"Datadog has created a model called Toto-Open-Base, which is built with 151 million parameters and trained on more than two trillion time-series data points, making it the largest pretraining dataset for any open-weights time-series foundation model."
"Li believes that developing models brings two advantages: AI becomes part of Datadog's platform, and it leads to better agents that can detect and predict anomalies more effectively."
"For AI systems to win trust, their output must be both explainable and verifiable. Using its own models makes that easier for Datadog, allowing them to create tools that monitor AI platforms for signs of hallucinated output."
Datadog is preparing to launch an updated AI model, Toto-Open-Base, designed to enhance its observability services. This model, built with 151 million parameters and trained on over two trillion time-series data points, aims to provide better anomaly detection and root cause analysis. The company emphasizes the importance of domain-specific models over generic ones. Datadog's chief product officer highlights the need for AI outputs to be explainable and verifiable to gain trust, especially in mission-critical IT environments.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]