Data Quality
topic on 7 shows · 12 statements across 12 episodes
Innovators & Investors 
We Live to Build 
Latent Space 
No Priors 
the MAD Podcast 
Big Technology 
20VC
12 statements about Data Quality, every show
Gupta: AI cannot fix bad data without upstream human validation
“If the same bad quality data goes, AI is not going to fix it. No one can fix it. There is someone who has to check it.”
Lacroix: Data quality improvements yield 10x the gains of model architecture tweaks
“Getting the data perfect, because we knew this was potentially not the most exciting part of the work, but it was absolutely critical, and any improvement on the data quality would, Tenex, the improvements that we would get by really improving on the model arc…”
Rai: The most useful enterprise problems are usually boring topics like data quality
“You know, so I think the most interesting problems which are useful is very, are usually very boring in nature. So data quality is a very boring topic. No one would say, get excited by data quality. But that's a huge problem.”
Sansbury suggests base AI models are hitting an asymptotic performance plateau.
“Models are getting better, but maybe they're approaching this kind of asymptotic barrier where are thirteen billion parameters really that much better than twelve billion parameters, right? Maybe nine is good. Maybe five. We, someone mathematically has done th…”
Morcos: Data diversity is the single most important factor in AI data quality
“If there's only one thing that you should take away from this entire interview about what is good for data quality, it's diversity.”
Edwin Chen: Data Quality Is the Primary Bottleneck to AI Progress
“So I would definitely rank data quality first, followed by compute, followed by the IRLs.”
Burke: Virtual cell accuracy is limited by data quality, not architecture
“One of our conjectures is that one of the reasons the models aren't doing well is not just simply model structure. We have a lot of rich structures that we understand in the ML space and machine learning space that the issue is the Data quality.”
Unified models enable faster error corrections via refinement clicks
“And we found that, you know, going from each phase, it both improved the efficiency and it improved the data quality. And in particular, when you get rid of this two-part model, one of the advantages is that when you make refinement clicks, so You prompt the m…”
Mensch: Compute scale alone hits a hard ceiling without high-quality data
“Scale isn't the only recipe, the only ingredients to the recipe you need to scale, but you also need to have proper data. Otherwise you reach some data quality limit.”
Rouif: AI data quality is more important than data size
“On data, what matters is quality more than size”
Mostaque: AI performance depends on data quality, not quantity
“It's not about more data, it's about better data.”