The REAL Reason People Are Scared of AI

Intro (0s)

• A friend mentioned that AI has the potential to destroy society and potentially end humanity, which raises concerns about the dangers of artificial intelligence to the social order (10s).
• There have been warnings about the dangers of AI, but they are often discussed in vague terms, such as a threat to democracy, losing control, and a danger to society (37s).
• To understand the dangers of AI, it's essential to look at the new laws and regulations being proposed, which provide insight into what lawmakers and regulators are worried about (52s).
• AI is a technology that uses computers to do tasks that human brains can't do well, and it has been developed over decades through the use of code and specific instructions (1m10s).
• The difference now is that AI software can teach itself how to do tasks, using tons of data from the world, and the goal is to create artificial intelligence that can predict accurate information and solve problems (1m34s).
• The potential of AI to change everything we do is what makes it dangerous, and it's essential to understand what is meant by "danger to humanity" and how AI could negatively affect humans (2m6s).
• To explore the dangers of AI, six scenarios will be examined, looking at how AI could negatively affect humans and what can be done to prevent it (2m19s).
• The concept of AI can be represented by a "black box" where humans provide data and instructions, and the AI figures it out by itself, which is where the potential promise and peril of this new technology lie (2m31s).

Predictive Policing (2m52s)

• The use of technology in policing should be done in a way that respects the principle of "innocent until proven guilty," rather than relying on predictive methods that may lead to wrongful accusations (2m54s).
• Car Artigas, an expert in machine learning and former Secretary of State of artificial intelligence in Spain, emphasizes the need for global governance of AI to understand its effects on society (3m4s).
• AI can be used to solve problems by analyzing large amounts of data, and its accuracy increases with the amount of data it is trained on (3m25s).
• Applying this approach to crime prediction, a police department could use data such as biometric information, location, and behavior to predict who may commit a crime (3m55s).
• However, this raises concerns about mass surveillance, infringement of privacy rights, and wrongful identification (4m48s).
• A recent example in Detroit, USA, illustrates the risk of wrongful identification, where an AI algorithm incorrectly matched a suspect's face to a driver's license record, leading to a wrongful arrest (5m6s).
• The use of AI in policing may lead to a "nightmare scenario" where police departments collect and track excessive data in the name of preventing crime (5m30s).
• The EU's new AI bill aims to address these concerns by emphasizing that people should be judged based on their actual behavior, rather than predictive models (5m49s).

Elections (6m2s)

• Experts are worried that AI will affect elections and democracy by eroding trust in the system, as a huge part of elections and democracy relies on trust in the system itself and the information received about candidates and election results (6m4s).
• One of the nightmare scenarios with AI is the use of deep fakes, which are becoming easier to make and can be used to spread misinformation, although humans are currently good at deciphering these fakes due to evolution training their brains to be discerning of human faces (6m41s).
• However, deep fakes and synthetic media are expected to get better quickly, and can be used to sway elections or spread misinformation, such as fake videos of politicians or leaders saying something they didn't say, or fake news about polling stations being taken over by a militia (7m5s).
• In the future, people may stop believing in the democratic system if they are exposed to too much misinformation, and this is a scarier result than the manipulation of images, as it can lead to a lack of trust in institutions (8m5s).
• To address this issue, lawmakers in California are requiring online platforms to find and label synthetic media, or take it down, and some bills even prohibit people from posting election-related content that has been generated or modified using AI (8m18s).
• The AI bill in Europe requires anyone who makes deep fakes or synthetic media to code in an invisible watermark that can be detected by software, and the AI act makes it compulsory by law to identify if something has been generated by a human or by AI (8m36s).

Social Scoring (9m12s)

• Social scoring is a method used by governments to control populations and can lead to unfair treatment or discrimination, where a person's behavior is tracked and tabulated to create a score that determines access to public services and housing loans (9m18s).
• A similar system already exists in the United States in the form of credit scoring, where corporations collect data about individuals and use an algorithm to assign a score that affects their ability to get loans, housing, jobs, and insurance (10m6s).
• This credit scoring system is discriminatory against certain groups and is normalized in the US, despite being invasive (10m40s).
• A more invasive social scoring system could be implemented, where employers can buy data and track employees' behavior on the job to evaluate their fitness for promotion or hiring (10m47s).
• AI-powered admission systems could also analyze applicants' data, including photos, essays, and social media handles, to decide who gets into university, but these systems can be biased and create discrimination against certain groups (11m19s).
• The Chinese government has a social credit system that assigns a score between 600 and 1300, determining access to schools, travel, jobs, and other services, with punishments for low-scoring individuals and perks for high-scoring ones (12m16s).
• China's social credit scoring system is not centralized and varies by region, but it is expected to become more robust with the development of more powerful artificial intelligence (12m55s).
• Lawmakers in Europe and the United States are concerned about the use of AI for social scoring, which is considered an unacceptable risk, and is prohibited in Europe as it involves ranking or classifying people based on their behavior in society (12m59s).
• There are numerous developments and stories about AI emerging every week, including a recent Nobel Prize awarded to machine learning researchers, with some likening it to the invention of penicillin, while others emphasize the risks of AI (13m29s).
• Ground News is a useful tool for navigating the various speculative takes on AI, as it aggregates news sources from around the world, analyzes and sorts news stories, and provides information on the factuality and ownership of news outlets (13m50s).
• Ground News is offering a 50% discount on their Vantage plan during the holiday season, and can be accessed by clicking the link or scanning a QR code (13m55s).
• The tool is particularly useful for coverage on AI, allowing users to have a well-rounded perspective on the topic, as well as many other news topics (14m26s).

Nuclear Weapons (14m57s)

• The concept of machines taking over, as depicted in sci-fi movies, is often different from the reality of the risks associated with artificial intelligence (AI), which can be more boring but more dangerous, as seen in scenarios like predictive policing and social scoring, but the fear surrounding AI and nuclear weapons is similar to the movie Terminator, where an AI-powered missile defense system becomes self-aware and launches a nuclear assault against humanity (15m20s).
• The real fear is giving machines too much autonomy to make high-stakes decisions about war, including launching nuclear weapons, as AI can synthesize a lot of information to make more accurate decisions than humans, taking into account more data than a human brain can hold at once (15m47s).
• As AI becomes better at reasoning, it will become better than humans at making decisions that get the desired result, which in war is a difficult thing, and in a future where many military systems are run by AI, it will become a bigger part of defense strategy (16m17s).
• A scenario where an AI system is in charge of making real-time decisions and sees an adversary conducting military tests, which it mistakenly interprets as a threat, could lead to a nuclear launch, although this scenario is unlikely (16m55s).
• Lawmakers have moved quickly to address this issue, with a bill floating around the Senate called the Block Nuclear Launch by Autonomous AI Act, which aims to prevent AI systems from launching nuclear weapons, and the US hopes other countries will follow suit (17m38s).
• While nuclear weapons may be off the table for AI systems, other powerful weapon systems are not, and AI is already being used in Ukraine and Israel to provide recommendations on strike targets, making war more frictionless, easier, and less transparent (17m58s).

Critical Sectors (18m32s)

• Critical sectors such as pipelines, water, electricity, transportation, food, and communication systems are essential for human survival, and most people wouldn't be able to stay alive for long if these systems went down (18m33s).
• Artificial intelligence and machine learning algorithms are being used to help run these critical sectors more efficiently, making millions of small decisions every minute, recognizing patterns and problems, and optimizing resources (18m57s).
• AI systems will soon be running water treatment plants, traffic lights, and public transportation, adjusting traffic flows in optimal ways and responding to real-time information (19m12s).
• This increased reliance on AI will make life better by reducing congestion and improving overall transportation efficiency, but it also raises concerns about biases in critical infrastructure management (19m40s).
• AI systems may prioritize the needs of wealthy individuals over vulnerable populations, such as the elderly, sick, or low-income areas, exacerbating existing inequalities (20m32s).
• The use of AI in critical sectors also raises concerns about the "black box" problem, where it is unclear how the AI is making decisions, which can lead to unintended consequences, such as contaminated water or traffic congestion (21m10s).
• Technical issues, such as faulty sensors or software updates, can also cause AI systems to malfunction, leading to serious consequences, such as people getting sick from contaminated water or being stuck in traffic (21m25s).
• The increased reliance on AI in critical sectors highlights the need for transparency and accountability in AI decision-making, as well as the need for human oversight and maintenance to prevent technical issues (21m43s).
• A hypothetical scenario is presented where a city's traffic system, run by a smart AI system, encounters an unexpected bug, causing chaos and disrupting critical services like ambulances and fire trucks, resulting in damage and even death (22m24s).
• The root cause of the bug is not immediately apparent, taking traffic technicians and engineers 5 days to identify and fix the issue, highlighting the risks of relying on complex AI systems for critical infrastructure (22m40s).
• The importance of keeping critical sectors and infrastructure running smoothly is emphasized, as they are crucial for maintaining public safety and preventing harm (22m54s).
• To mitigate these risks, it is suggested that lawmakers and companies using AI in critical systems should be required to "open up the black box" and demonstrate that their AI systems have been trained with representative data sets, are unbiased, and do not discriminate (23m14s).
• Companies using AI in critical systems will need to assess and mitigate risks, ensure robust cybersecurity measures are in place, and demonstrate a commitment to responsibility and safety in order to leverage the benefits of AI (23m42s).
• By taking a cautious and responsible approach to the development and deployment of AI in critical systems, it is possible to minimize risks and maximize benefits (23m58s).

Optimist’s Take (24m12s)

• Advances in AI could dramatically change the world for the better, particularly in the medical field, where AI could run hospitals and medical research, potentially saving lives and discovering new treatments for previously untreatable diseases (24m13s).
• AI systems could also allow for the prediction and preparation of extreme weather events, optimization of water use in agriculture, monitoring of soil health, and prediction of pest outbreaks, which would reduce the need for harmful pesticides and fertilizers (24m16s).
• The development of AI is expected to bring about significant benefits, and with the help of smart people, such as Card, who are working on legislation to establish "guard rails" around the technology, it is possible to develop AI responsibly and mitigate its risks (24m56s).
• The goal is to develop AI in a way that allows society to reap its benefits while minimizing its risks, and with responsible development, a positive future for AI is possible (25m4s).

Credits (25m15s)

• The video ends with music and a mention of a person named Carme, who is referred to as a "cool lady" (25m23s).
• Viewers are encouraged to support the creators on Patreon (25m25s).
• The video concludes with a farewell message, stating that the creators will see their audience in the next video (25m28s).