In the wake of the 2024 US presidential election, one fact became clear: disinformation was spreading online at an alarming rate. shapes the views of Americans about each candidate, in addition to a variety of various issues, including public health, climate change and immigration. Generative AI – with its ability to supply deepfakes in seconds and its propensity to hallucinate facts – will only exacerbate the issue.
Factuallya startup that took part within the TechCrunch Disrupt Battlefield 200 in October is preparing for the onslaught. The company, which won the very best pitch within the Security, Privacy and Social category, has developed a business-to-business tool that permits live fact-checking of text, video and audio. The company's pitch: to assist corporations save hours of research and mitigate any reputational or legal liability risks.
The Norwegian startup continues to be in its infancy; Faktiverse has raised around $1.45 million in pre-seed funding since its launch in 2020. However, in line with Maria Amelie, CEO and co-founder of Fativerse, the corporate has already began working with media and financial partners, including one in all Norway's largest banks.
Factiverse even enabled live fact checking US presidential debates which was utilized by several media partners, said Amelie.
“We usually are not an LLM (large language model). “We developed a special form of model based on information retrieval,” Amelie told TechCrunch.
As a former technology journalist and published writer, Amelie has first-hand experience fighting facts. She worked with Fativerse co-founder and CTO Vinay Setty, who’s an associate professor of machine learning on the University of Stavanger, to launch the B2B-focused startup.
According to Amelie, the Factiverse model is predicated on high-quality, well-curated and credible data from reliable sources and fact-checkers world wide, slightly than the “junk food data” that generative AI is trained on.
“We train our AI model to intuitively think like someone who has a whole lot of experience researching information,” Amelie said.
The model, based on machine learning and natural language processing, is capable of discover claims and search the online in real time – from serps like Google and Bing to AI serps like You.com to academic papers.
“The funniest part is that we don’t show you what shows up first in these serps,” Amelie said. “We actually suggest to you which of them sources in your topic are probably the most credible or have been probably the most credible prior to now… We actually research the world related to the subject and sometimes even who’s quoted in an article.”
Factionverse today says it outperforms GPT-4, Mistral 7-b and GPT-3 in its ability to discover fact-checkable claims in 114 languages. The company's model also outperforms LLMs in determining the veracity of a claim. Amelie said Factiverse's success rate is about 80% and the goal is to enhance as the corporate takes on latest customers world wide.
“We come up with the money for to be the very best, but we’re here within the US to be the fastest,” Amelie told TechCrunch. She also noted that the corporate would really like to lift a seed round in 2025. “We are on the lookout for customers and investors who want to take a position in trust and credibility,” she said.
