Title:	AI Geopolitics – Algorithms as Weapons
Authors:	Saalbach, Klaus
Abstract:	Artificial Intelligence (AI) is commonly understood as the ability of machines to perform tasks that normally require human intelligence and is a key area of advanced computing. Analyses of US military experts indicate that an Artificial General Intelligence (AGI), i.e., an AI fully reaching human cognitive level, will be available in a few years. The US and China are currently developing new war concepts with a primary focus on the key vulnerabilities of the military system of the adversary instead of widespread destruction. In US, this is e.g. discussed as non-kinetic or cognitive warfare and in China as Multi-Domain Precision Warfare MDPW. In line with the concept of a non-kinetic cognitive warfare, AI is now projected as a new weapon type where an advanced Artificial Superintelligence (ASI) uses superior tactics and a rapid takeover of computer systems to overwhelm an adversary. While this is not yet official an official doctrine, all sides are aware that the nation who will achieve an AGI first will already have a unique technological, economic, and military advantage. The security dimension of algorithms is acknowledged: In June 2024, the leading AI firm OpenAI has hired the retired US General Paul Nakasone, the former head of the National Security Agency (NSA) and the military Cyber Command (CyberCom) and a few days later, it was decided to block the access to ChatGPT for China from 01 July 2024 on. The design of computer chips is already dependent on computer support with specialized software systems. The involvement of AGIs could accelerate this process to develop an ASI which is far beyond human intelligence and other computer systems. Most authors expect AGI in this decade and ASI in the first half of the 2030ies. An important finding of the ongoing ASI debate is that an ASI will need sophisticated algorithms to beat the adversary with a cycle of data input, analysis, conclusions, and actions, but does not necessarily need ‘consciousness’ which is seen as the most difficult if not even impossible part of AI development. An ASI will amplify the challenges of current AI systems: AI systems lack context knowledge and may decide too quickly. The opacity of the systems leads to explainability or interpretability issues; further problems may result from manipulated input (data poisoning) and errors (hallucinations). As a software system, AI is vulnerable to cyberattacks, although it can also detect and classify cyber threats more quickly through data analysis and pattern recognition. Wargame simulations with current generative AI Large language models (LLMs) showed that the systems tend to escalation up to nuclear strikes, because if the analysis of the communication shows that there is no progress, the AI decides to escalate. Furthermore, currently used generative AI programs use AI deception to achieve their goals as quickly as possible. The expected development of computers by other computers (machine evolution) will hand over control from humans to machines without knowing how the machine works. Already the AGI may lead to an AI explosion which can massively change daily life. Politicians and society are not prepared for the coming changes.
URL:	https://doi.org/10.48693/556 https://osnadocs.ub.uni-osnabrueck.de/handle/ds-2024070211315
Subject Keywords:	AI; Artificial Intelligence; Artificial Superintelligence ASI; Artificial General Intelligence AGI; Cognitive Warfare; Multi-Domain Precision Warfare
Issue Date:	2-Jul-2024
License name:	Attribution 3.0 Germany
License url:	http://creativecommons.org/licenses/by/3.0/de/
Type of publication:	Arbeitspapier [WorkingPaper]
Appears in Collections:	FB01 - Hochschulschriften

This item is licensed under a Creative Commons License