Facticity.AI is a new AI-powered fact-checking tool developed by Singapore startup AI Seer, claiming 92% accuracy in verifying claims across text and video content - significantly higher than competitors like Bing CoPilot and Perplexity.
The tool was successfully tested on recent content related to the US election, analyzing claims across different media formats including videos and podcasts.
AI Seer is partnering with ORA to decentralize Facticity through an Initial Model Offering (IMO), using blockchain technology to ensure the tool remains globally accessible, censorship-resistant, and community-governed.
“In a world of disinformation, Facticity.AI, an artificially intelligent fact-checking tool developed by Singapore start-up AI Seer, seeks to tip the scales in the truth’s favor. Facticity’s online tool, released in beta in June, checks claims in text and video, and provides references and links to reliable sources. The company claims 92% accuracy, beating competitor tools like Bing CoPilot, Perplexity, and Originality.ai (which claims about 72% accuracy on fact-checking). Facticity was put to the test during the most recent presidential debate, checking almost 250 claims in near real-time. Jacob Kozhipatt, a former Business Insider reporter who worked with Facticity to manually verify its checks, says he was impressed by “how much more and how much faster it can fact check than humans.”
— Time Magazine, Best Inventions of 2024, Facticity
In today’s digital landscape, misinformation proliferates easily, and the demand for reliable fact-checking has never been higher. Facticity was created to combat this issue, leveraging AI to deliver high-speed, accurate fact verification. By combining large language models (LLMs) and extensive databases into an agentic pipeline, Facticity is capable of delivering accurate, nuanced fact-checking at a reported 92% accuracy rate.
Facticity’s capabilities were put to the test during recent U.S. presidential rallies in Nevada, where it analyzed statements from Kamala Harris and Donald Trump:
For Harris’s 25-minute rally video clip, Facticity flagged 30 claims:
• True: 17
• False: 6
• Unverifiable: 7
For Trump’s 90-minute rally video clip, Facticity flagged 72 claims:
• True: 24
• False: 39
• Unverifiable: 9
Facticity’s flexibility extends across media formats, such as podcasts, where it recently evaluated a conversation on The Joe Rogan Experience. These tests demonstrate Facticity’s ability to perform fact-checking in real-time across diverse media, making it an adaptable tool for today’s rapidly evolving information environment.
Facticity’s service is a crucial tool for humanity in the current precipice of technological development. Misinformation, disinformation, and lies impede human progress by sabotaging collaboration and resource efficiency. Truth facilitates greater, more efficient collaboration and allows individuals to make better choices for themselves and their spheres of influence. There is no limit to how Facticity can be integrated and used by people across the world to evaluate the media they consume, even while the rate of false media production increases with AI.
To ensure Facticity’s long-term neutrality, accessibility, and resilience, Seer is collaborating with ORA to decentralize Facticity’s AI model through an Initial Model Offering (IMO). The IMO not only ensures that Facticity remains globally accessible and censorship-resistant, but it also invites stakeholders into a token-based governance model that fosters transparency and community-driven development.
The IMO framework combines two main components:
ERC-7007: Verifiable Inference Asset – securely anchors Facticity’s AI model onchain.
ERC-7641: Intrinsic RevShare Token – distributes revenue based on token ownership, allowing stakeholders to benefit from Facticity’s usage.
This model guarantees that Facticity’s fact-checking service remains accessible and fair, with decisions about model updates and expansions governed by its community. Token holders participate in the governance of Facticity, and the token fuels its ongoing development and integration into new applications.
In some areas of the world there is intentional misinformation by governments. In others, social media leads to the spread of misinformation and disinformation, intentionally or unintentionally. For example, the US has sounded the alarm about intentional misinformation efforts from foreign adversaries in the buildup to the 2024 election.
In the era of Internet technology, individuals have a greater ability to make informed decisions for the health of themselves and their community when they can access services like Facticity’s. Facticity can be used by anyone to check the legitimacy of the information they see online. Facticity’s fact-checking service empowers democracy and freedom. It gives individuals the ability to make more consistent, high-quality decisions across various aspects of their lives which can be influenced by media. On average, Americans consume about 11 hours of daily media, which means Facticity has a large surface area on which to enact positive change.
It is imperative for the continued improvement of human society across the world that services like these are accessible. These services should also be credibly neutral, ideally with their continued development governed by all stakeholders that may use them. This ensures that fact-checking does not develop a bias toward a specific ideology, group, or perspective.
To ensure continued and global access to Facticity’s services, as well as verifiable neutrality and decentralized governance, tokenization on a decentralized blockchain provides an exciting option to examine. By tokenizing Facticity, we can render Facticity:
Globally accessible
Uncensorable
Governable by all stakeholders
Monetizable by all stakeholders
ORA and Facticity are collaborating to decentralize their fact-checking model, thereby ensuring:
Global access
Censor-resistance
Community governance
Additionally, by creating a token-based economy around Facticity’s AI, we establish a positive incentive mechanism to drive continued development and application of the AI model in new applications.
The existence of a Facticity token, which can collect and distribute revenue for any application of the model onchain, incentivizes Facticity’s community (and prospective stakeholders) to expand the use of Facticity’s model. This can lead to exciting new integrations, driven by the exponential power of network effects.
You can see how this incentive mechanism has come in to influence the development and application of the OpenLM model; the first AI model to be tokenized onchain.
Since its IMO, the OpenLM community has passed governance proposals to fine-tune their models and developed several applications of the model that provide a service onchain in exchange for a fee. These fees are then distributed back to the community. Read more about the first IMO here.
The IMO is a very customizable framework that can meet the ecosystem needs of many such models through token design.
Facticity.AI was created by AI Seer Pte. Ltd., a Draper Associates portfolio company and a deep tech company accredited by IMDA Spark which has published its patent on a Mult-Spectral Reality Detector that has finished lab trials and is being field trialed by expert polygraphers and academics. Tim Draper funded them to empower everyone to detect lies and misstatements of facts in everything they see, hear, and read.
Facticity.AI’s long-term aim is to make incredible information credible like never before by being the first to use Multiple LLMs on the last living pictographic languages in East Asia and Hieroglyphics in Egypt to create a Rosetta Stone for ancient cave drawings, and Nazca Lines and unravel human history back to its pre-written beginnings.
www.facticity.ai | x.com/facticityai
ORA is chain-agnostic infrastructure that bridges the gap between AI and blockchain.
We empower developers with the tools necessary to build end-to-end trustless and decentralized applications enhanced by verifiable AI.
ORA is live today, offering the capability to verify and run inference on the largest and most sophisticated AI models without limitations.
ORA
No comments yet