Future Trends in Artificial Intelligence
1 like1,503 views
The document discusses various future trends in artificial intelligence, including deep learning, big data analysis, emotional simulations in robotics, and advancements like the Blue Brain project aimed at replicating human brain functions. It also touches on concepts like mind uploading and self-improving AI systems that could lead to significant changes in technology and society. Potential applications of these advancements include more efficient traffic control systems and improved human-robot interactions.
Future Trends in Artificial Intelligence
- 2. Contents 1) DeepLearning 2) ChewingBigData 3) Simulating Feelings 4) BlueBrain 5) NeuromorphicApproach 6) Mind Uploading 7) AI Traffic Control Systems 8) Self-ImprovingAI 04/06/2013 P.S.Jagadeesh Kumar 2
- 3. (1) DeepLearning a) DeepLearningcan beviewedasamethodthat makes acomputerto solve problem independentlybygoingthroughmultiple level of learning. b) Insteadof writingacodethat instructshow to solveproblem,computerscan solvethe problemontheir ownbystudyingsimplebasicconcepts.It ismuchlike human understanding. c) Googleusesdeeplearningto optimizetheir searchresult, improve androidvoice recognitionandfor better adtargeting. d) Facebookusesdeeplearning(8 layer) to improve it’s newsfeed appandpredict the future behaviour of thepeople. e) Forexample; In first layer, changesin pixelsare noticed. Secondlayeridentifies the edgesandshapes. Thirdlayer getsinformation fromsecondlayer andsoon… 04/06/2013 Future Trendsin Artificial Intelligence 3
- 4. (2) ChewingBigData a) BigDatarefers to the collectionof ‘huge’ amountof data that cannotbeprocessedwith conventional database systems. b) Usuallyit’salot of unstructured data. It mightincludedata suchasimages,status updates,profiles, etc. that are spreadout in the socialnetworkingsitesandsuch data in largecompanies. c) Bigdata anddeeplearning are likepairs. d) Googleusedits deeplearningonits vastcollection of video clipsthat madeits computer identify catpicturesontheir own. e) Facebookisbeginningto build avirtual brain with its user’spersonaldata. Alsoit istrying to characterize the behaviour of peoplewiththeir ownpersonaldata. 04/06/2013 P.S.JagadeeshKumar 4
- 5. (3) SimulatingFeelings a) Computerdriven simulationscanshowemotionsin its digital face. b) Avirtual babycreatedbyDr.MarkSagar(Animation expertbehindAvatar andKingKong movies) reactsexactly like normalbabywoulddoinreal-time. c) It isaground-breakingadvancement in human-robotinteraction. d) Theneurochemicalreactionsare simulatedin the baby’ssimulatedbrainwhichin- turn drivesfacialexpressions. e) Oxytocin,CortisolandDopaminelevelsare simulatedandvary dependingonits social connection.(Thebabygetsupsetandcriesif we moveout of reachandgets happyif we arearound) f) Many suchexperimentscouldmakehumanto robot interactionlook more lively andreal. 04/06/2013 Future Trendsin Artificial Intelligence 5
- 6. (4) BlueBrain a) Bluebrain aimsin reverseengineeringabiologicalbrain andrecreatesit using computer simulation. Thesimulation will beaccurateat cellularlevel. b) IBM’sBlueGenecomputerisusedfor simulation purposeandhencethe name ‘Blue Brain’. c) Thescopeof the bluebrain project includesimprovement in fields of healthcare to advancedcomputingandbetter artificial intelligence.It isclaimedthat afull human brainsimulationwill bepossibleby2023. which isd) Blue brain studiesneuronsthroughthe biological data obtained, simulated in asupercomputerresultingin acomplete blueprint. e) Althoughthis method soundspromising,the power consumptionof blue brain is8 MW, anequivalent of powerconsumptionby1600 homes. 04/06/2013 P.S.Jagadeesh Kumar 6
- 7. (5) NeuromorphicApproach a) Neuromorphic engineeringaimsin replicating the ‘Brain’. b) Neuromorphiccomputers are sophisticatedcomputersystems similar in size, speedand energyconsumptionto that of ahumanbrain. c) Humanbrain consumesonly20 Watts of energywhereassupercomputersare consumingpowerin Megawatts. d) Neuromorphicengineeringaimsin creatingelectroniccircuitsthat fire neurons like humanbrain doesandthe dimensionsof these circuitsremain very small. e) Consumingjust5 Megawattsof power,theycan mimicI million neuronin a NeuroGrid. If this hardwareiscombined with the ability to deep learn,Terminator’s SkyNetwill becomeareal possibility. 04/06/2013 Future Trendsin Artificial Intelligence 7
- 8. (6) MindUploading a) Mind uploadinginvolvestransferring ourbrain to adigitalstoragemediumand moving consciousmemoryfrom oneto another. b) Althoughit soundstoo fictional, Googlefuturistclaimsthat humanswill beable to upload their entire brain to computersandbecomedigitallyimmortal. c) Mind uploadingandemulation couldmakeussuperintelligent , superfastand immortal in realworld. d) Thefuture will belike the ‘Matrix’ movie where everyonewill bepluggedinto computerswith BrainComputerInterface BCIs. e) Althoughsomeargueover the impracticality of this approach,minduploadingremainsthe ultimate embodimentof machineandmind. 04/06/2013 P.S.Jagadeesh Kumar 8
- 9. (7) AITrafficControlSystems a) Trafficiseverywhere,even onthe internet. b) Studiesreveal that atraffic signalsystemcontrolled byhumanbrainisvery efficient than the conventionalsystem. c) Sincethe humanbrain isgoodat traffic stuff, researchersare building ‘Machine Learning’basedAI traffic control systemthat couldcontrol the traffic efficiently. d) Theyalsolookforwardto the future wherevehicleswill beequippedwith GPSandWi-Fi to makethe traffic control a simple task. 04/06/2013 Future Trendsin Artificial Intelligence 9
- 10. (8) Self-ImprovingAI a) Sincethe basicnature of intelligence allowsfurther creation andinvention, a super intelligentsystemwouldfurther evolve throughrecursiveself improvement. b) Imaginehow wouldit beif acomputersoftwarecodesitself. Thisresultsin adaptation to thesituation. c) Theredesigningof cognitive algorithmsandupdationto abetter selfcouldturn it into somethingthat wasnot programmedinitially. It becomes‘SelfAware’. d) SelfImprovingAI couldstart intelligence explosion. 04/06/2013 P.S.Jagadeesh Kumar 10
- 11. ThankYou 04/06/2013 Future Trendsin Artificial Intelligence 11 Viewpublicationstats