In today's highly accessible media landscape, the potential of a drastically changing information environment is both real and concerning. Through the advent of generative artificial intelligence technologies like GPT-3, Dall-E, and deep fakes, it is becoming increasingly difficult to tell what is genuine and what is manufactured. Thanks to Generative Adversarial Networks (GANs), users can produce synthetic media with relative ease.

It is no longer simply a matter of fact-checking but rather understanding how these applications can influence our perception of information that has become readily available. The implications here are so much more than just guessing the correct answer. There may be an entirely new set of dangers for us to consider as we navigate this ever-changing world and what is often computational propaganda.

With this exponentially growing threat on our perception of reality, we must take steps to prevent malicious actors from exploiting this new medium. We must engage in attentive vigilance, establish strong cross-industry collaborations, and invest in research to monitor these new technologies. It is ultimately up to us as humans to ensure that generative AI is harnessed responsibly. This article will take a closer look at these generative technologies, weighing the pros and cons, and considering what future dangers we should anticipate.

Understanding the Basics of Generative AI & How It Is Used to Manipulate Reality

Generative AI is a new form of artificial intelligence that is gaining traction in the digital world. This technology uses algorithms and machines to generate content autonomously, such as text, images, and audio, by learning from existing data sets. Generative AI speeds up data processing and improves outcomes with increased accuracy and precision.

Companies can use this technology for many positive applications, such as automated customer service agents, automated asset creation for marketing campaigns, or even video game design. However, these same tools are also being utilized by bad actors to create false or deceptive information, manipulating public opinion, and distorting our perception of reality.

By providing insight into the applications and impacts of advanced technologies on our current reality, we can remain informed of what is revolutionizing our understanding of communication, manipulation, and general information production.

How Does It Work?

Generative AI works by leveraging machine learning models which feed on data sets (images, videos, etc.), analyzing them using deep neural networks before generating new content based on what it has "learned" from analyzing existing data.

These models are trained to recognize patterns in the data they consume before applying them to new content that appears realistic but is completely fabricated. In other words, this technology can create entirely synthetic media that looks real enough to fool humans into believing it is true with no actual human input or intervention.

How is Generative AI Making It Easier for Threat Actors to Drive Large-Scale Disinformation Campaigns?

As a strong online presence become increasingly crucial for businesses, malicious hackers and disfranchised actors are finding more ways to weaponize the digital space, often driving large-scale disinformation campaigns. Generative AI has made it easier than ever for these actors to warp news, media, and information to create fear, mistrust, and confusion.

Researchers from Harvard's Shorenstein Center maintain that experienced purveyors of manipulated media and malicious operators can use this technology to propagate prejudiced or harmful ideologies. Such actors can deliberately publicize disparities and conflict, enabling potentially divisive and inaccurate ideas to permeate the public consciousness as if they are factual.

This year, Meta, Facebook’s parent company, released its Meta Adversarial Threat Report, which served as an example of this activity being perpetrated at scale. In this report, a network originating from Saint Petersburg was found targeting five African countries with news designed to foster discontent about France's influence on the continent. Meta was able to tie this activity to the notorious Russian Internet Research Agency (IRA) — clearly illustrating how effective generative AI can be in manipulating online conversations for malign influence. That said, using productive AI technology to craft carefully balanced media demonstrates the evolving capacity of threat actors in this space.

How Low-Cost Systems Are Expediting the Evolution of Generative AI

Low-cost systems, such as those involving GPUs to process deep learning algorithms and employing massive data sets for training, has greatly expedited generative AI's evolution. This technology can process patterns from text, video games, voice recordings, and high-quality imagery to generate new data with unprecedented speed and detail. While this development has brought remarkable advances in many industries, particularly in entertainment, it has also negatively influenced the proliferation of disinformation.

Many types of misinformation have been algorithmically crafted with surprising realism, making them difficult for humans to distinguish from their truthful counterparts without proper verification. This increased proficiency of bot-generated content leads to greater dissemination of false information that can damage trust and faith in media sources and other communications platforms.

The potential for algorithmically crafted misinformation to be surprisingly realistic is quite disconcerting, as humans may often be unable to detect it accurately without proper verification. Last year, Tom Cruise's deepfake videos by Chris Umé and Miles Fisher garnered immense popularity on TikTok, achieving tens of millions of views and inspiring the creation of Metaphysic; A company that uses deepfake technology to create impossible advertisements and restores old films. Despite its fantastic potential, these capabilities also convey risk when used irresponsibly. The global community must remain vigilant in verifying information sources so that any content generated by deepfakes can be adequately identified and weeded out.

Open-Source Platforms and the Impact of Generative AI on Disinformation

Open-source platforms have enabled a new wave of democratized access to content and communication channels, with the potential to change how information is compiled and disseminated drastically. This accessibility has been used in many areas, but now, open-source platforms have become integral components of many AI applications.

Generative AI technologies like Natural Language Processing can produce vast amounts of low-effort and low-cost content designed to misinform or deceive readers by taking advantage of these open-source platforms. Notably, GPT-2 software can generate versions of news articles from a summary sentence that can appear convincing yet completely false.

These digitally-generated deceptions have proven difficult for tech companies and governments to identify and counter effectively. Open-source platforms may continue to fuel disinformation advances as technology evolves, making widespread propagation all the more likely.

For example, by utilizing generative AI models such as GPT-3 (the successor to GPT-2) in combination with Stable Diffusion's text-to-image generator, threat actors can now produce convincing images with captions that resemble original journalistic materials that are inaccurate or entirely fabricated.

Additionally, the release of GPT-4 is quickly approaching, bringing a significant milestone for artificial intelligence. With the emergence of new techniques in language processing, such as those used in GTP-3.5 and InstructGPT, it may no longer be necessary to rely on extensive parameters for the most advanced text-generating systems available today. While some have predicted that GPT-4 could require more than 100 trillion parameters - 600x as many as its predecessor - this seems increasingly unlikely as advances in language models continue to make great strides towards ever more extraordinary capabilities. Thus, this debate over the number of parameters likely to be needed for GPT-4 delineates which method is especially suitable for developing competent algorithms in natural language processing.

Large companies are fighting to counter visual misinformation through open source

Adobe's CAI (Content Authenticity Initiative) is setting the tone for mitigating digital misinformation with its announcement of a three-part open-source toolkit being released. Developers have made this powerful technology available for swift implementation, and its promise to counter the prevalence of online visual misinformation is inspiring.

The tools include a JavaScript (Software Development Kit) SDK for constructing platforms to showcase these credentials on browsers, a command line utility, and a Rust SDK that equips developers with the capability to generate desktop apps, and mobile apps, to craft, view, and confirm content particulars embedded in these programs. Adobe's decision to make their technology open-source heralds progress in the battle against digital fraudulence.

Exploring the Pros and Cons of Generative AI Technologies

Generative AI technologies also present certain risks when it comes to cybersecurity. For starters, malicious actors can use generative AI tools to create digital disinformation. This includes convincing fake news stories or deepfakes to spread disinformation across social media platforms and other online spaces.

As the debate around the use of generative AI technologies continues to heat up, it is essential to understand what both sides hold as their central arguments. This section will explore the pros and cons of using these technologies, weighing the potential benefits against the associated risks.

Generative AI technologies are revolutionizing the automation landscape. GPT-3, in particular, offers incredible versatility by allowing users to break down complex tasks into explicit and relatively simple steps. This is attributed to its innovative algorithm, which relies on a 175-billion parameter neural network that is more than one hundred times greater than GPT-2's in size. GPT-3 can create highly complex output, potentially generating natural language indistinguishable from humans, only achievable through years of training.

OpenAI's paper introducing GPT-3 showcased that researchers developed it beyond earlier AI models with utmost success across a broad range of language tests that showcase its extendibility and reliability. While this dynamic technology has opened up endless possibilities for automation pipelines, one concern is that it could contribute to a broader reach and realization of malicious strategies in the hands of bad actors.

One of the main concerns in using Generative AI Technologies is the precision and effectiveness of their outputs, as particularly demonstrated by recent tests on GPT-3.

Recently, researchers at the Center for Security and Emerging Technology (CSET) tested GPT -3's capacity with six content generation skills that are standard disinformation techniques: narrative reiteration, narrative elaboration, narrative manipulation, narrative seeding, narrative wedging, and narrative persuasion. The test results found that GPT-3 fared well regarding more straightforward tasks but met with significant difficulty when tasked with longer narratives. This raises questions about its potential to generate text of high quality.

Another issue that should have been addressed in detail is the need for more logical coherency exemplified in GPT-3 when writing more than a few paragraphs; the linkage between the sections can become unintelligible to readers.

Moreover, GPT-3 sometimes fails to tackle projects accurately due to its training data—which was compiled before major global events, including the current pandemic—and thus may not fully comprehend certain concepts or terms such as COVID.

There is no doubt that Generative AI is a powerful tool, yet these cons must be considered when approaching any project involving it.

The Reality of AI Disruption

AI is the new buzzword, the subject of much talk and speculation. However, there is much fearmongering surrounding AI technology, especially with future predictions that seem extreme. It's essential to understand how this technology is being used and its potential implications for the near-term future.

What We Need to be Aware of

The potential reality of AI disruption raises a host of ethical and industry-specific questions. One such example is that a former engineer working for Google's Responsible AI organization went public believing that the company's chatbot was sentient. As AI capabilities ismprove, they may become powerful tools for manipulating and deceiving people.

As algorithms become better than humans at specific tasks, there could be danger associated with blanket acceptance of all decisions made by AI, which renders it possible for malicious actors to introduce false information or produce content without any accountability. Though not all possible use cases may lead to dystopian scenarios, it is crucial to prepare and strategize to protect ourselves against potential malicious use of the technology in the future.

Arms Race

AI has been thrust into an arms race between governments, corporations, and individuals. This is particularly true for China, which intends to become the world leader in AI by 2030. It's also true for businesses trying to outpace each other with new technology and applications to gain competitive advantages. These attempts to get ahead have led to warnings from prominent figures like Bill Gates and Elon Musk about how dangerous AI can be if left unchecked.

Skewed Reality

Artificial Intelligence (AI) disruption is a real possibility soon, with its potential to drastically reshape day-to-day operations and redefine existing industries. Despite this, the full range of implications for AI disruption still needs to be discovered. While it is clear that it could lead to immense benefits for many businesses, the ramifications of this technological advancement reach far beyond profit margins.

There is an enormous potential for discrepancies in how we perceive reality if we fail to consider how AI programming can influence decision-making processes or create centralized technologies that compromise personal data privacy. We must consider many scenarios when weighing the cost versus benefit of AI disruption or run the risk of a distorted reality.


In summary, generative AI technology has the power to both positively and negatively impact our lives at a rapid pace. It can manipulate people into false belief systems and undermine public discourse by facilitating new realities and narratives. On the other hand, AI models that specialize in narrative intelligence can be used to identify potentially harmful conversations and filter out malicious actors. Ultimately, leveraging the right tools and technologies is key to preventing disinformation and ensuring we can access accurate information that represents a nuanced view of perspectives in open conversation - one in which all voices are heard without manipulation or fabrication.

About Blackbird.AI
Blackbird.AI helps organizations detect and respond to threats that cause reputational and financial harm. Powered by their AI-Driven Narrative & Risk Intelligence Constellation Platform, organizations can proactively understand risks and threats to their reputation in real-time. Blackbird.AI was founded by a team of experts from artificial intelligence, and national security, with a mission to defend authenticity and fight narrative manipulation. Recognized by Forrester as a "Top Threat Intelligence Company," Blackbird.AI's technology is used by many of the world's largest organizations for strategic decision making


While all these recommendations seem to be sound, the likelihood that these measures can be agreed upon and implemented are becoming increasingly less likely in the U.S. and around the world. In fact, we have been moving in the opposite direction. Platforms have begun to roll back access for research communities, decrease moderation around misinformation, or strike down moderation altogether in the name of freedom of expression. The very notion of banning a popular platform in the U.S. would have seemed unthinkable a few short years ago, with organizations like the ACLU strongly voicing that a ban on TikTok would violate the First Amendment.

- asdf