let's go back to a reality! EEG-based Brain-Computer Interfaces...Prof. Fabio Babiloni Prego Professore! \clearpage % ========================================= negative \vspace*{0.7in} \begin{center} \color{blue} {\bf Dzi\c{e}kuj\c{e}!} \end{center} \end{huge} \end{document}