It's a wrap from Delhi!
The past few days have been amazing, had the great honor to be at the Plenary when Prime Minister @narendramodi addressed world AI leaders. His message was simple and powerful: AI must scale responsibly and inclusively, not just quickly.
One research talk that really stayed with me was Prof. @SuryaGanguli's session on Digital Twins. The idea of building mirrored systems and feedback control loops for AI felt very aligned with what weโve been building at @fiddler_ai: explainability, observability, and making AI systems accountable in real time.
As AI moves from models to autonomous agents, you need Visibility, Context, and Control.
Grateful to be part of these conversations at such an important moment for India and for AI globally. Some moments from our trip! cc @mainak
Dhanyawad India! ๐ฎ๐ณ ๐
Today, weโre excited to share that Fiddler AI has signed a definitive agreement to be acquired by @CrowdStrike.
We started Fiddler to help businesses understand AI. Together with CrowdStrike, we intend to turn that understanding into the confidence to put AI to work across the enterprise.
To our customers, partners, and team: thank you for trusting us with important systems and shaping our work through your hardest production problems.
Read more from Krishna: https://t.co/YIDJkH2CPg
Today, weโre excited to share that Fiddler AI has signed a definitive agreement to be acquired by @CrowdStrike.
We started Fiddler to help businesses understand AI. Together with CrowdStrike, we intend to turn that understanding into the confidence to put AI to work across the enterprise.
To our customers, partners, and team: thank you for trusting us with important systems and shaping our work through your hardest production problems.
Read more from Krishna: https://t.co/YIDJkH2CPg
More validation from our AI Science research on Evals:
โข Small classifiers, adapted with either full fine-tuning or low-rank adaptation (LoRA), can outperform generalist models 10x their size.
โข Generalist models retain a significant edge on inputs unlike the data the small classifiers were trained on.
โข Training on post-verdict reasoning lets large teacher models produce superior student classifiers without the latency penalty that reasoning imposes at prediction time.
https://t.co/FyxeqHKBjV
Interesting and important!!
That said, evaluating synthetic data (subjects) using traditional survey research methods with all its well-known biases is inherently limiting.
Imagine if the timeline were flipped and survey science came *after* synthetic data. How would we use surveys to validate synthetics, rather than the other way around? Flipping that script might be the breakthrough we need.
Today, we celebrated our 8th anniversary at @fiddler_ai, remembering some highlights from our journey from explaining machine learning models to building a control plane for AI agents. Our mission around building trust in AI has never been so important. Really grateful to all our employees, investors, customers and partners for being a part of this journey! More to come!