In a departure from the traditional cadence of local TV weather updates, Haystack News has embraced artificial intelligence to revolutionize how communities receive weather information. Using an automated workflow powered by Amazon’s AWS, the company utilizes Large Language Models (LLMs), text-to-speech technology, and an image generator to deliver AI-generated weather reports every hour. Currently serving 30 cities, including Cleveland, Houston, Cupertino, Kansas City, and St. Louis, Haystack plans an ambitious expansion to over 100 locations in the next two weeks, with the goal of reaching every U.S. city within months.
While the use of AI in news organizations remains a contentious topic, Haystack emphasizes its commitment to augmenting, not replacing, traditional news coverage. With licensing agreements with major media companies such as ABC, Fox, Nexstar, Hearst, and Scripps, Haystack aggregates local news for 97% of U.S. TV markets through its mobile and smart TV apps. CEO Daniel Barreto envisions AI’s role in enhancing news coverage, particularly in noncontroversial areas, with weather forecasting being a prime example due to its data-centric nature.
The inadequacies of current local TV weather coverage are apparent, as demonstrated by San Francisco’s micro-climates being condensed into a single regional report. Haystack aims to address this gap, recognizing that variations in weather can be drastic, especially in urban areas and even more pronounced in poorer rural regions disproportionately impacted by extreme weather events. Acknowledging the importance of weather information in daily life, particularly for rural residents, Haystack’s AI-generated forecasts are positioned to provide more localized, accurate, and accessible updates.
Founded in 2013 with financial backing from European TV-maker Vestel and other investors, Haystack embarked on the development of its AI-generated weather forecasts six months ago. The process involved extensive trial and error, internal testing, and refinement. The simplified formula involves feeding local weather data into a large language model, generating a script for a concise one-minute forecast. This script undergoes conversion through a text-to-speech engine, complemented by automatically generated infographics, ensuring a seamless integration of AI technology into the realm of local weather reporting.
While Haystack News relies on weather data from reputable sources such as the National Weather Service and major vendors to minimize the risk of AI “hallucinations,” the company did face early challenges with the AI misinterpreting data, as acknowledged by CEO Daniel Barreto. Instances occurred where the AI erroneously associated cold weather with pleasant conditions, illustrating the potential pitfalls of relying solely on automated models. To maintain user trust, avoiding such errors is crucial for Haystack, according to Erickson Strategy & Insights analyst Paul Erickson. The risk of people associating inaccuracies in AI-generated weather reports with the entire service could lead to a loss of trust over time.
Beyond ensuring the accuracy of AI-generated content, scalability and cost control are significant considerations. Unlike other video content, weather reports have a short lifespan, expiring within hours, making repurposing and monetization on social media challenging. This temporal constraint questions the feasibility of broader AI ventures in the news industry, such as those attempting to create entire networks with AI anchors and synthetic voice-overs.
Contrasting with experiments requiring substantial manual intervention, Haystack has successfully automated its weather reporting process and is exploring the application of AI in other data-heavy domains, including local traffic reports. Barreto envisions a broader integration of AI in traditional TV broadcasting, foreseeing its omnipresence in the long term. While other news networks experiment with AI-generated anchors, Haystack’s approach stands out as a scalable and practical implementation, showcasing the potential for AI to become a staple in various facets of media production.