Machine Learning Models
topic on 14 shows · 30 statements across 27 episodes
Another Podcast 
We Live to Build 
Latent Space 
Lenny's Podcast 
No Priors 
Sourcery 
Capital Allocators 
Top Founders 
the MAD Podcast 
How I Built This 
the a16z Podcast 
Big Technology 
TBPN 
20VC
30 statements about Machine Learning Models, every show
Chess.com relies on proprietary data and machine learning to detect cheating
“We spend a lot of time, a lot of money, a lot of resources on anti-cheating basically, and we have a whole set of tools, which I'm not going to get into because I don't like to give ideas to people on how we do it, but we track a lot of stuff. We have more dat…”
Xin: Databricks uses ML models to optimize database algorithms at runtime
“They use that to build a model. Like a machine learning model, not an L, a machine learning model. Machine learning model basically can very, very quickly tell us how any algorithm and how any implementation will perform for any specific type of queries with v…”
Kulik: ML models deliver 100x to 1000x speedup per optimization dimension
“Usually, just even for a not so accurate machine learning model, you get, you know, at least a hundred to a thousand-fold speed up for every dimension you're optimizing over”
Reganti: Workflow Automation Always Requires Combining ML Models and Deterministic Code
“Whenever you're trying to automate some part of a workflow, it's never the case that you could use an AI agent and that will kind of solve your problems, right?
It's always, you probably have a machine learning model that's going to do some part of the job.
Yo…”
Mahr: MDT trains its machine learning models on 50 years of data
“We train our models on roughly 50 years worth of data, which I say that to some potential investors and they're surprised. We think that market data from the 19 seventies and eighties is still useful for forecasting mispricing.”
LLMs cannot beat traditional machine learning for core ranking systems yet
“In core ranking systems I haven't seen anything yet that's, like, much better than the state-of-the-art machine learning models. So I'd say we're, at least in core ranking systems, probably not a lot of LLMs.”
Shadbolt: Claims of Quantum Machine Learning Acceleration Are Premature Hype
“Quantum machine learning is really nice to think about. It sounds awesome in like a press release. It's really early is how I put it. Like the development of the theory is not yet at a stage where you can confidently say quantum computing is going to be awesom…”
Morcos: Data curation choices fundamentally determine machine learning model performance
“There are a ton of choices you would make in that process, ranging from how you're going to filter the data, how you're going to sequence the data, what synthetic data you're going to generate, if any how you're going to batch the data, all of those things. An…”
Morcos: Proper data curation enables smaller models with equal or better performance
“Help the folks we work with to train models much faster to much better performance and to also help them train much smaller models to the same or better performance, which I actually think is some of the most exciting stuff going forward. But fundamentally, th…”
Masad: Randomness in machine learning is a feature enabling creativity
“Input-output machine learning models have inherent randomness, and that's a feature, not a bug that creates creativity, right?”
AI Models Will Test and Verify Their Own Work Within Six Months
“Like I think over the next three or six months, I think we're going to see machine learning models being able to test and verify their work.”
Hitchcock: SurrealDB allows users to run AI models directly inside databases
“One of the pieces of functionality we have in CeruleeB is the ability to bring a model right inside the database so that you can run that model. It's a custom trade model or an off-the-shelf model. You can run that model right alongside your data.”
Biewald: Simple operational errors cause more model failures than data drift
“People talk a lot about data drift in the industry. And that's this idea that like, you know, like language changes over time and you want to know that it's changing and sort of like have your model you know, notice that and update it. But I guess like what I …”
Guu: Continuous Real-Time ML Weight Updating Is Rare Due to Validation Costs
“Those I initially thought would have been more popular in kind of production grade settings, but they come with a maintenance cost, which is if you have something updating live, you don't have the opportunity to validate and check that everything is going well…”
Evans: ML models lack structural understanding and rely entirely on statistics
“And the same thing with machine learning models, this is what we've always been saying. You know, get right back to kind of, I don't have any structural understanding of what it is. You know, they may recognize the difference, they may be able to tell a cat fr…”
Menker: Gro Intelligence runs 28 frameworks to generate 2 million unique models
“Today we have 28 modeling frameworks in the system... And so today, those 28 modeling frameworks actually develop two million unique models by Grow.”
ML models make delivery platforms far more profitable than greedy solvers
“While a greedy solver can get you maybe initially 70% of the way there, that's fine for a while. Layering on, you know, all of this, you know, additional machinery made it, you know, much more profitable over time.”
Schneider: We may never deeply understand the inner workings of ML models
“I don't know that we can come to much better of an understanding about what's going on in the middle or what that means, because there's so many little things happening.”
Douetteau: Model auditing and regulation are bigger enterprise bottlenecks than ML performance
“In, in some use cases, it's not machine learning per se, or the performance of machine learning models that is the choke point in order to deliver value. It's the ability to actually meet the, meaning the regulatory, the regulation constraints, meaning literal…”
Pinterest uses 50 machine learning models in its pin safety pipeline
“And so, there's a, that whole pipeline has got like about 50 different machine learning models in itself.”
Pinterest wants non-coding data scientists to deploy ML models to production
“And to get to a point where we can have people that are data scientists that don't even code, that can just build models and be able to deploy those to production. So that's really what we want to get to within Pinterest as well.”
Training dedicated models for distinct cohorts beats fitting general models
“If there's a strong enough use case for a different cohort or segment like new users, typically it'll make more sense just to train a new model rather than trying to shoehorn it into a general model.”
Alex Wang: Synthetic data is nowhere close to impacting machine learning progress
“It hasn't really worked in practice in most situations, particularly when it comes to visual data or even textual data. It really hasn't worked, and a lot of the reasons why it's sort of limited in terms of its overall impact is that at the end of the day, whe…”
Stoica: ML models degrade over time as real-world data evolves
“The models you developed on some data set, and for instance, the data or the queries are going to evolve over time. And because the environment or the world around you evolves, what you learned and which is embedded in the model may not be as relevant or as go…”
Nguyen: Human Intuition Functions Like Parameters in a Machine Learning Model
“Well, what we think of as intuition are actually, you can think of as parameters inside a machine learning model.”
Bloom: Machine learning's destiny is personalized models for everyone
“The manifest destiny is everybody has their own machine learning models built upon their own past behavior, perhaps leveraging some of the insights that you wind up getting from the whole system.”
SigOpt fine-tunes existing client models instead of entirely replacing them
“So instead of just taking a raw data set of decades of fraud data and giving them some model to like rip and replace what they already have, we sit on top of what they have and provide this additive boost by fine tuning it.”
Machine learning models often predict outcomes without explaining why they happen
“The models that you can create can often tell you what might happen, but they can't always tell you why.”
Achin: Complex machine learning models can be interpreted using modern techniques
“I would argue, though, that even very complicated models can be interpreted using modern techniques.”
Wallach: Investigating correct model predictions helps contextualize how models treat certainty
“I don't know if there are necessarily any sort of general purpose, like this is going to fix everything kind of solutions, but I would say yes, digging into why your model is making certain predictions, even when those predictions are correct, can kind of help…”