(public).sospep.events ~ view task ~ CREATE-article

<h1 style="text-align: center;">CREATE-article# #</h1>
<h2>[previously]</h2>
<ol>
<li><strong>[2017-01-15][00:00] NEW task IN ?</strong></li>
<ol>
<li>] # 6798 - CREATE-article# # (this)&nbsp;</li>
</ol>
<li><strong>[00:00] NEW article IN week-one-agenda</strong></li>
<ol>
<li>&nbsp;# # - [08:00] VIDEO SUMMARY-mooc-ai-video-001 - overview of ai</li>
<li>&nbsp;# # - [13:00]&nbsp;VIDEO SUMMARY-mooc-ai-video-002 - applications of ai</li>
<li>&nbsp;# # - [07:00] VIDEO SUMMARY-mooc-ai-video-003 - history of ai</li>
<li>&nbsp;# # - [18:00] VIDEO SUMMARY-mooc-ai-video-004 - course logistics&nbsp;</li>
</ol></ol>
<h2>[currently]</h2>
<ol>
<li>] &nbsp;text book - Artificial Intelligence: A Modern Approach - http://aima.cs.berkeley.edu/&nbsp;</li>
<li>] PR</li>
</ol>
<h2>[next]</h2>
<ol>
<li>]&nbsp;</li>
</ol>
<p>&nbsp;</p>
<hr />
<h1 style="text-align: center;">video 1</h1>
<p>&nbsp;</p>
<ol>
<li><strong>[00:00] NEW task IN&nbsp;</strong></li>
<ol>
<li>x] REVIEW VIDEO[08:45]&nbsp;</li>
</ol>
<li><strong>[00:00] welcome&nbsp;</strong></li>
<ol>
<li>] conversing with a computer using natural language has always been a dream&nbsp;</li>
<li>] attractive field for movie industry - her, terminator, ai, space odyssey, HAL, ex-machina</li>
</ol>
<li><strong>[00:00][WHAT] defining AI</strong></li>
<ol>
<li>] definition: intelligence @webster : &nbsp;ability to learn and solve problems</li>
<li>] that definition is independant of human or machine</li>
<li>] ai @wikipedia - is the intelligence exhibited by machines/software,&nbsp;</li>
<li>] ai @mcCarthy - science and engineering of making intelligent machines, McCarthy is pioneer in field, coined the term ai</li>
<li>&gt;] ai @russel&amp;norvig, authors of book(), - study and design of intelligence agents, where an intelligence agent is a system that percieves it environment and takes actions that maximizes its chances of success ( definition adopted for course )</li>
</ol>
<li><strong>[00:00][WHY by&nbsp;] andrew ng (researcher) -&nbsp;</strong></li>
<ol>
<li>] just as industrial revolution freed us up from physical drudgery, ai has the potential to free humanity from mental drudgery</li>
</ol>
<li><strong>[00:00][WHAT - 4 schools of thought from Russell</strong></li>
<ol>
<li>] thinking humanly - machines with minds -&nbsp;</li>
<li>] thinking rationally - using math and logic to do ai - PRB not all knowledge can be expressed with logical notations</li>
<li>] acting humanly - machines do things like humans, - ex turing test - ai can fool human tester, ex airplanes, fl</li>
<li>] acting rationally - &nbsp;doing the right thing, acheive the best expected outcome, - computation intelligence is the study of the design of intelligent agents.</li>
</ol></ol>
<div>&nbsp;</div>
<div>
<div><hr /></div>
<div>
<h1 style="text-align: center;">video 2</h1>
</div>
<ol>
<li><strong>[00:00] NEW task IN&nbsp;</strong></li>
<ol>
<li>x] REVIEW VIDEO[13:05] applications of ai</li>
</ol>
<li><strong>[00:30] first thoughts - virtual assistants</strong></li>
<ol>
<li>] AAPL siri, AMZN echo, GOOG now, MSFT cortana</li>
<li>] leverage deep neural networks to handle speech recognition and natural language understanding</li>
<li>] status making good progress, can use them for simple queries</li>
</ol>
<li><strong>[01:00] hand writing recognitions</strong></li>
<ol>
<li>] US postal service interested to automate mail processing in mid 80's</li>
<li>] yann lecun published solutions at the end of 80's, used neural networks to recognize hand written digits</li>
</ol>
<li><strong>[01:30] cheque processing</strong></li>
<ol>
<li>] same approach as for mail processing</li>
<li>] saved hundreds of millions for banks</li>
</ol>
<li><strong>[02:00] machine translation</strong></li>
<ol>
<li>] started with us gov trying to translate russian text to english</li>
<li>] first systems completely failed</li>
<li>] many funding cuts after initial failures impeded further progress</li>
<li>] today we have statistical machine transalate, leverages the vast amount of translated machine corpuses</li>
<li>] huge progress has been made, google translate&nbsp;</li>
</ol>
<li><strong>[00:00] robotics</strong></li>
<ol>
<li>] a lot of progress in this feild,&nbsp;</li>
<li>] NAO, ASIMO,&nbsp;</li>
<li>] surgery, housekeeping,&nbsp;</li>
</ol>
<li><strong>[04:00] reccomendation systems</strong></li>
<ol>
<li>] example amazon, netflix</li>
<li>] review yours/like-you {purchases, viewing, readings, ...} and refer you to others similar/related</li>
</ol>
<li><strong>[05:00] email enhancements</strong></li>
<ol>
<li>] &nbsp;spam filter,&nbsp;</li>
<li>] learns from your interactions, what you do,&nbsp;</li>
<li>] uses technigues like naive bayes classifiers</li>
<li>] makes ad recommendations</li>
</ol>
<li><strong>[05:30] facial detection</strong></li>
<ol>
<li>] uses viola jones method developed in 2001,one of the most successfull applications of ai</li>
<li>] using sliding window of 24x24 pixel window over image, to recognize different rectange features in them</li>
<li>] uses machine learning algorithm adaboost,&nbsp;</li>
<li>] very successfull, fails sometimes when problems with luminosity or orientation&nbsp;</li>
</ol>
<li><strong>[08:20] facial recognition</strong></li>
<ol>
<li>] detection identifies face, recognition tries to determine who it is</li>
<li>] very difficult because glasses/no glasses, smiling/not, looking left/looking right, ...</li>
</ol>
<li><strong>[08:50] medical imaging detection</strong></li>
<ol>
<li>] detection of breast cancer in medical images</li>
</ol>
<li><strong>[09:00] adversial search (games )</strong></li>
<ol>
<li>] 1997 world chess champion defeated by Deep Blue, IBM deep learning computer</li>
<li>] 2011 IBM watson defeats top jeopardy players&nbsp;</li>
</ol>
<li><strong>[09:50] natural language understanding and extraction</strong></li>
<ol>
<li>] utilized by watson&nbsp;</li>
</ol>
<li><strong>[10:00]&nbsp;</strong></li>
<ol>
<li>] lee sedol world champion in Go, defeated by GOOG alphago</li>
<li>] alphaGo utilizes deep learning, reinforcement learning and search algorithms, to complete the best moves</li>
<li>] a complicated search problem, because of branching factors, number of moves</li>
</ol>
<li><strong>[10:50] autonomous driving</strong></li>
<ol>
<li>] flourishing area, been a dream for a long time</li>
<li>] DARPA grand challenge, started in 2005, had to drive 132 miles</li>
<li>] lot of logistics and policy challenges</li>
</ol>
<li><strong>[12:20] more</strong></li>
<ol>
<li>] social network analysis, route finding, protein design, document summarization, transportation scheduling, information extraction, ....</li>
<li>] And there are many, many other applications that could actually leverage AI capabilities to make our life better.</li>
</ol></ol></div>
<div><hr />
<div>
<h1 style="text-align: center;">video 3</h1>
</div>
<ol>
<li><strong>[00:00] NEW task IN&nbsp;</strong></li>
<ol>
<li>x] REVIEW VIDEO[07:45]&nbsp;</li>
</ol>
<li><strong>[00:00] foundation</strong></li>
<ol>
<li>] contributions from many different fields including philosophy, mathematics, economics, logistics,&nbsp;</li>
</ol>
<li><strong>[00:00] philosophy</strong></li>
<ol>
<li>] philosopher were first contributors going back to Aristotle in 4000BC&nbsp;</li>
</ol>
<li><strong>[00:00] mathematics</strong></li>
<ol>
<li>] logic, formal representation and proof,&nbsp;</li>
<li>] computation,&nbsp;</li>
<li>] algorithms, first was euclid, GCD greatest common divisor,&nbsp;</li>
<li>] probability, theory of -&nbsp;</li>
</ol>
<li><strong>[01:45] economics</strong></li>
<ol>
<li>] formal theory of rational decisions, to maximize payoff or utility,&nbsp;</li>
<li>] combine descison theory and probability theory for decision making under uncertainty</li>
<li>] addressed game theory, which is an agent planning to maximize it utility in the presence of an opponent who is planning against him</li>
<li>] fomalized markov decision processes as a class of sequential decision problems with the markov property</li>
</ol>
<li><strong>[02:00] neuroscience</strong></li>
<ol>
<li>] how the brain functions</li>
<li>] compares simliarity to how computers function</li>
</ol>
<li><strong>[00:00] pschyology</strong></li>
<ol>
<li>]&nbsp;</li>
</ol>
<li><strong>[00:00] computer engineering</strong></li>
<ol>
<li>] feedback from the environment</li>
</ol>
<li><strong>[00:00] control theory and cybernetics</strong></li>
<ol>
<li>] feedback from the environment</li>
</ol>
<li><strong>[00:00] linguistics</strong></li>
<ol>
<li>] how language and thinking are related</li>
</ol>
<li><strong>[03:30] founders</strong></li>
<ol>
<li>] list&nbsp;</li>
</ol>
<li><strong>[] resources</strong></li>
<ol>
<li>] list&nbsp;</li>
</ol>
<li><strong>[04:20] history of AI</strong></li>
<ol>
<li>] broad field with a long history, now maturing,&nbsp;</li>
<li>] turing</li>
<li>] dartmouth meeting</li>
<li>] MIT video thinking machine - enthusiasm</li>
<li>] chess and checkers program</li>
<li>]&nbsp;</li>
</ol></ol>
<div><hr /></div>
<div>
<h1 style="text-align: center;">video 4</h1>
<ol>
<li><strong>[00:00] NEW task IN&nbsp;</strong></li>
<ol>
<li>x] REVIEW VIDEO[18:45]&nbsp;</li>
</ol>
<li><strong>[00:00] prerequisities</strong></li>
<ol>
<li>] programming</li>
<li>] probability, understanding of -&nbsp;</li>
</ol>
<li><strong>[00:00] level</strong></li>
<ol>
<li>] masters level course, expect some difficulty</li>
<li>] language = python, v2 or v3</li>
</ol>
<li><strong>[00:00] readings&nbsp;</strong></li>
<ol>
<li>] text book</li>
</ol>
<li><strong>[00:00] what you will learn&nbsp;</strong></li>
<ol>
<li>] &nbsp;</li>
</ol>
<li><strong>[02:30] roadmap&nbsp;</strong></li>
<ol>
<li>] &nbsp;</li>
</ol>
<li><strong>[00:00]&nbsp;</strong></li>
<ol>
<li>] function &lt;-takes precepts, -&gt; goes actions&nbsp;</li>
</ol>
<li><strong>[04:15] search agents&nbsp;</strong></li>
<ol>
<li>] &nbsp;agents that work towards a goal,&nbsp;</li>
<li>] consider the impact of <strong>actions</strong> on future <strong>states</strong></li>
<li>] <strong>job</strong> is to <strong>identify the actions</strong> that lead to the <strong>goal</strong></li>
<li>] paths come with different costs and depths</li>
<li>] 2 kinds - uninformed(no domain knowledge, bfs,dfs,ucs) &amp; informed(adds heuristics to reach goal faster, ) greedy search, A* etc</li>
<li>] example = 8 queen problem, route finding,&nbsp;</li>
<li>] concept methods =&nbsp;</li>
</ol>
<li><strong>[06:30] adversarial search ( games )</strong></li>
<ol>
<li>] checkers solved problem, by chinook,&nbsp;</li>
<li>] adversial&nbsp;</li>
<li>]&nbsp;</li>
<li>] concept methods = minimax algorithm</li>
<li>] example =&nbsp;</li>
</ol>
<li><strong>[07:45] machine learning&nbsp;</strong></li>
<ol>
<li>] how do we create programs that improve with experience</li>
</ol>
<li><strong>[00:00] maching learning - supervised learning&nbsp;</strong></li>
<ol>
<li>] binary classification, labels</li>
<li>] concepts methods = supervised learning, classification, k nearest neighbors, perceptrons, neural networks, linear regression, etc</li>
<li>] example =&nbsp;</li>
</ol>
<li><strong>[10:30] machine learning - unsupervised learning</strong></li>
<ol>
<li>] date segmentation or clustering, no labels</li>
<li>] concepts methods = unsupervised learning, clustering, k-means, associations rules</li>
<li>] example =&nbsp;</li>
</ol>
<li><strong>[11:15] constraint satisfaction problems</strong></li>
<ol>
<li>] search problem too, we dont care about the path, just the goal</li>
<li>] all paths are of the same depth, problem is formulated using variables, domains and constraints</li>
<li>] concepts methods = problem formalization, backtracking search, arc consistency,&nbsp;</li>
<li>] example = soduku, grid of 9x9 cells,&nbsp;</li>
</ol>
<li><strong>[13:00] logical agents</strong></li>
<ol>
<li>] logic can be used by an agent to model the world</li>
<li>] concepts methods = modus ponens, sound and complete inference, horn clauses,&nbsp;</li>
<li>] example =&nbsp;</li>
</ol>
<li><strong>[14:30] reinforcement learning</strong></li>
<ol>
<li>] agent evolves in stochastic and uncertain environment, agent learns from the reinforcement or reward, learning approaches for decision making where outcomes are stochastic, agent continues to plan and to learn to affect its environment, driven by maximizing rewards&nbsp;</li>
<li>] concepts methods =&nbsp;</li>
<li>] example =&nbsp;</li>
</ol>
<li><strong>[15:00] applications - natural language processing</strong></li>
<ol>
<li>]&nbsp;</li>
</ol>
<li><strong>[15:00] applications - vision perceptions</strong></li>
<ol>
<li>]&nbsp;</li>
</ol>
<li><strong>[15:00] applications - robotics</strong></li>
<ol>
<li>]&nbsp;</li>
</ol>
<li><strong>[] historicial moment</strong></li>
<ol>
<li>] Alan Turing</li>
</ol>
<li><strong>[] summary / conclusions</strong></li>
<ol>
<li>]&nbsp;</li>
</ol></ol></div>
</div>
<h2>&nbsp;</h2>
<h2>[reference]</h2>
<ol>
<li>] course-video-slides.pdf (local) - notes on the first four video lectures ] what is ai, ] applications of ai, ] foundation and history, ] course overview, ] wk1 quiz, due 01-25 ] wk1 discussion&nbsp;</li>
<li>]&nbsp;<a href="/view/article?id=5519" target="_blank">LIST-artificial-intelligence-usage-applications</a>&nbsp; - IN tech-dev-sw/research/index</li>
<li>] preparing-for-the-future-of-ai.pdf whitehouse (local) ] 58 pages -&nbsp;</li>
<li>] turing-paper (original) - http://www.turingarchive.org/viewer/?id=463&amp;title=3</li>
<li>+] What A.I. Can and Can't Do Right Now. &nbsp;<a href="https://hbr.org/2016/11/what-artificial-intelligence-can-and-cant-do-right-now" target="_blank">Harvard Business Review</a>. (my SRC = )</li>
</ol>

~ view task ~ TaskGroup] week-one FROM [ Plan] documenting-project IN [ calendars] prj-042-artificial-intelligence