0) big companies already are very comfortable using contracts to trust other people with their data. Maybe if they're inflexible on the 30-day requirement for fable some orgs will opt out but by-and-large it's already happening and it's not a blocker.
1) Cost will be a blocker. The level of token spend is untenable and the pareto curve is flattening. Most orgs are going to default to either using a distilled model from China or a distilled model from the model companies (e.g, Sonnet 5). It'd behoove Claude/OpenAI to offer a model router before another vendor wins that area.
2) Karp is selling his book. No one knows or cares what an 'ontology' is. From what I can tell, company's product is a tool that helps governments bomb people.
The problem is nobody can build a real business on models they don't control. Cursor focused on making a great AI coding experience but they didn't control Claude and got destroyed once Anthropic started explicitly training Claude to be better at using Claude code.
And then there's the cost front, a good model is soon going to be served at 90+% margin, meaning the model makers have a ton of flexibility in how they choose to subsidize their first party products. Cursor had to eat the full API price of Claude while Anthropic could sell Claude Code at a massive subsidy.
If you're an AI product then the model product synthesis is your product, and you don't get any of that from a proprietary closed model.
Palantir has sold their flavor of "consulting" (forward-deployed whatevers) as an alternative to the old-school flavors of consulting (management consultants). Now, Alex Karp is upset because there's a new, new flavor of consulting (agents) that they now have to sell against.
Not only are your FDEs old news, but the new hotness is an order of magnitude cheaper. Tough sell.
Is there active research in improving the density of models i.e. higher performance at lower parameter counts? Is this something that's even theoretically possible to do with the architecture?
I feel like I can have a much better model for how I can expect the cost of LLMs to change over the next few years. "Build more datacenters" reeks of "just one more lane bro" and because of Jevons paradox I'm skeptical it will bring costs down. Could ASICs and custom processors (ala Groq) move the needle here at all if the model density can't be improved?
Yes, LLM density has continuously improved. A recent 32B model beats older >100B models. It may level off at some point but no one knows when that will be.
Voice of American *capitalists* used to the idea of investing capital to generate rent. Maybe AI can rebalance the market of capital and individual labor.
I can't decide if in my life span we did anything beyond the internet to actually make life better as a primary driver of economic growth. it feels like _everything_ is just rent seeking now.
I surprised myself by reading some content without knowing where it came from and it had a position that owning more than one home is hoarding. It dug into the relationship of landlords and buildings providing apartment homes vs single family homes. I'm not going into the whole property is theft part of it but it really struck a chord with me. I have a personal friend who constantly goes on about how great renting properties is and how much money his family makes and it really smacked me in the face how it really shouldn't "be this way".
anyways... we feel this rent seeking everyday with the death-by-a-thousand-papaer-cuts that is SaaS and subscription revenue.
i'm getting tired as i get older and i'm not sure how to make the most of the next ~half of my career and life.
You’re right. These are not stupid people. They are likely very out of touch, but they are not stupid. They know exactly what they are doing and they are good at it.
Often times, a stupid take is useful to manipulate people or change the discourse. We argue about how dumb the take is and get distracted about what really matters.
It’s the same strategy that causes working people to fight each other and ignore the real enemy.
A few things:
0) big companies already are very comfortable using contracts to trust other people with their data. Maybe if they're inflexible on the 30-day requirement for fable some orgs will opt out but by-and-large it's already happening and it's not a blocker.
1) Cost will be a blocker. The level of token spend is untenable and the pareto curve is flattening. Most orgs are going to default to either using a distilled model from China or a distilled model from the model companies (e.g, Sonnet 5). It'd behoove Claude/OpenAI to offer a model router before another vendor wins that area.
2) Karp is selling his book. No one knows or cares what an 'ontology' is. From what I can tell, company's product is a tool that helps governments bomb people.
[dead]
The problem is nobody can build a real business on models they don't control. Cursor focused on making a great AI coding experience but they didn't control Claude and got destroyed once Anthropic started explicitly training Claude to be better at using Claude code. And then there's the cost front, a good model is soon going to be served at 90+% margin, meaning the model makers have a ton of flexibility in how they choose to subsidize their first party products. Cursor had to eat the full API price of Claude while Anthropic could sell Claude Code at a massive subsidy. If you're an AI product then the model product synthesis is your product, and you don't get any of that from a proprietary closed model.
Palantir has sold their flavor of "consulting" (forward-deployed whatevers) as an alternative to the old-school flavors of consulting (management consultants). Now, Alex Karp is upset because there's a new, new flavor of consulting (agents) that they now have to sell against.
Not only are your FDEs old news, but the new hotness is an order of magnitude cheaper. Tough sell.
Is there active research in improving the density of models i.e. higher performance at lower parameter counts? Is this something that's even theoretically possible to do with the architecture?
I feel like I can have a much better model for how I can expect the cost of LLMs to change over the next few years. "Build more datacenters" reeks of "just one more lane bro" and because of Jevons paradox I'm skeptical it will bring costs down. Could ASICs and custom processors (ala Groq) move the needle here at all if the model density can't be improved?
Taalas has a pretty impressive proof of concept, but there's been no news about them recently.
https://chatjimmy.ai/
Yes, LLM density has continuously improved. A recent 32B model beats older >100B models. It may level off at some point but no one knows when that will be.
Voice of American *capitalists* used to the idea of investing capital to generate rent. Maybe AI can rebalance the market of capital and individual labor.
I can't decide if in my life span we did anything beyond the internet to actually make life better as a primary driver of economic growth. it feels like _everything_ is just rent seeking now.
I surprised myself by reading some content without knowing where it came from and it had a position that owning more than one home is hoarding. It dug into the relationship of landlords and buildings providing apartment homes vs single family homes. I'm not going into the whole property is theft part of it but it really struck a chord with me. I have a personal friend who constantly goes on about how great renting properties is and how much money his family makes and it really smacked me in the face how it really shouldn't "be this way".
anyways... we feel this rent seeking everyday with the death-by-a-thousand-papaer-cuts that is SaaS and subscription revenue.
i'm getting tired as i get older and i'm not sure how to make the most of the next ~half of my career and life.
Archive is giving a 500 error. Anyone got another paywall bypass?
All I can say is, ‘fuck these people’.
No one is actually this stupid.
You’re right. These are not stupid people. They are likely very out of touch, but they are not stupid. They know exactly what they are doing and they are good at it.
Often times, a stupid take is useful to manipulate people or change the discourse. We argue about how dumb the take is and get distracted about what really matters.
It’s the same strategy that causes working people to fight each other and ignore the real enemy.