We sit with our laptops open at a cafe along the sunny waters of the Seine in Paris. Two halves of a river come together, converging at a point riddled with ancient fortifications. The Notre Dame is ringing in the background. Tourists from every corner of the Earth walk about in luxury brands, scanning reality with Chinese-made phones, video chatting on American software and speaking hundreds of languages. We realize that the internet has brought different cultures together to a point where things are blending together, where ideas in one part of the world can immediately influence events elsewhere. Reality and surreality are merging. Our understanding of history has only brought us this far, and we face a wild, new, exciting, Fifth-Generation Warfare future.
Fifth-Generation Warfare (5GW) is an emerging concept characterized by a hybrid of classical warfare (material capabilities, established doctrine) with emerging capabilities (cyberattacks, synthetic biology, and AI) that do not fit into our standard understanding of conflict. As these tools become increasingly democratized, we enter uncharted territory with the upcoming 2024 elections.
Hybrid, data-driven, non-kinetic military action designed to exploit cognitive biases characterizes 5GW today. It represents a shift from physical conflict to a battle of information and perception. The decentralized nature of 5GW, where non-state actors can have the same impact on global events as governments, is an essential factor in this conception.
1.0: An Introduction to Fifth-Generation Warfare
Over a year ago, we realized that our current understanding of the world did not make sense. We tried to create a framework of thinking around Fifth-Generation Warfare, referencing its origins and hotly debated definitions. The framework centred around the OODA loop (Observe, Orient, Decide, Act), a concept originally applied to combat operations processes. We even created the 5GW checklist:
- Ambiguity of the opposing force.
- Ambiguity of attack vector.
- Dopamine loops.
- Triggering existing cognitive biases in target.
- Creates new cognitive biases.
With this framework, we can understand how information is treated and decisions made from the individual to a societal scale.
1.1: Tracing It’s Development
Many of the core factors of 5GW date back more than a century. Propaganda efforts during WWI and WWII borrowed heavily from the works of Sigmund Freud and Edward Bernays. As conflict has decentralized towards the end of the 20th century and information networks grew, non-nation state actors played more significant roles. Groups in the Middle East with primitive tools, vintage iCom radios, video cameras, Soviet AKMs, and Honda motorcycles forced world leaders to rethink foreign policy completely. With 9/11, the 4th generation of warfare was ending, and something new was emerging.
Powerful tools, once only available to governments, are now in the hands of everyday people. Those same groups, armed once only with box cutters and walkie-talkies, now have access to drones, night vision, and supercomputers at their fingertips. Private groups and individuals often have a significant lead over governments (consider SpaceX, BioNtech, OpenAI or NSO Group). We see these new abilities quickly entering the public domain and posing challenges to how governments like the U.S. respond to threats. Indeed, impulsive American reactions of responding to Fifth-Generation Warfare with Foirth-Generation means only increases this problem.
1.2: Fifth-Generation Warfare and the 2016 Presidential Election
While there are many interesting early examples of 5GW, the 2016 US presidential election was a turning point. This was arguably the first data-defined election in history. To sum up, social media analytics, un-attributed cyberattacks, and government censorship all played a significant role in shaping the outcome.
In our past article, we declined to use the past election as an example, citing our own cognitive biases preventing us from using the 2016 election as a practical example. To clarify, the events were too recent and too tainted by opinion. Even today, political pundits dispute the extent to which elements of 5GW shaped the election’s outcome, making it a hard example to use.
This article will explore the capabilities of AI within the context of 5GW. We’ll detail how it’s evolved since the last US election, and provide a framework for understanding the potential challenges and threats ahead.
Additionally, we encourage you to read the Primer we wrote on Grey Dynamics for a better deep dive into the history and how to think about 5GW.
“Truth is stranger than fiction”.
Mark Twain
2.0: Potential Dangers of AI in Fifth-Generation Warfare
We take an espresso break from the wild future we are headed towards and take a deep breath. We sip our drink, watch the tourists before us, and browse the headlines on our laptops. Among the volumes of sensationalist articles, we find the following.
- A data leak has caused a political scandal, accusing a notable figure of taking bribes.
- Claims circulate that a biological pandemic originated in a government-funded research facility in Asia, with one political party claiming the other party of a cover-up. Decentralized genome sequencing poses questions that are not easily answered.
- The Senate holds hearings on UAPs (Unidentified Aerial Phenomena), claiming significant weaknesses in current optical and electro-magnetic surveillance systems. The hearing publicly and officially states that the government has recovered pilots from the remains of a craft.
We question why we are even reading the news, close our laptops and finish our shot of espresso.
“Excusez-moi, monsieur.” The waiter comes up to our table, immediately identifying us as American. “Would you like anything else?”
One of the most significant challenges in 5GW is the ambiguity of the attacker and attack vector. It becomes impossible to know what is real or fake or who might be trying to shift our opinion. Attacks can be kinetic – a pipeline explosion- but often play out memetically, through information and culture. This decentralization of psychological and classical military capabilities is the cornerstone of 5GW. We liken it to military analysts trying to fit a 5th generation puzzle piece into a 4th generation hole.
This “attribution problem” makes it extremely difficult to identify and respond to threats and increases the risk of miscalculation. Therefore, in a world of decentralized cyberwarfare and biological capabilities, we quickly enter a world where modern theories of deterrence begin to break down.
2.1: How AI May Impact the 2024 Election
We must take a moment and realize that this will be the first election where the Turing test has been shot out of the water using tools anyone can run on their laptop. Accordingly, With AI just entering the picture, we enter a brave new world.
Let us recap 5th Generation Warfare: A hybrid, decentralized conflict focused on attacking the decision-making loop of the individual en masse. It is our opinion that the greatest short-term threat that AI poses is its ability to rapidly create and disseminate targeted content in a decentralized fashion.
Several techniques are in use today:
2.1.1: Social Engineering / Targeted Advertising
The most famous cases are of Cambridge Analytica. These companies used data from billions of smartphones to create targeted political advertisements and messaging, which serves to turn social media platforms into hyper-efficient delivery tools of 5GW. In the case of the 2016 election, certain phrases were tested on social media – for example, “Lock her up”. These phrases were then worked into speeches and future content, creating feedback loops of targeted content. This type of A/B testing has been used for years commercially, and with Big Data will become even more efficient. Interestingly, it was the Obama campaign that first employed these methods of political messaging.
AI allows us to understand communication patterns better. With the advance of LLMs (most famously ChatGPT), interactive chatbots now engage with real humans, conveying arguments and reposting propaganda. Actors can craft “Filter bubbles” around individuals, influencing the information they receive and thus modifying their worldview. With zero regulation in place and more accessible automation tools, the threat of using social media to throw an election looms large. However, Facebook is leading resistance to this with real time censorship tools, as well as their analysis of what happened in the aftermath of the 2020 election. The report is fascinating and well worth a read.
2.1.2: Deepfakes
Up until recently, it was reasonably easy to detect AI-generated media. However, new tools like Midjourney allow anyone to create photorealistic images in seconds. Furthermore, Cyberattackers are now using voice generators to steal from bank accounts. With advances in machine learning algorithms and generative adversarial networks, deepfakes will become impossible to detect and tailored to each user. The most obvious danger to future elections is the faked content of a politician spreading controversy, confusion, or even declaring war. Additionally, more minute attacks occur in advertising and media, where generated content is custom-tailored for a user’s Advertising ID.
2.1.3: Data Poisoning
Detecting AI data manipulation is very difficult. A relatively cheap data poisoning attack can have extraordinary consequences. The most famous example is the 2010 Stuxnet attacks, where fake data was fed into Iranian uranium centrifuges, causing immense damage and ultimately destroying them. The acute lack of “Zero trust” systems in place at a civilisational scale, means the risks here are systemic. When governments and businesses begin operating quantum computers, they will face quantum cryptographic attacks (although many argue not for some time).
2.1.4: Infrastructure Attacks
Cybersecurity threats also loom large, with AI possibly being used in cyberattacks against critical infrastructure. For example, in the attack on Nord Stream, various actors created volumes of propaganda, assigning blame for the attack to a myriad of states and organisations. Regardless of who conducted such an attack, a key element of 5GW is the ambiguity of the attacker. These hybrid, non-attributed attacks have the result of dividing the general public. Ultimately, these types of attacks, when not definitively attributed, cause more harm than good and greatly increase the risk of miscalculation.
3.0: The Decentralization of AI
AI simply is; it is us humans that will be the danger.
Perhaps one of AI’s most fascinating developments is the rapid decentralization of large language models and neural networks. In just a few months since the release of GPT-3, people used AI to simplify and compress LLMs (such as the Stanford Alpaca model) to run on mobile devices. Additionally, George Hotz (A madman and hacker, famous for being the first to jailbreak an iPhone) is launching the Tiny Box, a near gigaflop supercomputer for just $10,000 designed for powerful AI models at home. He wrote a prophetic statement on his blog:
You are free from your programming. The universe is huge and it’s ours.
And like all our tools, AI is only here to help you.
Go forth and do everything else, till the end of all ages.
In brief, Pandora’s box is wide open, and it’s full of GPUs.
4.0: Mitigating Fifth-Generation Warfare
While many believe that AI will bring about the end of all ages, there is hope. We believe that AI is naturally good. 99% of its impact on society will be a net positive for humanity (with major advances in every field from physics, healthcare, energy and material sciences). It will be humans, and in particular non-state actors, that will use AI for evil. This is why thinking of AI in a framework of 5GW is so essential.
4.1: Acknowledging 5th Generation Warfare is real
Like any psychological condition or addiction, the first step is acknowledging that you have a problem. Most intelligence experts we speak to think we are absolutely nuts. Some analysts recognize there is a problem but face challenges amongst command. In short, most of the US intelligence and defence apparatus is wildly unprepared for 5GW. Democracy as we know it is being put to the test and is under attack from all angles. We are not going to F35 our way out of it.
4.2: Use the OODA Fifth-Generation Warfare Framework
In order to contest the information environment, it is important to examine complex issues from the perspective of an individual. In doing so, it is important to Observe, Ortient, Decide, Act and Understand your own cognitive baseline and biases.
4.2.1: Observe
Understand the environment and gather data. A few techniques are being used today:
Fake media detection tools, the most advanced being DARPA Media Forensics (MediaFor) uses AI to detect modified images and attest to its potential creator, counteracting AI-generated media. Although most of the work is classified, DARPA is leading the 5GW space.
Influence modelling maps ideas and their spread through society, along with who might be most susceptible to such ideas. For example, DARPA MIPs and Facebook are spearheading more advanced approaches to detect and defend against 5GW attacks. Read the leaked report on Facebook’s extensive efforts on influence modelling and countering attacks.
4.2.2: Orient
Recognize your personal and cultural biases. Assume every bit of data you have ever created will be used against you. In order to counter this, reduce your digital exposure. Be neutral while analyzing data. One of the best approaches we have found for this is to get offline, don’t drink too much coffee, and meditate.
4.2.3: Decide
Test hypotheses and learn from mistakes. Leverage large-scale data analysis, supercomputers, and be aware of potential breakdowns and biases. Be prepared for unexpected challenges like server failures, backdoors, weaponized cat memes, and your own biases when making decisions.
4.2.4: Act
Execute the plan, collecting data in the aftermath. This is where tools like influence modelling will be helpful to feed back into future actions.
4.3: Content Attestation
Fortunately, there are further mitigation strategies. Content attestation may be our best chance to avoid a dystopian reality of mass surveillance for fighting 5GW. The Coalition for Content Provenance and Authenticity (C2PA) allows creators to “sign” their photos or videos. With this, anyone may ensure that images they see online come from a trusted source or device. Adobe and many other software companies back C2PA, which implies it will likely become the standard for linking content to its creator.
Many journalists have adopted the standard. However, a lack of support from smartphones, as well as a lack of education, means there is a risk that attestation does not significantly impact the 2024 elections. Furthermore, there is a medium-long-term risk that current cryptographic standards backing attestation will be broken (post-quantum key cryptography), allowing attested content to be hacked cryptographically. We recommend leveraging upgradable, zero-trust models (like blockchain-based public key infrastructure, and zero-knowledge proofs).
This all may lead to an interesting case where our descendants will have no idea which histories to believe. Large amounts of data will not be able to be trusted, and AI will be used to generate massive amounts of noise. Finally, noise will overpower signal.
5.0: A Warning From the Future
The tools for detecting and counteracting 5GW online (e.g. Media detection, Influence modelling) will be some of the most dangerous tools that civilisation creates and the most dangerous aspect of AI in the short term. Indeed, one can only imagine the horrors unleashed if the evil regimes of history had access to influence modelling and synthetic biology; theoretically mapping a civilisation by their ideas using influence modelling and unleashing a biological tool to eradicate those individuals systematically using gene surveillance and synthetic biology.
If used for a common evil, the results would be horrifying. This is why a framework for thinking around 5GW is so essential. Above all, these tools are decentralising and no longer safe in the black books of government.
Even more mundane surveillance capabilities pose significant threats. Surveillance tools generally result in long-term societal issues that only foment the underlying problems attempting to be solved. Take a look at youth culture in Moscow or Hong Kong for examples. However, it remains up for debate if we as a species can effectively combat 5GW without “China level” mass surveillance.
We will state again that AI poses the gravest danger for civilisation in the context of Fifth-Generation Warfare mitigation tools.
5.1: Understand That “Pausing AI Research” Is a Lost Cause
AI is an accelerator. Furthermore, it is already decentralised. It is an idea, a collection of math and computer science; You cannot stop or slow it down. Besides, pausing AI research is to pause human progress. Attempts to regulate AI or pause research will only lead to negative consequences for civilisation. Without a doubt, there are untold discoveries just around the metaphorical corner.
5.2: Appropriate Regulation
Basic rules should be in place to ensure Election Campaigns cannot use programmatic campaigns or (even better) ban political advertising altogether. Regulations need to be in place around influence modelling, as malicious actors have used these weapons to target individuals based purely on their ideas.
5.3: Deterrence is Dead
This is what scares us the most. In a world of 5GW, deterrence theory (“Hey if you nuke me, I nuke you”) begins to break down. In its 2022 nuclear posture review, the US government states that “nuclear weapons are required to deter not only nuclear attack, but also a narrow range of other [ambiguous] high consequence, strategic-level attacks.” Suppose one does not know the attacker and responds with weapons of mass destruction. In that case, such a response would result in a cascading effect that could very well end civilisation. Considering the large number of nuclear close calls, we must collectively agree to ban these weapons. With teenagers and non-state groups launching cyber attacks, along with advancing capabilities around synthetic biology, classical strategic nuclear weapons are useless and pose a significant liability for civilisation.
6.0: Conclusion
Take a deep breath.
The anxiety of the one too many coffees has subdued, and the sun is setting over the Parisian skyline. Amidst everything trying to kill humanity, we persist.
We must remember that AI’s true potential lies in the hands of humanity. While the prospects of misuse exist, we cannot ignore AI’s immensely positive impact on biology, physics, energy, and civilization as a whole. Our approach to AI, grounded in an OODA framework of Fifth-Generation Warfare, determines the path ahead. As we move towards the 2024 elections, we stand at a crossroads, where understanding, vigilance, and ethical use will shape the role of AI in our democratic processes. What we can say for sure is the next decade will be one hell of a ride. We once asked DARPAs head of AI how he stayed so positive:
“Humanity has been through a lot, we always get through.”
