Artificial Intelligence & Robotics
Artificial Intelligence as the name says, its composed of two different words namely, artificial which means it is not natural and is human made and intelligence which means ability to interpret and process information and respond according to environment. Intelligence is usually possessed by humans as it’s a god gift. Humans using their intelligence have carried out huge game changing inventions such as electricity, computers, space travel and many more, it’s a never-ending list and will go on as long as there is life. Humans also using their intelligence built artificial intelligence which is basically a machine doing tasks given to it. According to Wikipedia, “Artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. In computer science AI research is defined as the study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, the term "artificial intelligence" is applied when a machine mimics "cognitive" functions that humans associate with other human minds, such as "learning" and "problem solving". In simple terms, it is the ability of a machine to respond to commands given by humans. For example, searching a query on google is also a form of Artificial Intelligence. This was all about what is Artificial Intelligence, now the next thing is how big is AI. Artificial Intelligence is almost used in every possible field such as in smart phones, in computers, in cars, medical to do surgeries, in space industry and also to carry out research. AI is a vast field of study and my focus here is on Artificial Intelligence and Robotics. There are various issues and concerns being raised in AI and Robotics since this field has seen the most growth in recent years specially after the Hanson Robotics building their Humanoid named Sophia. In the following paragraphs I will discuss the history, present state, future and limitations and advantages of AI and Robotics.
AI research was founded in 1956 in at a workshop in Dartmouth College which was attended by Allen Newel, Herbert Simon, John McCarthy, Marvin Minsky, and Arthur Samuel which also became the founders and leaders of the AI Research. They started building computers and started training them on the game of checkers. By 1960, machines became more stronger than an average human in the game of checkers. The machines were good at various stuff such as solving word problems in algebra, proving logical theorems and also speaking English. US Department of Defense invested heavily in AI research and leaders and founders of the research team were also optimistic. Herbert Simon stated that “machines within 20 years will do things which a man can do” . Marvin Minsky agreed and added “within a generation, problem of creating artificial intelligence will probably solved”. The founders and leaders were probably right because in today’s world there a lot of machines to help us out in our everyday life such as robotic vacuums, sensor lights and more. The AI. Research slowed down in 1974 which is also called the AI Winter era, that is, research went all cold due to various reasons and pressure from Congress in US to invest more in more productive projects. In early 1980’s the AI research revived due to commercial success of expert systems which could simulated the skills and knowledge of human experts. Not only this, Japan’s 5th generation computers inspired US and British Governments to restore investment in AI research. By 1985, the AI research crossed Billion Dollars but again went cold in 1987 after the collapse of Lisp Machine. Late 1990’s and early 21st Century marked the use of AI into fields of Data Mining, Medical Diagnosis other areas due to its computational power. Deep Blue became the first computer chess-playing system to beat the reigning world chess champion, Garry Kasparov on 11 May 1997.
In 2011, a Jeopardy! quiz show exhibition match, IBM's question answering system, Watson, defeated the two greatest Jeopardy!champions, Brad Rutter and Ken Jennings, by a significant margin. Faster computers, algorithmic improvements, and access to large amounts of data enabled advances in machine learning and perception; data-hungry deep learning methods started to dominate accuracy benchmarks around 2012. The Kinect, which provides a 3D body–motion interface for the Xbox 360 and the Xbox One, uses algorithms that emerged from lengthy AI research as do intelligent personal assistants in smartphones. In March 2016, AlphaGo won 4 out of 5 games of Go in a match with Go champion Lee Sedol, becoming the first computer Go-playing system to beat a professional Go player without handicaps. In the 2017 Future of Go Summit, AlphaGo won a three-game match with Ke Jie, who at the time continuously held the world No. 1 ranking for two years. This marked the completion of a significant milestone in the development of Artificial Intelligence as Go is an extremely complex game, more so than Chess.
According to Bloomberg's Jack Clark, 2015 was a landmark year for artificial intelligence, with the number of software projects that use AI within Google increased from a "sporadic usage" in 2012 to more than 2,700 projects. Clark also presents factual data indicating that error rates in image processing tasks have fallen significantly since 2011.[48] He attributes this to an increase in affordable neural networks, due to a rise in cloud computing infrastructure and to an increase in research tools and datasets.[11]Other cited examples include Microsoft's development of a Skype system that can automatically translate from one language to another and Facebook's system that can describe images to blind people.[48] In a 2017 survey, one in five companies reported they had "incorporated AI in some offerings or processes"(https://en.wikipedia.org/wiki/Artificial_intelligence#History).
Though the future of Artificial Intelligence and Robotics remain uncertain but we can still hold onto to its present state. Dan Rufener of New Relic attended International Conference for Learning Representatives (ICLR) where he learned about intriguing projects in AI. He describes one of the papers named Zero-Shot Visual Imitation in which the developers trained a robot using AI to tie a knot and navigate through the building in just one demonstration without any specific guidance. In another presentation named Integer Deep Neural Network Training, attendees saw how an autonomous bicycle can follow people on its own. The conference also revealed various limitations AI faces such as LabSix fooled an AI robot with a 3D printed Turtle which was a rifle for the robot. Also, Machines still lack general intelligence unlike humans, but they are getting smart and will have huge impact on our society economically. (https://blog.newrelic.com/engineering/state-of-artificial-intelligence/)
Robotics is very essential in modern era. Though there are industrial robots which stay indoors at work place, do not resemble humans and some also speak but not as humans. These industrial robots work in assembly line, that is, they do whatever task is given as command to them. Predictions were that by 2018 every household will have a robot as an assistant at home which obviously didn’t happened but we do have very advancements in the field of robotics where humanoids are being built but still being used only for industrial purposes. Though reaching ultimate goal where these humanoids can decide what to do and make decisions on its own could be a risk factor as well. Though the use of AI and neural networks have completely changed the scenario of robots but still these algorithm still work on old technology which was founded in 1940’s to 1950’s. What it actually needs is to shift focus from old technology to new technology so that these robots can interact as well like a human does with the people. (https://www.technative.io/the-state-of-robotics-in-2018/)
This talks show aired on FRESH AIR, host Terry Gross interviews Cade Metz who’s a tech journalist and he told a lot about AI and robotics and neural networks. The show starts with Gross introducing audience about AI and technology followed by introducing Metz to the audience and asking him to talk about AI. Metz answers Gross by explaining that anything that can be automated and do things on its own is artificial intelligence but the real change which became a boon for the AI industry was in the past 5 years when scientist started building new algorithms called Neural Networks for AI capable machines. According to Metz, neural networks are systems which learn things on their own, which he explained by giving an example how Facebook distinguishes between photos of people based on their face because on the back end Facebook uses neural networks in order to identify and sort images based on people. Metz further explains that the neural network basically mimics the neurons of a human brain but it is all mathematical algorithms which makes it do that. Further explained by another example in which Metz describes how these neural network can be used to scan human body for identifying signs of cancer and it works on the same principle as Facebook. Gross talks about how smart speakers work like Google Home, Amazon Echo or Apple Home Pod Mini to which Metz replies that on the same principle as other AI works, that is, they are made to learn and respond to basic commands given by humans. Metz also mentioned the limitation here AI faces is that it can respond to anything human commands because it can just mimic human brain but developing an artificial brain which work absolutely same as a human is quite far and limits the neural network. Metz also explained that speech formed in response to the command can be human generated (in case of Google, because it learns too much from human given commands) whereas, in devices from Apple, Siri is a computer-generated voice. Metz also describes the limitations of AI, that anyone could make them do anything and fool the system which he further elaborated using the example of self-driving cars which use same AI and neural networks to operate, but a minor fault could result in vicious consequences. So basically AI is not mimicking human brain because scientists do not know how the brain works since its so complex and if scientists knew how brain works then they could have built a brain in digital form transferring data between neurons. In reality AI works on pyramid structure in which the base consists of all the operations and algorithms which are used to transfer data on the top layers which is called learning. AI cannot respond to complex situations a human brain can or can’t understand how emotions work as human brain can. AI learns things by traditional method of trial and error but its so fast that as soon as it learns something it rapidly expands itself to learn more and all the possible solution, for example a game of chess in which AI can learn all possible moves in fraction of seconds then a normal human. AI can be hacked since its just a machine and software and can be tampered with. Some scientists fear that AI could surpass humans and won’t allow them to turn them off and its quite possible because AI works on neural networks which means they learn things, that is, if they are made to learn something and they like it they may not allow someone to turn them selves off. AI is very useful since people are using Machine Learning algorithms which design much better software than humans, that is, AI is building AI.
(www.npr.org/2018/03/15/593863645/robots-are-now-creating-new-robots-tech-reporter-says)
AI and robotics will bring out a revolution in the whole industry by replacing humans at work places specially those areas which require extensively human labor. Amazon has already started this trend where both robots and humans work together at their warehouses. Robots is the future of our society but only if combined with Artificial Intelligence and Neural Networks in a specific way which would benefit the humans. Though assuming that robots will replace humans completely is still undetermined reason being that there is no bridge between knowledge and data technology and it wouldn’t even happen overnight. Instead it would take decades to bridge that gap. Also, if they do so it can’t completely outcast humans since how can we trust a machine and also it would require human supervision. Revolution is coming but won’t happen overnight and won’t affect every sector at the same time, it will spread slowly among the sectors and would give society time to accept it.
(http://shapingthefuture.economist.com/robot-revolution-ai-and-the-future-of-work/)
Everyone agrees that for a robot to work in an assembly line is pretty easy than being intelligent like a human. And in fact, this is true because robots do not have that intellect which humans do, but making them do things which us humans can do easily through artificial intelligence is a challenge for humans. We know computers can work and process tasks very quickly using AI because they already have the data stored in it and computers processing power for data is very fast. Also, AI means developing a machine that could implement human thought process which would mean that it could think and respond to situation instantly using their artificial intelligence just like humans do. But Roboticists are nowhere near this, in-fact they have achieved this kind of intelligence using limited AI. Some modern robots even have the ability to learn patterns and actions in a limited environment. In this the robot is shown a certain action once for instance, a hand shake which the robot stores in its memory and performs the same action next time whenever robot encounters that similar situation. Some robots can also interact socially such as Kismet, a robot at M.I.T’s Artificial Intelligence Lab. Kismet responds back by recognizing human body language. Kismet and other humanoids at M.I.T’s Artificial Intelligence Lab operate on a unconventional control structure which the is a low-level structure and program’s director Rodney Brooks thinks this model is accurate because it does most things automatically just like we all humans do. The real challenge which still rests is how to make AI understand natural intelligence and implement it. Roboticists do understand that human brain contains billions of neurons attached to each other transferring electrical signals but where do they add up to answer reasoning is still incomprehensible. Scientists believe that robots will change the way we look at humans and human intelligence, working along with humanoids gives the best way to understand human intelligence but there’s also a risk that the working along with humanoids at work places would turn human generation into cyborgs. Either way, robots will play a key role in future where they would move from industries to regular households just like computer did in the 80’s.(How Robots Work” Tom Harris https://science.howstuffworks.com/robot6.htm)
Everything comes with its limitations, so is AI and robotics. A few of the the typical limitation any kind of AI would have is that it lacks common sense, that is, it cannot process simple things own its own rather it requires training even to pick a glass. Whereas a new born human baby tends to grab things as he/she grows but a robot with AI still needs to be taught how to garb things. In technical terms, AI uses a method called “backpropagation” also called backprop which requires extensive set of values. Moreover they are hard to debug since it works on neural networks which means requires a mathematical geography to be debugged unlike regular programs. It’s a kind of black box who’s contents cannot be defined as reliable unless they are out of the box. Also, as mentioned earlier they require knowledge since they lack common sense which makes them shallow.( https://www.wired.com/story/greedy-brittle-opaque-and-shallow-the-downsides-to-deep-learning/) Robots embedded with AI are very costly and are limited to industrial use for now, they lack emotions and will struggle to make a judgment. Also they can’t improve overtime since they don’t have the ability to learn and gain experience with time as humans do. Robots could also replace humans at work place in near future which means cutting out human labor which would result in less jobs for humans in near future.
AI and robotics have an upper hand in various aspects over humans, since they are machines and highly trained on data, chances they will make mistakes are very less likely to happen. This means that robots embedded with AI will be much more accurate than humans. Humanoids could be used for space exploration since they can be programmed to last longer in space than actual humans and they won’t even have problems which normal humans do. Another benefit of using a humanoid in space exploration is that they could upload data over the cloud fast side by side. Other advantages include in the medical industry where humanoids could assist doctor in carrying out operations or in car industries where they can be used to build cars faster than humans. Replacing humans by humanoids is a great advantage for owners because it’s a one-time investment and output would be much more than before ignoring other factors such as scheduled maintenance of these machines. These humanoids could pe made to work longer tha humans and would boon the production unit of any industry using humanoids. They can also be used in Wars as soldiers by the defense department.