TimeGPT: Production Ready Time Series Foundation Model for Forecasting

The website (https://www.nixtla.io/) says all sorts of companies are using this: Ford, Lyft, Nestle, T-Mobile, Wayfair, etc etc.

That's not actually true, right? Surely that's a lie? I find it very unlikely that these huge companies are actually using this, especially considering how sketchy the website is (the About page doesn't list any employees/executives, there are no case studies, no white papers describing what this product even is.)

I get the impression this product only exists to scam investors who have unrealistic expectations of AI.

Closed-source. I'm not sure why this is interesting. ML API companies hoping to hit it rich are a dime a dozen.

With the exception of a few like OpenAI, I can't imagine much success over open source models.

If it can genuinely predict stocks, start a hedge fund.

The link to the anomaly detection use case is broken https://docs.nixtla.io/docs/tutorials-anomaly_detection

Training a model with multiple billion floating point parameters on only 100 billion data points feels like a bad idea.

> TimeGPT outperforms a comprehensive range of well-established statistical and cutting-edge deep learning models, consistently ranking among the top three performers across various frequencies.

I'm immediately skeptical af if there is nothing more explicit than this. That's handwaving bs. Top three means nothing if I don't have proper data.