Locator: 49753CUDA.
From wiki, a screenshot -- it's interesting how seldom we see CUDA mentioned -- and yet it appears to be the "key to the magic kingdom."
*****************************
Stanford
Note: I only supply the AI prompt; the answer is way to long to post. But you can easily put the prompt into a chatbot and see how this develops. I use ChatGPT.
On another note, Nvidia's success is due to CUDA, discovered and developed by a Stanford University graduate student whose work caught the attention of both Nvidia and DARPA. That simply amazes me.
So, this AI prompt:
Years ago, it seemed to me that it was the East Coast -- Harvard, Princeton, MIT -- that had the lead on almost everything that mattered -- banking, finance, tech, and, of course liberal arts (Harvard, Yale), but then some years ago it seemed this all switched to the west coast (Bay Area, specifically) and then even more specifically, Stanford. Every time I turn around, going deep into the history of something, a graduate student at Stanford seems to crop up. Again, tonight, reviewing CUDA over at wiki, the line, "Both Nvidia and AMD teamed with Stanford University to create a GPU-based client." Then we learn that it was a PhD graduate STUDENT who was key in "discovering" how GPUs could be used for general purpose parallel computing beyond just graphics (Ian Buck --> Brook --> CUDA). Nvidia noticed Buck's work and from there, the rest is history, as they say. Is it just my imagination or is there some special sauce at Stanford? In finance, it was a Stanford alumnus that brought us Schwab, SoFi, and Robinhood. This doesn't seem like coincidence. It's almost as if Harvard Business School is really, really good at writing "case studies" for students to study, but at Stanford, the alumni put things into action. What is Stanford's special sauce?
Then, the follow-up prompt:
Compare the Stanford and MIT entrepreneurship pipelines would be the most interesting.
Absolutely fascinating.
So, this led me to the next ChatGPT prompt:
This reminds me of something I mentioned some weeks ago: making chips is almost mundane in the big scheme of things. The brilliance comes from figuring out what to do with these chips, how to optimize them, how to maximize their potential. Investors (and the public in general) are fixated on Intel (or at least they were six months ago) but at the end of the day, they are just making chips. On the other hand, companies like Nvidia (and maybe AMD, I don't know) are trying to figure out to get the most out of the chips they are developing. The Nvidia GPUs --> Ian Buck --> Nvidia / DARPA --> is just such an incredible example. I know it's not that simple or straightforward, but as important as "a chip" is these days, it's the language that is developed to maximize the potential of those chips, and it's the ability how to stack or package various chips (GPUs, CPUs, NPUs, XPUs, memory) etc that is really the secret sauce. Focusing on chip manufacturing (like Intel) seems so 1980s where as packaging, stacking, language (so various chips can talk to each other), etc is so 2020s. Do you see Nvidia (and perhaps companies like AMD) more than just chip makers (like Intel) or am I misreading what Intel is doing?
Again, absolutely fascinating.
And, finally this prompt:
Too much to digest right now. We'll come back to this but you've hit the nail on the head. And I guess I should have seen that myself. In this sphere, it's back to basics -- trying to imitate (for lack of a better word) how the human brain works. A gazillion neurons, but it's not how many neurons exist (not how many transistors six on a chip) but how many dendrites and synapses each neuron has but even more importantly, how good these dendrites are in locating the "right" node (or the next "right" neuron) -- and interestingly, AI is doing just that -- which I still think is "magic" and impossible to explain. Also, interestingly, it seems the basic of the human brain's success is the communication between neurons (dendrites) is often (always) hit-and-miss, and evolves. AI, on the other hand, is using math (statistics, vectors, etc) to get Nvidia's dendrites to communicate with each other, which then allows you to responds to my rambling thoughts.
It's all about dendrites.
