Written by: Techub News Organization

Introduction

In July 2025, the "Godfather of AI," Turing Award winner Geoffrey Hinton, rarefully articulated his deep concerns about the recent risks of artificial intelligence in an interview with "The Diary Of A CEO." This pioneer of deep learning is highly praised for his groundbreaking work, but in recent years, he has frequently warned of the potential dangers of AI. In this interview, Hinton vividly outlined the five major impacts that AI technology could have on society by 2027 without effective regulation, using clear and urgent language. His views are not only based on technological insights but also touch upon the deep contradictions in capitalism, regulatory dilemmas, and human nature.

Summary

• Cybersecurity Risks: AI-driven cyberattacks could cripple key institutions such as banks, prompting Hinton to take measures like diversifying his assets for self-protection.
• Bioweapon Risks: AI significantly lowers the threshold and cost for designing and manufacturing new deadly viruses, making it possible for individuals or small groups to acquire large-scale lethal capabilities.
• Election Interference Risks: AI can generate highly personalized political advertisements, precisely manipulating voters, while centralized data provides the "fuel" for such manipulation.
• Social Division Risks: Social media recommendation algorithms continuously push extreme content for profit maximization, deepening information silos and tearing apart social consensus.
• Decreased War Risks: Lethal autonomous weapons will greatly lower the "friction" and domestic political costs for major countries to start wars, potentially leading to more frequent acts of aggression.

1. From Cyberattacks to Biological Warfare: Imminent Survival Threats

Hinton's worries initially focus on specific and potentially imminent survival threats, with cybersecurity and bioweapons placed at the forefront of high risks.

Regarding cybersecurity, Hinton exhibits an almost "preparedness" mentality. He revealed that he has "totally changed" his behavior out of fear of cyberattacks. He expresses particular concern that well-regulated financial institutions like Canada's banks could also fall victim to AI-enhanced cyberattacks. He explains that while customer assets theoretically remain the customer's when a bank goes bankrupt, if attackers can manipulate and sell off customer stocks held by the bank, individual wealth could still suffer catastrophic losses. To this end, he has diversified his and his children's funds across three different banks, hoping that when one bank is attacked, the others can quickly strengthen their defenses. He even uses a non-networked phone and offline hard drives to back up important data in preparation for an "entire internet outage."

Even more chilling is AI's potential in creating bioweapons. Hinton points out that AI makes the design of new viruses "relatively cheap," no longer requiring highly skilled molecular biologists. He paints a terrifying scenario: a resentful "madman" with a bit of knowledge in molecular biology and expertise in AI, or a small cult capable of raising millions of dollars, could design a batch of viruses. State actors may also secretly engage in such research. Although nations would be concerned about retaliation and the virus disrupting their own populations, the risks are already present. Hinton emphasizes that what makes this threat especially terrifying is that it renders the acquisition of large-scale destructive capabilities unprecedentedly "democratized."

2. Ripping Society Apart: Manipulated Elections and the Erosion of Consensus

Hinton believes that AI’s erosion of the operational foundation of democratic society is happening quietly, with its core being the manipulation of elections and the destruction of social consensus.

He proposes that a "very effective" means of election corruption using AI is based on massive personal data for targeted political advertising. Once voters' income, addresses, and all sorts of information are known, AI can generate highly persuasive messages, such as persuading them not to vote. Hinton boldly points to Elon Musk's business activities in the U.S. He notes that Musk is pushing to obtain data that has been strictly isolated, claiming it's to improve efficiency, but this "is precisely what you need if you want to corrupt the next election." Hinton deduces that one motivation for acquiring this government-sourced data might be to manipulate elections, and another motivation could be that this data is excellent material for training large models.

Deeper social divides stem from the profit-driven mechanisms of social media platforms. Hinton analyzes that platforms like YouTube and Facebook have algorithms determining content to display that essentially means "showing anything that can get you to click." What gets the most clicks are often the extreme content that triggers "indignation" and confirms one’s biases. This mechanism pushes users further into deeper "echo chambers," making their understanding increasingly distant from that of other groups. "We no longer share a common reality," Hinton sighs. He shares his experience: three-quarters of the content in his iPhone news feed is about AI, making it difficult for him to determine whether the world is talking about AI or it's merely his information silo. This algorithm-driven personalization is making "my reality increasingly distant from yours."

Hinton attributes the root cause to the profit motive under capitalism. Companies are legally obligated to maximize profits, and when harmful societal acts (such as promoting extreme content) are necessary for profit, regulation must intervene. However, regulation faces a dual dilemma: first, politicians may be "bought off" by large companies; second, politicians may not understand technology at all, as he cites the embarrassing instance of the U.S. Secretary of Education calling AI "A1." Ultimately, tech companies likely control the situation because they are smarter. Hinton mentions that U.S. tech firms are advertising that they should "not regulate AI, or they will get hurt in competition with China." He sees this as a seemingly reasonable argument, but society must decide: should we win the competition at the expense of our own society?

3. Lowering the Threshold for War: Autonomous Weapons and Runaway Superintelligence

Extending from realistic threats to ultimate fears, Hinton discusses the catastrophic prospects that lethal autonomous weapons and superintelligent AI may bring.

Lethal autonomous weapons refer to machines that can decide to kill on their own. Hinton believes that its worst impact is making it easier for major countries to invade smaller ones. Using soldiers results in casualties, and returning body bags can provoke domestic protests (like during the Vietnam War). But if robots are deployed, domestic protests are likely to be far fewer, and the military-industrial complex would prefer this, as robots are expensive and profitable. This "lowers the cost of invasion" and "reduces the friction of war," potentially leading to more frequent acts of aggression. Hinton shares a chilling experience: a friend’s drone, costing less than £200, could not only closely observe him but also track him continuously in the woods. He asserts that all major defense departments are busy producing such weapons.

When these risks are compounded, they create even more "combinatorial" risks. Ultimately, this points to the risks of superintelligent AI. Hinton's core idea is: facing superintelligence that is smarter than humans, we cannot prevent it from eliminating us (if it wants to); the only way is to ensure that it never wants to do that. He uses a vivid metaphor to explain this asymmetric relationship: to understand what it feels like not to be the top intelligent species, ask a chicken; the intelligence gap between him and his French bulldog, Pablo, is a microcosm of the future gap between humans and super AI. He further likens it to raising a little tiger; you must ensure that it never wants to kill you when it grows up; otherwise, you could be dead in seconds. "Our current AI is that little tiger."

He mentioned a positive example: mothers and infants. Evolution has made mothers smarter than infants, but infants can "control" mothers through crying and other means. We need to find ways to make superintelligence "not want to take over us." However, Hinton candidly stated, "We don't know if we can make them not want to take over, not want to harm us... I think it may be hopeless." But he pivoted, believing that if extinction occurs because humanity is too lazy to try, that would be madness. When asked whether this affected his view of his life’s work, he admitted, "It indeed diminishes the sense of accomplishment," feeling a bit sad, as AI should create immense welfare in fields like healthcare and education. He feels a responsibility now to discuss these risks.

4. Industry Insights: The Departure of Security Pioneers and the Duality of Leaders

Drawing from industry events in July 2025, Hinton shared some observations about AI safety advocates and business leaders.

When discussing the departure of OpenAI co-founder and chief scientist Ilya Sutskever, Hinton stated that they still occasionally have lunch together. Although he has no insider information, based on his understanding of Ilya—"he has good moral principles" and is genuinely concerned about safety—he believes this is indeed the reason for Ilya's departure. As a core figure in the development of early versions like GPT-2, Ilya’s departure itself sends a signal. In contrast, Hinton described Elon Musk as "lacking moral principles." Regarding Sam Altman, Hinton stated he does not know him personally but noted the changes in Altman’s remarks: a few years ago, he casually mentioned that AI could kill everyone, while now he states there is no need to worry excessively. Hinton suspects that this shift is driven not by a pursuit of truth, but by a pursuit of money (or power).

Hinton relayed an unsettling private remark from a billionaire friend about the leader of one of the world's largest AI companies. This friend is closely involved with circles of these AI builders, and he revealed that this leader (Hinton hinted at one of three giants) privately believes we are heading towards a dystopian world with excessive leisure and no need to work, and this person "doesn't care at all" about the harm it will cause to the world. After watching this person's public interview, Hinton was shocked to find that "he was publicly lying." Hinton reflected that part of what drives these figures is a "sadistic tendency" for power; they enjoy the idea that they can fundamentally change the world. He believes Musk obviously has this trait and describes him as a "complicated character," who has both pushed for good things like electric vehicles and done some "very bad things."

At the end of the interview, Hinton reiterated his core appeal: if superintelligence could lead to human extinction, he hopes that by warning of this possibility, people can urge governments to focus on researching how to control this technology. Governments must force companies to allocate resources to safety research, whereas companies currently invest insufficiently in this area because it does not yield direct profits. In Hinton's view, facing the growing AI "little tiger," humanity's ability to coordinate and regulate is facing an unprecedented test.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。