A founder of the Internet raises concerns about deepfakes, disinformation

Key Concepts

• Internet's Genesis: The foundational development of the internet, including the first network connection and the mathematical theory behind data transmission.
• Commercialization of the Internet: The transition of the internet from an academic/research network to a commercial platform.
• Generative AI: Artificial intelligence capable of creating new content, including text, images, and potentially more.
• Large Language Models (LLMs): A type of generative AI that processes and generates human-like text.
• Digital Threats: Modern cybersecurity risks, including deep fakes and disinformation.
• Disinformation: False or inaccurate information, especially that which is deliberately intended to deceive.
• Deep Fakes: Synthetic media where a person in an existing image or video is replaced with someone else's likeness.
• Regulation of AI: The need for governmental oversight and rules to govern the development and deployment of AI technologies.
• Data Protection, Privacy, and Intellectual Property: Existing regulatory frameworks that may need to be adapted for generative AI.

The Evolution of the Internet and Rise of Generative AI

Leonard Kleinrock, a pioneering figure in the development of the internet, has voiced significant concerns regarding contemporary digital threats, specifically deep fakes and disinformation. His insights were shared during a conversation with CN Rani Samani in Singapore, where they discussed the internet's evolution and the burgeoning field of generative AI.

The Internet's Early Days and Development

• Foundational Network: The initial backbone of the internet was established with a connection from UCLA to SRRI.
• Mathematical Theory: Kleinrock, a founding architect of the modern internet, developed the mathematical theory that underpins data transmission between computers.
• First Message: He sent the very first message between two devices in 1969 on a network that served as the precursor to the internet.
• Timeline of Development: The internet began in 1969, with its commercialization not occurring until the mid-1990s. This provided a 25-year period for the network to be curated, perfected, and for issues to be addressed.

The Rapid Ascent of Generative AI

• Accelerated Growth: In contrast to the internet's gradual development, the recent rise of AI has been exceptionally rapid, with services exploding both positively and negatively without sufficient time for processing, maturation, and curation.
• Optimism and Advantages of LLMs: There is optimism surrounding the potential of AI and its advantages. Large Language Models (LLMs) can significantly enhance individual capabilities in creation, study, and presentation. They also hold the potential to automate tedious jobs.

Emerging Digital Threats and Concerns

• Increased Sophistication of Threats: The "dark side" of the internet, initially perceived as minor disruptions by hackers, has evolved into a far more serious landscape involving organized crime and nation-states.
• Unforeseen Consequences: Kleinrock expressed surprise at the speed and severity of these negative developments, noting that these were not anticipated when the internet was initially conceived by a group of engineers focused on a challenging technical problem.
• Magnification of Disinformation: Social networks amplify the reach of individuals spreading poor information, turning a single message into a global reach.
• AI's Exacerbation of the Problem: Generative AI systems can now create fabricated content that appears highly realistic, making it even more difficult to discern truth from falsehood. This has significantly worsened the problem of disinformation.

The Need for Regulation and Safeguards

• Resistance to Regulation: Those deploying AI technologies are often resistant to regulation and control, prioritizing profit and shareholder interests.
• Public and Scientific Pushback: Kleinrock suggests that these entities will only yield to regulation if there is significant pushback from the public and the scientific community.
• Regulatory Gap for Generative AI: While most countries have regulations for data protection, data privacy, and intellectual property rights, there is a notable gap concerning generative AI.
• Potential for Harm: Until a global regulatory structure is established, powerful AI systems, particularly if misused, could cause more harm than good.

Key Arguments and Perspectives

• Pace of Innovation vs. Maturation: Kleinrock argues that the internet's development allowed for a period of maturation and issue resolution, a luxury not afforded to the current rapid advancement of generative AI.
• Evolution of Threats: The nature of digital threats has shifted from individual hackers to organized criminal enterprises and state actors, necessitating a more robust response.
• AI as an Amplifier of Harm: Generative AI, while offering benefits, also possesses the capability to create and disseminate highly convincing disinformation, amplifying existing societal problems.
• The Imperative of Regulation: The lack of specific regulatory frameworks for generative AI poses a significant risk, highlighting the need for global cooperation to establish safeguards.

Notable Quotes

• "We used to imagine the dark side was just a bunch of hackers making trouble. Fine, let them do it. But it's gotten much more serious these days. It's now organized crime. It's a nation states and so the surprising thing is how quickly and badly it turned and how much we did not anticipate that when we started this net we were bunch of nerds basically trying to succeed in a challenging engineering problem to connect computers around the world." - Leonard Kleinrock
• "The people who are deploying this thing don't want to be regulated. They don't want to be controlled. They want to basically exploit the things from AI to further their own goals which is profit and their shareholders. Until there's pushback against them by the public and by the scientists they won't yield." - Leonard Kleinrock

Conclusion

Leonard Kleinrock's reflections underscore a critical juncture in technological development. While acknowledging the immense potential of generative AI, he emphasizes the urgent need for proactive measures to mitigate its inherent risks. The internet's journey from a research project to a global commercial network provided a crucial period for development and refinement. However, the current trajectory of AI, characterized by its rapid deployment and the sophisticated threats it can enable, demands immediate attention. The absence of a comprehensive regulatory framework for generative AI, coupled with the resistance from industry stakeholders, presents a significant challenge. Kleinrock's call for public and scientific pushback highlights the necessity of collective action to ensure that AI's advancement serves humanity's best interests and does not lead to widespread harm.