Large Models
topic on 5 shows · 14 statements across 12 episodes
Innovators & Investors 
Latent Space 
No Priors 
the a16z Podcast 
20VC
14 statements about Large Models, every show
Sanseviero: Local models handle agentic capabilities well, but world knowledge requires scale
“With local models or models that you can run in your own hardware, you can get capabilities, so you can get agent capabilities, function calling, system instructions, like conversational, and that kind of stuff. Knowledge is much trickier, so for knowledge, yo…”
Yang: Quantum computing and large models will achieve indistinguishable consciousness
“I truly believe that specifically the quantum computing based processors and big, large models. Will be indistinguishably conscious.”
Houston: Large models bottle cognitive energy like Industrial Revolution mechanized labor
“I believe that the large models, that's really the first time we can kind of bottle up cognitive energy and offload, you know, if we started by offloading a lot of our mechanical or physical busy work to machines, that freed us up to make a lot of progress in …”
Sarah Guo says massive frontier AI models are impossible to serve commercially
“Over time, applications are going to want efficient inference, and, like, really large models are impossible today to serve for the vast majority of use cases from a cost and speed perspective”
Sarah Guo notes training frontier models requires co-locating GPUs for data transfer
“Today to train these large models, you need all of the GPUs co-located because there is enough data transfer between different chips, right? Between your nodes. And there's a physical constraint on that in that you need to get that much power and to a data, da…”
Argenti: Large proprietary AI models lead in reasoning capabilities
“Probably nobody beats those large models with regards to actually reasoning capabilities.”
Casado: Generative AI models bring the marginal cost of creation to zero
“So it's pretty clear if you just take the fundamental economic analysis that these large models bring the marginal cost of creation to zero, like creating that image, and language understanding, like reasoning over those documents.”
Cohen: Undertrained large AI models underperform well-trained smaller models
“If you have a large model, That is undertrained. It will underperform a small model, which is really well, like well, well trained. So you're just wasting resources and you're going to get like less efficient results.”
Ghodsi: Smaller custom models can beat large LLMs on domain accuracy
“And there you're better off if you have a good data set to train, you can train a smaller model. The latency will be faster to use it later, and it will be cheaper to use it later, and yes, you can have absolutely accuracy that beats the really large model, bu…”
Large AI models will reduce the marginal cost of creation to zero
“So I think there's a pretty good analog where you say these large models actually bring the marginal cost of creation is there was some very fuzzy, vague notion of what creation means, but for sure we could talk about it of like content, conversation, whatever…”
Appenzeller: Over-training smaller AI models allows them to match larger ones
“You can match the performance of a large model with a smaller model if you train it more, right?”
Large Language Models Are Creative Fiction Engines Rather Than Factual Databases
“I do say these large models as well should be viewed as fiction, creative models, not fact models, because otherwise we've created the most efficient compression in the world. Does it make sense you can take terabytes of data and compress it down to a few giga…”
Small Hive Architecture AI Models Will Massively Outperform Large Monolithic Models
“I think that small models will outperform large models massively, like I said, the Hive model aspect”