InkCanon

12 ms·

121.
by InkCanon 2y ago
I second this. Time spend is Pareto distributed, and you spend most of your time doing something because you don't know something. Research, homework, coding, you spend the most time caught in a loop for something you don't know
122.
by InkCanon 2y ago
I don't think I was informed by sensational headlines. I was well into talking to people I knew about how DS's performance relative to compute was a game changer much before the stock crash. It's not binary where you either h
123.
by InkCanon 2y ago
I used to think the same way wrt Nvidia stock when it tanked - compute is clearly diminishing returns. Tech companies subsequently announced capex equal to or greater than expected in compute. I smacked myself on the forehead when I realize
124.
by InkCanon 2y ago
I'm also somewhat confused by it. To my (admittedly limited) knowledge of immigration in the US, it was not really endorsed by any major figure or movement. In fact there were many anti immigration bills and laws written during that ti
125.
by InkCanon 2y ago
Not sure if it's just me, but it looks like all SOTA companies are doubling down to chase the new benchmark, which beyond hype, doesn't seem to translate into real world uses. Why don't these companies just plug it into a pop
126.
by InkCanon 2y ago
I think they definitely are. To customers, there's close to zero perceptible benefit when X model scores significantly higher on some benchmark they never heard of (or even know what is about). So they're desperately trying to mak
127.
by InkCanon 2y ago
Because interviews were always an attempt to discern a signal few hours interview into an accurate prediction of performance from months to years. AIs generate a lot of nosie to mask that. Interviewees can just pass the question to the AI,
128.
by InkCanon 2y ago
To my understanding, OpenAI won the case where it argued training was covered under fair use and did not infringe on copyright.
129.
by InkCanon 2y ago
It's like that Dr Phil episode where he meets the guy who created Bum Fights!
130.
by InkCanon 2y ago
I think it's a desire for minimalism. In a world filled with websites saturated with ads, clickbait and constant attention grabs, this is like walking from an overcrowded city street into a Zen garden.
131.
by InkCanon 2y ago
Exactly. Any rational person will predict Nvidia will own the space. But projected over a long period, a 10% decrease in growth is massive
132.
by InkCanon 2y ago
To clarify, a legitimate benchmark for training is to calculate the running cost, not capex cost. Because obviously the latter would drop dramatically with the number of models you train. But to put into context, Meta wants to spend 50B on
133.
by InkCanon 2y ago
You're on the dot here. Most people are looking at it wrong. Someone here once put it succinctly - Nvidia isn't selling GPUs, it's selling Infiniband clusters. The whole NVLink/CUDA ecosystem is the key, and a mass of se
134.
by InkCanon 2y ago
People here quoting Jevons paradox are grossly misapplying it. Jevons paradox will predict an increase in usage of compute for AI. Not Nvidia specifically. The specific insane valuation of Nvidia comes from an unending stream of hundreds in
135.
by InkCanon 2y ago
Jevons paradox has been quoted to death here - it really needs pointing out that it points to a commodity in general , not Nvidia. Nvidia is not selling just coal, it's selling a super purified ultra powerful coal only it can make. Je
136.
by InkCanon 2y ago
But if they don't sell them, are they spending 50 billion a year on open source goodwill?
137.
by InkCanon 2y ago
If anything Meta and other foundation model companies should be dropping. What's the moat? DeepSeek is releasing stuff for free.
138.
by InkCanon 2y ago
It's arguable how good a strategy it is to check against other P/E. During the tech bubble people would say X is cheap, because Y is trading at 100P/E instead of 200
139.
by InkCanon 2y ago
How do you pass these models code bases?
140.
by InkCanon 2y ago
The question depends what you mean by first principles. Usage of the phrase "first principles" has sprawled into many different things since (I think) Musk first mentioned it as a way to learn. The original, philosophical meaning
141.
by InkCanon 2y ago
Kind of cute they gave credit to the AI for writing its own paper.
142.
by InkCanon 2y ago
Contextually, yes. DeepSeek is just a hundred or so engineers. There's not much promotion to speak of. The promo culture of google seems well corroborated by many ex employees
143.
by InkCanon 2y ago
DeepSeek really is taking out OpenAI at the knees. It's shocking that the first direct peer competition to OpenAI is also doing it for an order of magnitude less as a side project.
144.
by InkCanon 2y ago
The way it has destroyed the sacred commandment that you need massive compute to win in AI is earthshaking. Every tech company is spending tens of billions in AI compute every year. OpenAI starts charging 200/mo and trying to drum up 5
145.
by InkCanon 2y ago
IMO thinking about it as business domains first isn't good. From "first principles" you have technology (which is rapidly changing). After the technology has been made sellable, people come up with a label for it to put what
146.
by InkCanon 2y ago
Nice product. I assume you're using some kind of RAG on quarterly releases? I personally would pay ~$30/mo for something that uses the same data but a different way. I want a "Stripe for company data". There are financia
147.
by InkCanon 2y ago
This. Virtually anything that can be changed, can be innovated. And people seem to underestimate how complex the world is. Even something as unchanging as a commodity can increase in demand massively because of second order effects - like l
148.
by InkCanon 2y ago
But there's still a ton of innovation going on in cars. Tesla and BYD are some of the biggest companies in the world. And conversely, rockets was regarded as a sleepy, government contracting business for a long time. I don't parti
149.
by InkCanon 2y ago
On the last point: "The industry has matured. The low-hanging fruit of the mobile/web era has largely been picked, making truly innovative opportunities harder to find." It is a repeating pattern in history that people think
150.
by InkCanon 2y ago
"How do browsers take advantage of this fact? By reserving 4GB of memory for every single WebAssembly module." Does reserve mean it has exclusive access to? Because it can't possibly be that every single wasm module takes 4GB

More ›