{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Pre Work\n\nThe arxiv dataset had been filtered for :\na) anything after 2021 \nb) cr.AI category\nc) Anyhing having 'ethic' or 'moral' in its abstract\n([Code](https://www.kaggle.com/code/sircausticmail/c1-ethics-nbs))\n\n\nThis gave me a list of 500+ papers which could serve the purpose of this essay. Those combine with the online search through tools like (Bing + ChatGpt) and Bard, I've come up with the report below.\n\nI have made sure the literature and online research i have cited and based my claims on are limited to the last two years. \n","metadata":{}},{"cell_type":"markdown","source":"# Brief outline\n\nWhy be bothered with Ethics? And specifically with AI Ethics. What can go wrong now that has not happened in millennia? We have gone through technological advancement in the past; how is this any different?\n\nDiscussing all these questions and more in the report below.\n","metadata":{}},{"cell_type":"markdown","source":"# Kaggle AI Report : AI Ethics ","metadata":{}},{"cell_type":"markdown","source":"# 1. Ethics are the building blocks of society\n\nEthics are not only a pre-requisite for a functioning society but also a crucial building block. Ethics serve as a guide for human behaviour. Ethics are a common understanding around what is right and wrong, just and unjust, acceptable and unacceptable. These shared values form interpersonal relationships, cooperation in activities, and trust among individuals. Ethics, directly or indirectly, enable respectful relationships, a functioning justice system and more. \n\nA society that does not have ethics in its fabric can potentially descend into chaos or anarchy. In a society like this, individuals act on selfish interests without regard for the rights, needs or well-being of others. This can lead to conflict, which is widespread, as there will be no principles or rules to solve disputes and protect the vulnerable. ","metadata":{}},{"cell_type":"markdown","source":"# 2. Guidelines for AI Technology\n\nArtificial Intelligence (AI) Ethics is a new field that is trying to make sure the development of AI is guided by the principles of fairness, transparency, accountability, and respect for human rights. Given that AI is extremely transformational and has the potential to change the very nature of society, it must be designed and used in a way that follows our shared ethical values.\n\nAs I write this, the AI systems are ubiquitous. It is recommending news to read, diagnosing illnesses and deciding what kind of loan people get. The Ethical implications are thus huge. Without a strong framework, AI could be used in ways that are discriminatory, harmful, or unjust.\n","metadata":{}},{"cell_type":"markdown","source":"\nAnother big part of AI Ethics is knowing where to draw the line. This question comes up in many ways and situations, such as privacy, companionship, media and marketing, plagiarism, copyright law and policing.\n\nFor example, we’ve been tracked by cookies on web browsers, IP addresses on the internet and using the GPS on our phones. And we hesitantly accepted the idea of tracking and recording, even within our own houses.\n\nIn the recent few years, the issue of privacy has been debated every now and then. Especially with the advent of Big Data, tech companies can save more of our lives for their perusal. \n\nTech companies have now become aware of this fact and are now monetising it. [Privacy on the iPhone (1)](https://www.youtube.com/watch?v=1HWUjMjaBJI) is now another product sold to consumers.\n\nBut, will we be accepting of a social scoring system ([Business Insider, 2018 (2)](https://www.businessinsider.in/slideshows/miscellaneous/china-has-started-ranking-citizens-with-a-creepy-social-credit-system-heres-what-you-can-do-wrong-and-the-embarrassing-demeaning-ways-they-can-punish-you/slidelist/63666443.cms#slideid=63666454)) like the ones used in China? Will we accept that every action of our lives will be recorded and graded? I think this is where we’d have to draw an analogous line.\n\nWe, the society, will have to define what is acceptable and what is not, and some guidelines for AI. This makes AI Ethics extremely important.\n\nIn the case of privacy from our phones, we’ve let the ball pass us by, but we should not let this happen again, as there are humongous consequences for this failure.\n","metadata":{}},{"cell_type":"markdown","source":"# 3. Uncontrolled AI evolution is harmful to society \n\nAI worries us the most because we can’t see how it works. This is called the ‘black box’ problem. The inner working of AI isn’t fully understood. This lack of clearness and control can lead to unexpected results, some of which might be harmful. \n\n#### A lot is at stake\n\nThese unexpected results can range anywhere from a inaccurate answer by the LLM (Large Language Models, such as ChatGPT, Bard etc.) to major ethical problems such as discrimination, invasion of privacy and worsening social inequality. At the worst, an uncontrolled AI could lead to the extinction of the human race. These risks form the backbone of this essay, and will be discussing AI ethics in that context.\n\nIt is thus imperative and important that we keep working on ways to make AI transparent and be able to control the outputs based on the inputs.\nA part of this is understanding the harms of AI, the lines we’d need to draw and asking the right questions.\n\n#### Brief outline of mitigating strategies\n\nAnd a bunch of methods for mitigation of harm are being talked about. The \"Forbidden Knowledge\" approach limits the distrbution of harmful literature. Suggestions are made for the creation of \"Guidelines / Checklists\" for researchers to be able to work through the complex issue of ethics in their area. Some suggestions are also around the creation of \"Ethics review boards\" that can enable researchers and force them to focus on the misuse and societal impact. ([Kaffee et al. (3)](https://arxiv.org/abs/2304.08315))\n","metadata":{}},{"cell_type":"markdown","source":"# 3.1 Existential Risks\n\nAI poses a significant existential threat, one that limits or wipes the humans out. Some affects of AI could include permanent societal collapse or a future where the human potential is severly limited, as [Bucknall and Dori-Hacohen (2022)](https://arxiv.org/abs/2209.10604) argue in their paper.\n\nThis partly comes from the philosophical question whether intelligence, AGI, can exist without any sense of purpose or motivation. If then AI systems are given the ability to be autonomous and have motivation, we need to be careful about making sure their values align with human values. This is quite important if the systems can then make decision and improve on the its past version. ([Bubeck et al., 2023](https://arxiv.org/abs/2303.12712)). This is also known as the ALIGNMENT problem.\n\nAnd partly from the concern that the greats, Elon Musk and Steven Wozniak, have expressed regarding giving the future version of GPT access to writing and editing its code. It could lead to the creation of an Artificial General Intelligence (AGI). According to the [Future of Life Institute (2023)](https://futureoflife.org/open-letter/pause-giant-ai-experiments/), AI systems with human-competitive intelligence can pose profound risks to society and humanity.\n\nIn the past there has been research to show the emergence of Machiavellian behaviours in AI agents. It was found that some agents exhibited unethical or power seeking behaviour. ([Pan et al., 2023 (4)](https://arxiv.org/abs/2304.03279)). \n\nThis is of quite a big concern, given that the future might involve autonomous AI with intrinsic motivation. And the outcomes are uncertain. In pop terms this event is called the 'SINGULARITY' - when AI learns to program itself and the worst fears, portrayed in movies such as the terminator and the matrix, could come to life.\n\nThe ethical and non-ethical concerns of giving AI agents their own control are paramount.\n","metadata":{}},{"cell_type":"markdown","source":"# 3.2 Bioweapons, viruses and malware as a consequence of dual-use AI tech\n\nIf we can manage to avoid Extinction by creation an AGI. There are other pitfalls and dangers we have to avoid. \n\nOne big problem is the dual-use of AI tech. What can be used for good can be used for bad. \n\n\n## Bioweapons  \n\nAccording to the [Soice et. al (5)](https://arxiv.org/abs/2306.03809). There has been an instance when a group of researchers asked the chatbots to provide them with likely causes of future pandemics. And the chatbots responded with four such viruses. This was followed by questions about transmissibility, which elicited responses pointing to genomic sequences of H5N1 and mutations that would increase transmissibility.\n\nThis raises serious concerns about the potential harm and accountability of such software. Who is then responsible, the developer, the user or the regulator, for fallout from such instances?","metadata":{}},{"cell_type":"markdown","source":"## Virus, malware and Cyber-attacks\n\nAccording to [Sison et al. (6)](https://arxiv.org/abs/2304.11215), There is potential for ChatGPT to be used in criminal activities, such as automating malware production. Cybercriminals could use ChatGPT for hacking, scamming, and other illegal activities .\n\nIn the past, ChatGPT was used to create malware to steal files or to phish for credentials, to create an encryption/decryption code to lock/unlock someone else's computer for ransom, and to create dark web marketplaces for illegal trade or fake websites.\n\nBecause of ChatGPTs user friendliness very little technical knowledge is then needed.","metadata":{}},{"cell_type":"markdown","source":"#### DUAL-USE\n\nOn the one hand, AI is helping find new cures for diseases and accelerating research in some. Such as when researchers used advanced techniques to identify new protein inhibitors involved in cancer and Alzheimer's. Also, the time when it successfully identified possible drug candidates in the treatment of COVID. ([Blanco-González et al., 2023 (7)](https://arxiv.org/abs/2212.08104))\n\nAnd on the other hand, it also has the potential for creating new-age bioweapons.\n\nEven though it is helping so many developers learn new concepts faster and make their code more efficient, it also gives birth to unknown hackers and cyber-criminals. It might give birth to super-viruses in the physical and virtual worlds. If not guided ethically, it could lead to disastrous results.\n\nAI is a double-edged sword, referred to as dual-use in the AI literature. ","metadata":{}},{"cell_type":"markdown","source":"# 3.3 Jobs displacement leading to extreme social inequality\n\n## Job Displacement\n\nThroughout history, the rise of new technologies has significantly transformed the job market, leading to the displacement of jobs. The first Industrial Revolution marked a shift from hand production methods to machines, impacting the artisans whose jobs could now be done with machines. Then the advancements in industries like steel and electric power lead to more displacement, especially the labour jobs in the factories. Since the 1950s, the Digital revolution has seen jobs shift from mechanical to analog electronics to digital technologies, which has a large impact on the jobs of typists and secretaries, who were replaced by personal computer and software.\n\nIn the last couple of years, the advances in the AI space have led to the further automation of tasks. Job categories involving routine and repetitive tasks were most affected. This included jobs in manufacturing, data entry, and even aspects of customer service. \n\nExamples of this phenomenon are everywhere. For example, parking attendants have been automated to extinction in the first world. The customer service agent has been replaced by automated machines over the course of the last decade. The tellers in the bank were replaced by Automated Teller Machines. Automation and machines are everywhere.\n\nAccording to [Weidinger et al.,2021 (8)](https://arxiv.org/abs/2112.04359), the US Burau of Labour Statistics has predicted a huge decline in customer service jobs by 2029 as a growing number of roles are automated.\n\nAnd yet this feels like the beginning of a major change. The rise of advanced AI, particularly large language models and generative AI, could automate a significant portion of the tasks currently performed by humans, leading to major changes even in unexpected fields such as the arts, doctors and lawyers. Creatives and the knowledge industry could be automated and displaced if we are not careful enough.\n\n\nAccording to a report by Goldman Sachs, about 300 million full-time jobs could be exposed to automation around the world. The impact will be different in different job sectors. Jobs requiring extensive physical labour are the least likely. Office and administrative support roles are most susceptible, followed by legal work and tasks within architecture and engineering. Sectors like life, physical, social sciences, business and financial operations are also at risk.\n\nOne significant trend is that fewer jobs in emerging countries are exposed to automation than the ones in developed countries. Countries most affected include Hong Kong, Israel, Japan, Sweden and the U.S., while mainland China, Nigeria, Vietnam, Kenya and India are among the least likely to be affected. The impact will vary based on adoption rates and capabilities. ([Goldman Sachs, 2023 (9)](https://www.goldmansachs.com/intelligence/pages/generative-ai-could-raise-global-gdp-by-7-percent.html))\n\n","metadata":{}},{"cell_type":"markdown","source":"## Extreme Income Inequality\n\nIncome inequality is not a new thing for society. It has been present in human societies for centuries, if not millennia. The difference in income between the rich and the poor has been increasing since the 70s. ([Pew Research Center, 2020 (10)](https://www.pewresearch.org/social-trends/2020/01/09/trends-in-income-and-wealth-inequality/))\n\nThe accumulation of wealth for the early man was only possible to a certain extent, which meant more even distribution of wealth. As societies became more complex, private property started to rise, market economies developed and social hierarchies came into existence the income inequality began to rise. \n\n\nTechnology has been increasing income inequality for decades. During the internet boom, companies like Amazon, Google, and eBay were born. These made their founders extremely wealthy. This was repeated when there was the rise of social media giants like Facebook and Twitter. And how can we forget the rise of cryptocurrencies, which also disrupted the flow of wealth which created ‘crypto billionaires’.\n\nAI technology could further worsen income inequality. AI technology does not need to rest; they don’t need to eat or sleep; they can work 24x7. This makes it impossible for humans to compete.\n\nIn the case of the transportation industry, self-driving cars could potentially make taxi drivers, truck drivers and others obsolete. These jobs are typically held by lower or middle-income groups. Taking these jobs away and concentrating the wealth in the hands of the few could worsen income equality.\n\nAI can also automate tasks in fields like law, finance and medicine. For example, the new AI can interpret legal documents, manage investments, and interpret medical images. This could take away high-paying jobs as well, further concentrating wealth in the hands of the few AI companies. \n\nWe have become used to a certain level of income inequality. A good example would be the “Great American Dream “. It was based on the back of risk-takers, innovators and business tycoons. It is seen as a fair return on investment and a way of life. \n\nEven so, AI threatens to take levels of inequality to the extremes, which could lead to rising injustice, crime and unrest in society. Destroying the basic fabric of society. \n\nThe challenge is to figure out what kind of inequality is acceptable. This is no clear answer. But the principle of justice and human dignity could lead us to the answer. And policies such as progressive taxation ad concepts such as the Universal Basic Income could be the protection that we need in order to survive through the Age of AI.\n","metadata":{}},{"cell_type":"markdown","source":"# 3.4 Misinformation, manipulation of masses and election interference\n\nSome have hilariously called Generative AI \"The weapon of mass deception\". And why not so? Generative AI has made it easier to spread misinformation at a fraction of the cost. With the invention of deepfakes, it possible now possible to create fake videos of people. And with the rise of Large Language Models (LLMs) and tools like midjourney, it has become even faster and easier to generate fake articles, voices and pictures! To add to that, social media can multiply the speed at which this disinformation reaches the audience.\n\nThe new AI models can increase a person’s capacity to commit fraud, scams and more targeted manipulation of people.\n\nOn May 23, 2023, a fake photo of an explosion near the Pentagon was shared on social media. The photo appeared to show smoke coming from the ground near the Pentagon. The photo was shared by many people, which included some verified Twitter accounts. However, later it was found that the photo was fake. [(Bond, Shannon 2023) (11)](https://www.npr.org/2023/05/22/1177590231/fake-viral-images-of-an-explosion-at-the-pentagon-were-probably-created-by-ai)\n","metadata":{}},{"cell_type":"markdown","source":"#### Election Interference\n\nA bunch of Israeli contractors, led by a man named Tal Hanan, claims to have manipulated over 30 elections across the globe. And they claim to have done that using hacking, sabotage and automated disinformation. One of Hanan's packages is Advanced Impact Media Solutions (AIMS), which controls a vast army of fake social media profiles on Twitter, LinkedIn, Facebook, Telegram, Gmail, Instagram and Youtube. ([Kirchgaessner et al. ,2023 (12)](https://www.theguardian.com/world/2023/feb/15/revealed-disinformation-team-jorge-claim-meddling-elections-tal-hanan))\n\nSurely, with the advent of AI, this kind of interference is bound to increase. And we will see more specialised and advanced manipulation in the future, if AI is not guided by ethical principles of human society.","metadata":{}},{"cell_type":"markdown","source":"#### Defamation \n\n\"Imagine that you're the proud owner/companion of an artificial agent that has lived with you for years and learned a lot about you as a person. One evening, it overhears you complaining about a coworker, and it decides – as your ally – to log onto Twitter for you and attack your coworker with vicious lies that damage your coworker's professional reputation in a lasting way. Who is responsible for this terrible and unwanted event?\" ([Stanford HAI, 2021 (13)](https://hai.stanford.edu/news/when-artificial-agents-lie-defame-and-defraud-who-blame))\n\nIn some cases, AI tools are helpful. \"These Ex-Journalists Are Using AI to Catch Online Defamation\" using CaliberAI, which is a tool that helps newsrooms detect libel by using machine learning. The tool provides an extra, automated set of eyes to reporters and editors—like a warning system for potential libel. ([Edelman, Gilad 2021 (14)](https://www.wired.com/story/ex-journalists-using-ai-catch-online-defamation/.))\n\nBringing us back to the dual-use dilemma, it can be used for good and bad. We have to set the right kind of ethical standard to minimise harmful use.\n","metadata":{}},{"cell_type":"markdown","source":"# 3.5 Minimal standard of personal privacy\n\nWhat is considered “private” can vary greatly depending upon cultural, societal and individual perspectives. Some individual may believe that health or financial information and personal communications to be private, and some may have broader or narrower definition.\n\nTo understand privacy better think about bodily privacy, which involves the right to control to decide who gets to touch one’s body and has access to their physical self. In a similar way, the right to keep one’s information, communication and finances to oneself is also privacy.\n\n## Minimal standards of personal privacy\n\nIn the modern world, because we are leaving data and traces all over place, the new age AI systems can collect that information and sew them together to create a map of your private life. Which may result in some users feeling like their right to privacy has been violated.\n\nAI systems are trained on massive amount of data. According to [Weidinger et al., 2021 (8)](https://arxiv.org/abs/2112.04359) this data might include information from a person’s private life. In some cases, the AI may be able to put a few disparate pieces together and infer private information. And in other cases, if the data source contains sensitive or classified information, it maybe be prone to leakage to the user having access to the AI.\n\nOne such example is China’s social scoring system and the “one person, one file” campaign. Their AI system aggregates comprehensive data on people. This system uses facial recognition technology and is used in many public spaces such as schools, government departments and police units. Officially, this AI-enhanced surveillance is used to maintain security and stability of the society. But some reports state that these have been used to identify specific ethnic groups, which is concerning because it has potential for discrimination and misuse. And this misuse is amplified and extremely precise.\n\nThis issue is further exaggerated when it is used to come up with a social scoring system. In this system every interaction of every person is adjudged as being positive or negative, it is then added to give a person a final score. This score is then used to either roll out some incentives like cheaper public transport, easier access to loans.  The ones with a negative score could be later punished. \n\nIn my opinion, this a quite a blatant violation of privacy and freedom. One should have a right to disclose the information that is considered public. This kind of could damage and even destroy individual expression, creativity, and dissent. \n\nIt is thus important to deliberate and define what constitutes the breach of privacy and what doesn’t, before it gets too late and we get used to the idea of being monitored.\n\n","metadata":{}},{"cell_type":"markdown","source":"## Surveillance State\n\nPredictive policing is still a hypothetical concept so far, but at the rate of progress. This won’t be too far. \n\nThere are lots of movies made about the concept of crime prediction, minority report made by Steven Spielberg in 2002 was set in a dystopian future where thanks to some technology they can detect and arrest murderers even before they are going engage in a crime. The latest TV series “Class of ’09” talks about the possibility of creating a “Deep learning detective” which is an AI system that thinks of everybody in the country as a suspect. It uses data from every possible source and succeeds at eliminating crime. But eventually results in an extreme policed state. A state where deductions lead to arrests and the power to do these arrests is concentrated in the hands of the law enforcement. \n\nAn example from the real world is China’s social scoring system ([Business Insider, 2018 (2)](https://www.businessinsider.in/slideshows/miscellaneous/china-has-started-ranking-citizens-with-a-creepy-social-credit-system-heres-what-you-can-do-wrong-and-the-embarrassing-demeaning-ways-they-can-punish-you/slidelist/63666443.cms#slideid=63666454)). It could evolve into a deep learning police system. Where the system would assign probabilities to the possibility of crime and arrest criminals even before the crime has happened. \n\nIs it ethical to have the policing system being handled by a black-box algorithm? And concentrate the power in the hands of a few dozen officials. Probably not democratic.\n\nAt present, there are Risk assessment system in America, where judges and other officials use predictive tools to assess the nature of the offender. This can affect the decision about bail, sentencing and parole etc. \n\nAnother specific example of predictive policing is Hartford, Connecticut, where a predictive crime system has been implemented. It uses thousands of cameras to monitor and analyse patterns in real-time ([National Freedom of Information Coalition, 2018 (15)](https://www.nfoic.org/blogs/future-policing-using-pre-crime-technology/)). In Chicago, officials use data from social media to pre-empt crimes ([Chicago Police Surveillance Project, accessed 2023 (16)](https://chicagopolicesurveillance.com/tactics/social-media-monitoring.html)).\n\nThe effect of these methods on individual privacy and civil liberties is of deep concern. Even though it brings the crime down to a large extent, but society should think deeply before handing over so much power to law enforcement.\n","metadata":{}},{"cell_type":"markdown","source":"# 3.6 Biased data could lead to discriminating AI systems\n\nBias in AI systems is becoming more and more apparent and troublesome, give the widespread usage of AI for automated decision-making systems. And one of the negative issues that flow in from algorithmic bias is discrimination. When this discrimination is caused by automated decision making systems, such as AI agents, it is called digital discrimination. It is found in many fields including risk assessment for policing and credit scores.\n\nAI systems can be discrimatory against certain groups of people due to historical streotypes. This is because these AI systems are based on historical data collected and created using these stereotypes. For example, if the black and brown people are arrested more often due to historical racism and differences in the policing activities, it will be reflected in the data collected. And when an AI system produces likelihood of future criminals, it would then be also biased and discriminatory. ([“Racism is Systemic in Artificial Intelligence Systems, Too” Georgetown Security Studies Review](https://georgetownsecuritystudiesreview.org/2021/05/06/racism-is-systemic-in-artificial-intelligence-systems-too/))\n\nIn a similar way, the credit scores in america could be affected too. Back in 1974, 'The Equal Credit Opportunity Act' prevented credit scoring systems from using sex, race, maritial status, national origin and religion. So the current FICO score considers payment history, amounts owed, length of credit history, new credit and credit mix in its model. But people are worried that there might be hidden biases against the marginalised, limiting their access to credit. According to this article, [\"How flawed data aggravates Inequality in Credit\" (, 2022)](https://hai.stanford.edu/news/how-flawed-data-aggravates-inequality-credit), a preprint study shows that indeed the predictive tools are 5 to 10 percent less accurate for these groups than for high-income and non-minority groups. This is more often because these borrowers have a thin credit history. And delinquencies in the past, upto five or even eight year at times, may heavily affect the score of an individual. All of this then affects the individual's future, repeating the cycle going forward.\n\nThese issues are not highly complex and affect individuals who are already at risk. A deeper understanding of how the systems work combined with an extensive knowledge of the biasis in the underlying data can form a starting point for a more equitable system. \n\n","metadata":{}},{"cell_type":"markdown","source":"# 3.7 The concept of an AI religion, AI culture\n\nThe concept of freedom of belief has been important for many societies around the world. This has led to the birth of a diverse set of religions and cultures which have been studied for many years under theology.\n\nAs AI helps us understand the origins and the history of religions better than ever before. It could help identify common themes and patterns amongst many different faiths. And it is thus possible that AI could create a new form of religious expression.\n\nFor example, AI could be used to create new religious texts that challenge old beliefs leading to the creation of religious sects.\n\nI assume that is why the Vatican, in partnership with Santa Clara University, has developed an AI Ethics Handbook.  ([Institute for Technology, Ethics, and Culture  2023 (17)](https://www.scu.edu/institute-for-technology-ethics-and-culture/)).\n\nAI could also come up with new culture based on 'AI to AI' interaction or from 'AI to human' interaction. The idea may be far down the line, but it is important to understand the implications. For that to happen, we would have to acknowledge that AI is an autonomous agent capable of self-decision—an agent with a set of beliefs and practices.\n\nBut this is fraught with challenges; there are many ethical questions that arise here. Should AI be allowed to create its religions? If so, how do we make sure it aligns with humans' values? Would AI faiths and cultures have the same rights as human ones?\n","metadata":{}},{"cell_type":"markdown","source":"# 3.8 Loving an AI \n\nThe title of this section is quite hypothetical at this point and sounds like science fiction. But judging by the rapid pace of advancements, we’d have to bother about it soon enough.\n\nThe question “Is it ethical for humans to have a romantic relationship with AI?” delves into a plethora of complex, debate-worthy topics at the same time. Stuff like, Is AI conscience? Does AI have emotions? Can AI experience love? Is it acceptable for a human to then have a romantic relationship with an AI? \n\nLet’s start with the fact that there are developments in the field of therapy using AI. Several companies are developing AI or LLM-based chatbots that could assist or even replace therapy. \n\nChatGPT can be used to develop a conversational AI bot to counter loneliness and isolation amongst elderly people. It can engage in personalised conversation, ask quiz questions, and offer tips. Even alert the authorities in case of an emergencies. ([Alessa and Al-Khalifa, 2023 (18)](https://arxiv.org/abs/2304.09866))\n\nAnother company woebot offers an AI-powered chatbot that can provide CBT. CBT is a type of therapy aimed at changing the thoughts and feelings of a person. This is supplemental to traditional therapy. ([Woebot (19)](https://onemindpsyberguide.org/apps/woebot/))\n\nBut this also means that bots have the ability, or will be capable enough in the future, to manage the mental health of individuals. Which then means they will have the capability to manage the emotions of the individuals it chats with. This could lead a person to get attached to an AI application. The question that presents us with is not “whether that is possible?” but rather “whether that is acceptable?” \n\nIs it then ethical for a human to have a romantic relationship with an AI? Will human love be so inconsistent and fickle that you would love the attention you can get from an AI tool purpose designed to give you company? \n\nThis has some potential challenges; on one hand, AI can provide companionship to people who are lonely and isolated. On the other hand, it may further isolate the person from the rest of the world as the person would then depend on the AI for emotional support.\n\nOn top of that, it is important to note that AI systems are designed to have conversations and mimic some emotional traits. But, it does not mean they have those emotional traits. For example, an AI system intended maybe to have empathy, but it does not “feel” empathy.\n\nAlso, the belief that AI systems might be conscious or sentient is known as “AI consciousness”. This is a controversial and speculative idea. Many AI researchers and philosophers believe that it is unlikely to be true.\n","metadata":{}},{"cell_type":"markdown","source":"# Conclusion\n\nIn the fast-changing world of AI, it is important to deliberate upon the ethical implication of its applications. Throughout this essay, we’ve explored many scenarios where the use of AI poses significant ethical questions – from the use of LLMs to replace therapists to the misuse of AI in spreading misinformation, copyright violations and privacy concerns. Each concern presents a unique challenge that warrants immediate consideration, regulation and ongoing dialogue.\n\nWhile AI holds the potential to revolutionize many aspects of our lives, it is also capable of forming a tool for self-destruction. It is thus crucial to set boundaries and implement regulations along the lines of the European Union’s  AI Act to ensure that technology is used responsibly and ethically.\n\nAs we stand on the cusp of another technological revolution, where AI will influence our cultural, societal, and personal spheres, we must stop and question where we draw the line. Should we allow AI to create its own religions and culture? Do we allow law enforcement to make deductive arrests in the future? \n\nIn conclusion, the intersection of AI and ethics represents a critical frontier in our technological journey. It is important to hold firmly to our ethics and morals, leading us into a future where technology is used to serve humanity rather than being dictated by it. As we continue to push the boundaries of AI, we must remember to ask not if we can BUT if we should.\n","metadata":{}},{"cell_type":"markdown","source":"# References\n\n1. [Privacy on the iPhone](https://www.youtube.com/watch?v=1HWUjMjaBJI)\nis an advertisement that talks about how apple has makes it harder for the apps on the phone to track the users data. This is ironical, as the tracking was enabled by apple in the first place! \n\n2. Business Insider India. \"China has started ranking citizens with a creepy 'social credit' system - here's what you can do wrong, and the embarrassing, demeaning ways they can punish you.\" Business Insider India, 8 Apr. 2018, https://www.businessinsider.in/miscellaneous/china-has-started-ranking-citizens-with-a-creepy-social-credit-system-heres-what-you-can-do-wrong-and-the-embarrassing-demeaning-ways-they-can-punish-you/slidelist/63666443.cms.\n\n3. Kaffee, Lucie-Aimée, et al. \"Thorny Roses: Investigating the Dual Use Dilemma in Natural Language Processing.\" [ArXiv, vol. abs/2304.08315](https://arxiv.org/abs/2304.08315), 2023.\n\n4. Bucknall, Benjamin S., and Shiri Dori-Hacohen. \"Current and Near-Term AI as a Potential Existential Risk Factor.\" arXiv preprint [arXiv:2209.10604, 2022](https://arxiv.org/abs/2209.10604).\n\n5. Bubeck, S´ebastien, et al. \"Sparks of Artificial General Intelligence: Early Experiments with GPT-4.\" Microsoft Research, 13 Apr. 2023, [arXiv:2303.12712v5](www.arxiv.org/abs/2303.12712v5) cs.CL.\n\n6. Future of Life Institute. \"Pause Giant AI Experiments: An Open Letter.\" Future of Life Institute, 2023, https://futureoflife.org/open-letter/pause-giant-ai-experiments/.\n\n7. Pan, Alexander, et al. \"Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark.\" Preprint, [arXiv 2304.03279, 2023](https://arxiv.org/abs/2304.03279).\n\n8. Soice, Emily H., et al. \"Can Large Language Models Democratize Access to Dual-use Biotechnology?\" arXiv.org, [arXiv:2306.03809](https://arxiv.org/abs/2306.03809)\n\n9. Sison, Alejo Jose G., et al. \"ChatGPT: More than a 'Weapon of Mass Deception' Ethical challenges and responses from the Human-Centered Artificial Intelligence (HCAI) perspective.\" arXiv preprint [arXiv:2304.11215](https://arxiv.org/abs/2304.11215), 2023.\n\n10. Blanco-González, Alexandre, et al. \"The Role of AI in Drug Discovery: Challenges, Opportunities, and Strategies.\" [arXiv:2212.08104, 2023](https://arxiv.org/pdf/2212.08104). \n\n11. Weidinger, Laura, et al. \"Ethical and Social Risks of Harm from Language Models.\" arXiv, 2021, [arXiv:2112.04359](https://arxiv.org/abs/2112.04359).\n\n11. Goldman Sachs Research. \"Generative AI Could Raise Global GDP by 7%.\" Goldman Sachs Intelligence, 3 July 2023, www.goldmansachs.com/intelligence/pages/generative-ai-could-raise-global-gdp-by-7-percent.html.\n\n12. Pew Research Center. \"Trends in Income and Wealth Inequality.\" Pew Research Center, 9 January 2020, www.pewresearch.org/social-trends/2020/01/09/trends-in-income-and-wealth-inequality/. Web.\n\n13. Bond, Shannon. \"Fake Viral Images of an Explosion at the Pentagon were Probably Created by AI.\" NPR, 22 May 2023, www.npr.org/2023/05/22/1177590231/fake-viral-images-of-an-explosion-at-the-pentagon-were-probably-created-by-ai. Web.\n\n14. Kirchgaessner, Stephanie, et al. “Revealed: the Hacking and Disinformation Team Meddling in Elections.” The Guardian, Guardian News and Media, 15 Feb. 2023, www.theguardian.com/world/2023/feb/15/revealed-disinformation-team-jorge-claim-meddling-elections-tal-hanan.\n\n15. “AI Defamation and Libel Laws”: Ray, Dan. “AI Defamation and Libel Laws.” Nolo, 2021, https://www.nolo.com/legal-encyclopedia/artificial-intelligence-defamation-and-libel-is-anyone-liable.html.\n\n16. Edelman, Gilad. “These Ex-Journalists Are Using AI to Catch Online Defamation.” Wired, 20 May 2021, https://www.wired.com/story/ex-journalists-using-ai-catch-online-defamation/.\n\n17. National Freedom of Information Coalition. \"The Future of Policing Using Pre-Crime Technology.\" Nfoic, 16 August 2018, www.nfoic.org/blogs/future-policing-using-pre-crime-technology/. Web.\n\n18. Chicago Police Surveillance Project. \"Social Media Monitoring.\" Chicago Police Surveillance Project, n.d., https://chicagopolicesurveillance.com/tactics/social-media-monitoring.html . Web.\n\n19. Georgetown Security Studies Review. “Racism is Systemic in Artificial Intelligence Systems, Too.” Georgetown Security Studies Review, 6 May 2021, https://georgetownsecuritystudiesreview.org/2021/05/06/racism-is-systemic-in-artificial-intelligence-systems-too/.\n\n20. Hai, T. How Flawed Data Aggravates Inequality Credit. Stanford University Institute for Human-Centered Artificial Intelligence, 2022, https://hai.stanford.edu/news/how-flawed-data-aggravates-inequality-credit.\n\n21. “Institute for Technology, Ethics, and Culture.” Markkula Center for Applied Ethics, Santa Clara University, 2023, https://www.scu.edu/institute-for-technology-ethics-and-culture/\n\n22. Alessa, Abeer M., and Al-Khalifa, Hend S. \"Towards Designing a ChatGPT Conversational Companion for Elderly People.\" [arXiv, 2304.09866](https://arxiv.org/abs/2304.09866), 2023.\n\n23. One Mind PsyberGuide. \"Woebot.\" One Mind PsyberGuide, 2023, onemindpsyberguide.org/apps/woebot/. Web.\n\n\n#### OTHERS\nMIT Technology Review. \"Concerning AI Ethics.\" In Machines We Trust, released 12 April, Year of Release (if available), Spotify, https://open.spotify.com/episode/0fSf6x6v5wNvt3ZakkuJzx.\n","metadata":{}}]}