Computers Quote by Andrew Ng
“Want to train a machine translation system? Train it on a gazillion pairs of sentences of parallel corpora, and that creates a lot of breakthrough results. Increasingly, I'm seeing results on small data where you want to try to take in results even if you have 1,000 images.”
About This Quote
Source Talk: Machine Learning Conference, 2022
Training translation models on massive parallel data yields breakthroughs, yet even small datasets can produce valuable results with careful techniques.
In simple terms: Big data helps, but small data can work too.
Key Takeaway
Leverage both large and small data wisely.
Themes
machine learning
translation
data efficiency
innovation
scalability
Mood
practical
optimistic
Type
technical
inspirational
When to use this quote
- building MT systems
- research labs
- startup product development
Key Concepts
Transfer learning
few‑shot learning
data augmentation
Questions to Reflect On
- How can you maximize performance with limited data?
- What trade‑offs exist between data size and quality?
A Different Perspective
Limited data may not capture rare language nuances.
![“[Brazil] went to the UN and said, "We need new standards for this." We need to take a look at what they're calling "data sovereignty."” quote by Edward Snowden](https://sms4smile.com/quotes/img-quotes/af7/edward-snowden-555_brazil_brazil-went-to-the-un-and-said-we-need-n_14906.jpg)
