Industry sign banking colorado form computer
artificial intelligence is a term that has taken an increasingly bigger space in our collective conscience artificial intelligence is all around us everywhere visible and invisible artificial intelligence is making our lives better and more convenient it is a pillar of modern technology it is a fascinating field filled with endless possibilities for new avenues of development and it opens doors for many exciting breakthroughs in the future welcome to this video on ai where we will learn about the term ai its history its applications in the present and its potential in the future the term artificial intelligence was coined by john mccarthy in 1956 first let us know the definition of ai artificial intelligence is the intelligence exhibited by machines as opposed to the natural intelligence exhibited by human beings and animals poole mackworth and gerbil researchers at the university of british columbia write in their book computational intelligence a logical approach that artificial intelligence or a i is the established name for the field we have defined as computational intelligence or ci they also prefer to use the term synthetic intelligence using a pearl as an analogy a synthetic pearl may not be natural but still it is a pearl now we have seen the definition of ai we can move on to the next section the history of ai the history of ai is quite fascinating and can be chased back thousands of years though the history of modern ai had its beginning in the 20th century myths and tales of inorganic intelligent anthropoids can be traced back to 400 bc in greek and jewish mythology and folklore in the story of the argonauts written 2 400 years ago an automaton is mentioned that is a machine endowed with life called talos in jewish folklore that is a myth of an anthropoid made of clay known as a golem the concept of artificial intelligence is based on the idea that thinking can be mechanized or emulated by a machine leibniz thomas hobbes and rene descartes explored the simplification of thinking into a system leibniz inspired by 13th century mathematician and polymath ramon lal wanted to make a machine that could reason and think and give definite answers he argued that thoughts no matter how complex were made up of simple and complex fundamental concepts he imagined a machine called the great instrument of reason leibniz also imagined a time when to resolve disputes between thinkers they could simply say let's calculate and find a solution so ai by no means is a new concept but the birth of ai in its modern form starts with the dartmouth conference during a brainstorming session several researchers at dartmouth set out to make computers think the dartmouth conference is where ai was given its name and work on it began it's where our modern history of ai begins early ai researchers were very optimistic and dare i say unrealistically so the public was extremely surprised with anything that a computer did this led government organizations to pour huge sums of money into ai research in the hope of making a breakthrough early programs could solve mathematical equations which dumbfounded the public the public did not even think anything of this caliber possible early ai programs used a very basic algorithm they would try going forward and when hitting a wall or getting an unsatisfactory result they would go back and try to turn and try again and again until reaching the result this is a very straightforward and basic way to solve any problem but this was very inefficient the first ai winter saw a drastic decrease in interest and funding in ai it is sort of a nuclear winter but for ai it is quite a fitting metaphor nuclear winter is the hypothetical prolonged period of global cooling after a nuclear firestorm a similar thing happened during the first ai winter a.i burst onto the scene and its researchers were too optimistic the organizations that poured a lot of money into ai research were expecting more the great optimism in the early years of ai pushed the expectations to new heights the researchers believed that they were only a few decades away from creating fully intelligent computers which was a very unrealistic expectation was a lot of time push technology forward as governments try to do anything and everything to win a war but this backfired on the field of ai information is everything in a war knowing your enemy's next step is crucial for winning wars during the cold war the u.s wanted as much information about the soviet union's plans as possible they acquired many reports and documents from the ussr through spice and espionage operations they wanted an instant machine translation that was promised by ai the us government poured millions of dollars and aggressively pushed for instant machine translation but researchers underestimated the challenge of word sense disambiguation word sense disambiguation is the problem concerned with identifying in what context a word is used in a sentence that is a famous story that illustrates the difficulty of solving word sense disambiguation it is said that an early machine translation system translated part of the bible verse matthew 26 41 the spirit is willing but the flesh is weak into russian and back to english as the whiskey is strong but the meat is rotten this was because most machine translation at the time was done using a sort of massive bilingual dictionary it failed to account for word sense disambiguation which led to critical failures throughout the history of ai there were booms and winters or ups and downs we are going to skip all of that and come to the near present and the present ai finds use in anything that requires intelligence that is ai's greatest strength and that's what makes ai so dangerous so to begin the present section let us talk about its application its integration and the ai effect ai can be used to create another ai this is what the technique of nas or neural architecture search is all about an a n or artificial neural network is a type of layered ai that has several layers of data this is a huge over simplification for this video a ns have to be designed by hand nis automates this process to make better neural networks than can be made by hand using ai to make another ai results in better and faster ai creation and development agriculture is one of the oldest and most important professions the agriculture industry has to cope with several problems including increasing demand due to increasing population ai can help solve this problem ai is helping to transform the field of agriculture ai is making agriculture more efficient easier and better soil is indispensable for agriculture monitoring the soil plays an important role in agriculture ai helps farmers analyze their soil health to maximize yields plant x is an app that analyzes soil to detect infected crops using machine learning plant x makes soil detection accessible to small farmers who may not have the money or the tools to get to some of the more inaccessible options the agriculture sector is experiencing a huge shortage of workers as well civilization has moved away from small self-sustaining clustered societies to massive cities where agriculture is outsourced and as interest in agriculture drops fewer and fewer people want to get into agriculture ai is helping to solve this problem too harvest cro robotics has made a strawberry harvesting robot that can harvest eight acres in a day and replace 30 workers it is estimated that 40 percent of the revenue from farming is used to pay workers employing robots cuts down on that and helps solve the worker shortage ai finds use in education as well ai may be bringing about a huge revolution in the education sector teachers cannot be around students at all times this is where tutors come into the picture tutors help students catch up to the syllabus because everybody progresses at different rates quoting bill gates the students who are ahead feel bored and the students who fall behind feel hey i'm not good at this to resolve this problem we are turning to personalized learning assisted by ai for instance thinksters a math tutoring platform uses ai and machine learning to track students growth ai helps analyze the growth of students and their strengths and weaknesses and helps the human tutor address the weaknesses of the students thinksters use curveballs and they call these distractors which are used to throw off the students the ai tracks the students ability to solve similar problems similarly third-space learning a math tutoring company also wants to integrate ai into its services to track progress and optimize learning most of the teachers time is spent on grading tests and assignments ai helps in this process ai can great papers it is easier for computers to grade objective questions you may be familiar with grammarly which can detect grammar mistakes and spelling errors and can determine the complexity of the sentences ai can even help in grading essays but ai grading can be volatile we should not implement half big technology with loopholes because that could be a lot at stake there have been many cases where terrible ai grading implementation ended in disaster a student found a way to cheat at genuity's grading system lazar simmons school used software called ingenuity to grade essays initially lesser got bad grades even though he wrote acceptable and appropriate answers but he found a way to game the system a genuity's algorithm looked for certain keywords and depending on the prominence of these keywords it would grade the essay this is terrible horrendous implementation of ai grading especially because this software is used by 20 000 schools it is a massive oversight of the fact that a perfectly acceptable answer can be given without the use of particular keywords human connection is imperative in the learning process in this sphere ai does not aim to replace human teachers but to aid them because even if artificial intelligence had emotions we would not feel a human connection to ai replacing humans with ai in every field is not wise since the human connection is crucial in some professions despite ai's capability to feel emotions humans may not have the capability to empathize with ai teachers have a huge amount of work and responsibility aside from the grading mentioned previously they have to control a class of around 30 students while teaching an increasingly bigger syllabus ai can assist teachers by collecting data on students and tracking their progress or using speech recognition to control students ai can make learning more accessible to students with disabilities for instance ai speech recognition can be used to show subtitles for what the teacher is teaching ai makes teaching more efficient and makes the classroom more accessible and approachable the future of ai in classrooms is not ai led but ai assisted david kellerman an engineering professor used ai to make a bot that can answer questions relevant to the lesson and also that can suggest videos of a lecture to innovate we don't always have to create more complex new technology the technology of today is amazing innovative creative use of the brilliant technology of the present can do wonders david kellerman says that students are always passive observers or consumers we need to have that active participation since the implementation of the bot and interaction through microsoft teams he says that the class satisfaction went from 75 to 99 students started helping each other and the classroom transformed into sort of an ecosystem please note that this was before the pandemic ai may transform the role of teachers by freeing them from mundane automatable tasks it also gives more time for creative and unautomatable tasks so that the teachers can spend more time with the students and also personally tracking the progress of the students ai is transforming the classroom in a way that allows teachers to establish a more solid connection with their students it also helps all students progress at the same rate through personalized learning ai has a long history in finance especially in fraud and suspicious activity detection in 1987 security pacific national bank used artificial intelligence to detect the use of fraudulent debit cards before the use of ai in fraud detection reports had to be reviewed manually which was a very cumbersome job due to the inefficiency of manually reviewing banking activity some banks considered forgoing the step to minimize their losses nowadays ai is widely used in banking using ai is beneficial to both the bank and the clients despite the expensiveness of development of ai in the 80s ai paid for itself in minimization of loss for the bank ai reduces fraud by detecting suspicious behavior patterns and it flags them for instance let us say a person withdraws his salary rupees 40 000 every 30th of the month but one day he receives a huge sum of money two or three times his salary that is withdrawn the same day this would raise a red flag if the ai deems the activity to be money laundering it would alert the bank or if the ai deems it to be a stolen debit card it'll alert the person to whom the card is registered ai finds fascinating use in trading high frequency traders use artificial intelligence to make a huge number of trades and make split-second decisions using a massive amount of data it uses very complex algorithms to detect and act on market trends and gives traders who use hft massive advantages the advantage is so massive that investors exchange our iex and new york city stock exchange uses 38 miles or 61 kilometers of cables to delay the exchange in order to give traders without hft a chance hft is a very cutthroat field the competition is heavy firms and traders try to reduce the latency or delay of the data flow every millisecond microsecond and picosecond costs money so companies and traders try to move closer and closer to the exchange sometimes even inside the exchange hft has been met with criticism like the above mentioned unfairness that it breathes such quick decisions can move the market in unreasonable ways for instance the dow jones industrial average dropped by a thousand points for just 20 minutes it was found that a big order triggered off a sell-off which caused a drop according to nasdaq 50 percent of trading in the us consists of hft hft has its advantages and disadvantages but hft is here to stay for better offers ai is also used in underwriting underwriting is a service where financial institutions guarantee payment in case of a damage or financial loss and agree to bear the financial risk of such a guarantee ai is used to underwriting to assess the risk of consumer of a consumer of default this helps the underwriters increase their profits this is also beneficial to the customer as they get faster insurance as ai is faster and better assessing risk ai underwriting is an emerging field it is not established yet so far ai underwriting is mostly done by startups that offer their services to banks and other financial institutions humans are extremely biased you would think that ai would be an exception ai is designed by humans and the data fed to ai is selected by humans for example amazon was developing a hiring algorithm that would read a resume and choose the best candidate it recognized words such as women's and women and marked it a negative score which caused the algorithm to be biased towards men you may be familiar with the personalized ads served by facebook the algorithm targets specific people to maximize their probability of clicking facebook's algorithm served women with ads for nursing and secretary jobs and served men with janitor and taxi job ads again it is important to remember that most ai is not bond biased but taught to be biased buyers can be caused by many different reasons first of all the sample could be the cause of bias machine learning algorithms use huge amounts of data to do their tasks if the sample itself is biased then the machine learning system learns to be biased for instance a hiring algorithm to work well it needs to be fed with a huge amount of unbiased data if the humans who are contributing to the data are biased the hiring algorithm mimics and amplifies their bias ai could also be purposefully trained to mimic real-wo
ld biases facebook's ai is a perfect example of this prejudice bias ai could also be biased due to exclusion this is known as exclusion bias if a facial recognition algorithm is trained using only pictures of white people it may not be able to recognize the faces of people of color exclusion bias occurs due to an important data point being left out ai doesn't only mimic our biases but amplifies them this is because some ai work on generalization they are also optimized for accuracy but they are not optimized for unbiasedness when a 75 percent of a times a kitchen is shown with the presence of a woman an image recognition ai sees that and it generalizes and automatically assumes that every time there's a kitchen there is a woman to the ai that 25 is insignificant as it is mathematically inconsequential bias hurts the people who are subjected to it ai finds itself a part of many different systems and bias in ai can hurt people in many ways the bias mostly hurts people of color and women who are already oppressed and marginalized but ai is way better in making unbiased decisions than human ai only considers relevant criteria to make a decision so how do scientists and researchers fix the bias in ai it is important to understand that fixing ai bias is a very tough undertaking first the algorithm should be free of bias the algorithm should be tested for fairness and unbiasedness before being implemented skipping this test can lead to disastrous results for machine learning algorithms the data fed should be representative and inclusive when developing a product or technology for all people white brown or black male or female the data should fed should be representative of all demographics there are also ai guidelines like google's ai guidelines and ibm fairness 360 guidelines that ai development must follow to prevent ai bias we must also diversify the ai field a more diverse development field or management would be better equipped to review and iron out ai bias we're diving into a world of ai leader ai assisted technology when we give ai such an important role we must make sure that doesn't hurt anybody due to bias a bias has already affected the present if we don't iron out bias in ai it will only become more damaging in the future ai is helping us explore space the final frontier it is changing the very nature of space exploration it is estimated that the ai space exploration market is worth two billion dollars space is a very dangerous place the vacuum is filled with hidden dangers like radiation meteors supernovas and so on understanding and avoiding these dangers is crucial to effective and successful space exploration scientists use ai to chart unknown or uncharted galaxies supernovas and stars and even black holes ai is also used to study planets and exoplanets by utilizing data gathered from probes rovers and observatories nasa discovered the exoplanet kepler-90i by using machine learning and ai ai is also helping humans explore space ai helps astronauts communicate with earth and also keeps them company and helps them not feel lonely self-driving cars have been a staple of fiction self-driving cars are now a reality due to ai fully autonomous cars or self-driving cars use ai to navigate traffic and find their way to the destination the degree of automation varies from driver assisted to full automation ai helps drivers be safer ai can take control or assist in an emergency it can also monitor blind spots it can also break in case of a possible accident it is presumed that the mass use of self-driving cars will reduce crashes by 90 percent of course there are many advantages to self-driving cars but several technical moral and legal challenges must be overcome to make mass use of self-driving cars a reality making cars that drive autonomously is a massive challenge driving is an activity that requires a massive amount of data processing and requires many split second decisions it requires a massive amount of data for ai to learn how to drive here are the technical challenges in the present for self-driving cars social interaction is surprisingly important for driving i am not talking about the normal kind for safe driving drivers must communicate to drive the ai must know the intention of the other drivers human or ai as of now it is very difficult for ai to perform this complex social interaction so what do we do to fix this ai could communicate with other self-driving cars and understand their intentions but could take a while for self-driving cars to storm the streets but for now ai will have to interact with humans ai will have to be trained to understand humans and act accordingly there's also loss of independence and control that comes with handing over control of your car to an ai there is also the perceived unsafety of driving a self-driving car even though self-driving cars are actually safer than cars with human drivers in most situations this is the reason why people are afraid of flying we perceive flying is more dangerous although the risk of fatality in car crashes is one is to 14 and airline crashes is one in nine thousand eight twenty one your tow times more likely to meet our end by choking on food than an airplane clash we must put in place several safety measures to make sure that people feel safe in a self-driving car there are also moral and legal issues with self-driving cars self-driving cars have to be approved for public use the problem is that we won't have enough data about their safety until they are approved car fatality are one mile in a hundred million miles but google has only driven their self-driving cars for 1.3 million miles that amount of data is not enough to conclusively determine the safety of self-driving cars another issue is cyber security self-driving cars can open up the possibility of hacking or a software compromise malware could even be used to compromise a self-driving vehicle cyber security is an important and serious issue but it is fixable via proper testing and consistent updates to the software but it is impossible to know exactly what sort of cyber security issues we will face in the future with self-driving cars tesla has announced a fully autonomous car beta it has faced a new problem and caused a few crashes it seems to be a software issue the thing about self-driving cars is that most issues are software based and can be fixed with a software update from artificial narrow intelligence ai will progress to the next level what is known as artificial general intelligence or agi artificial general intelligence is a type of ai that may emerge in the future that will be able to perform any intellectual task that a human can artificial general intelligence is prominently featured in several fictional works how can we determine if an ai is intelligent enough to be considered general ai many ai researchers and scientists have proposed some tests to determine whether an ai can be considered an agi the first and most famous test is a turing test proposed by alan turing in 1950 which stated that for an ai to be considered intelligent it must be able to carry a conversation with a human that is virtually indistinguishable from a human the second is the coffee test proposed by steve wozniak co-founder of apple to pass the coffee test the ai must be able to navigate a standard american household and find out how to make a coffee the third test is the employment test proposed by ben gertzell in this test ai has to enroll at a college and obtain a degree using only fat tools that are available to all students when will ai progress to the point of being considered agi it is extremely hard to pinpoint a time in the future when agi would be developed as i have pointed out many times before we are modeling ai after the unknown so we do not know what advances need to be made to make agi a reality furthermore it is impossible to know when these breakthroughs will be made a great breakthrough may be made tomorrow that may put agi in striking distance reika juve a futurist believes that there's a 50 possibility of agi being achieved by 2099 while his contemporary rodney brooks co-founder of irobot believes that agi will be developed only by 2200. the reason for these inconsistent predictions is the unpredictable nature of ai development besides the method by which we are able to achieve agi are still unknown some believe deep neural networks machine learning and big data will help achieve agi whereas some research researchers think that the technology to achieve agi has not yet been invented some researchers think that the basic tools to create agi have already been invented and all that is required is time and effort we have so far talked about the greatness and boundless benefits of ai so now let us talk about the dangers of intelligent ai despite massive benefits that may come and have come out of powerful ai there are also bound to be some risks that come with agi so what are the risks that come with intelligent and super intelligent ai ai is coded with efficiency in mind and without common sense this can be dangerous nick bostrom philosopher and professor at the university of oxford gives the paper clip thought experiment if we instruct a super ai to make paper clips and we do not code common sense into it it'll be relentless in its pursuit to make as many paper clips as possible and will continue making paper clips if it deems that the best way to make paper clips is to wipe out humanity and make the earth into a paper clip factory it will do as it has no consideration of the cost and gravity of its actions and all it cares about is making paper clips nick bostrom also points out that the existential threat of ai does not come about due to a sudden spark of consciousness that makes ai resent humans and actively seek to destroy them but rather a misalignment of intentions that inadvertently ends up hurting us to prevent this we must code common sense into ai but trying to code common sense into ai or more specifically machine learning works machine learning works by identifying patterns and training data and recognizing what needs to be done so far there have been a few moderately successful attempts at common sense for instance agency a leading computer science and ai researcher set out to create an ai that could understand the intent behind the sentence the ai that she created examines blogs and idiom entries from wikitionary and takes out the names from the sentences extracted and crowdsources the meaning of the sentences and uses them to learn the implied meaning behind a sentence for instance if the sentence george prepared dinner for his family on christmas is given to the ai then the ai can recognize that george is trying to impress his family this ai has 50 chance of being accurate there is of course more to common sense than just understanding the implied meaning behind sentences to have common sense ai must know a lot of basic facts humans have more capabilities than pattern recognizing machines one of the possible approaches may be this let's say we scan the entire brain in immaculate details down to every neuron and run a computer simulation of the brain so perfectly accurate that it works indistinguishably from a brain this sounds perfect right scan the brain stimulate it and boom done well there are several issues involved in this approach the first is the possibility of the fundamental concept of this approach will we be able to scan the brain in such detail that we can make a perfect simulation of the brain certain recent advancements in the field of neuro imaging have made it possible to image and map the brain in great detail there are different techniques of neuroimaging used for different purposes for instance cat scans are computed axial tomography is typically used for quickly viewing brain injuries and mris are magnetic resonance imaging and fmris and ped scans are other types of neural imaging techniques combining all the data from these different techniques we can build a great model of the brain but it may not be adequate for a full one-to-one simulation of the brain well let's say we somehow managed to fully scan and map the brain there is still a problem the brain is extremely complex and using current computing power and even the most powerful and expensive supercomputers we cannot simulate the entire brain advocates of this approach point out that we are expected to reach this level of compute computing power required to simulate the brain in a few decades since the time frame of these advances necessary for the success of this approach is inconclusive i will leave the approach and let the viewer make up your minds on the possibility of the approach the next stage we have to progress to after agi is artificial super intelligence or asi related to the amount of time required to go from narrow ai to general ai the leap from agi to asi will be shorter artificial super intelligence or super ai or asi is ai that is better than humans in all areas of interest it will exceed the capability of humans and agi in all domains asi is where the capability and application of ai become interesting up to the point of asi ai makes our lives easier and makes other achievements easier to attain but asi will unlock completely new opportunities for the future of mankind it will help humans propel themselves into a completely new era of advancement it could be the key to achieving exploration beyond our solar system and achieving mind deploring and other great possibilities do we need asi will it spell doom for us is the rapid advancement of ai a good or bad thing ai is found in more and more things now and a small issue in ai could have disastrous consequences the reason for why ironing out any problem in ai is so hard and finding out ai's flaws and intentions are difficult are that ai is a black box when data is fed the output comes out the actual process of the transformation that takes place is veiled the way that machine learning algorithms work makes this veiled nature of the process inherent to ai if the black box is exposed it can be tinkered with and a person with malicious intent could make the ai act maliciously there are also advantages to having the black box exposed but it is impossible without getting machine learning and disfiguring it to the point of being unrecognizable but transparency of the development of ai is indisputably good closeted development of ai will pave the way for powerful possibly malicious and nefarious ai open transparent artificial intelligence development will allow the public and other coders to know exactly what the ai is trying to do and can detect if the ai is biased or if there is a flaw in ai if the power of ai is concentrated into the hands of the rich powerful and influential and the development of artificial intelligence is kept secret the future of ai is grim fortunately many tech giants like elon musk and google are striving to make ai open and for all using open ai elon musk is trying to prevent any catastrophe that may entail in the privatization and concentration of the power of ai in the hands of powerful agents open ai's goal is to develop powerful useful and capable ai for the benefit of the people of their ai is open source that is the code of the ai is publicly available and open for review and examination by all ai being a powerful tool it can be used for good or bad the use of ai must be transparent if the government starts using ai to develop a secret weapon that can be used to cause massive damage it could be devastating transparency in ai development may also prevent us from losing control of ai super intelligent ai must be always monitored and thus there must be some safety measures in place like a kill switch you