I Found the NVIDIA GTC Keynote to be quite interesting. The first topic, The Omniverse seemed useful, but I had some difficulty figuring out how exactly it is used. I know they said it was a digital platform for 3D simulation, but some of the examples they used for it didn’t seem to line up entirely. I was interested in how it is used for the Isaac Simulator. The use of the digital twins made with the Omni Verse in the Isaac Simulator, fascinates me because it seems like a great way to train robots without any danger to the people, though digital environments are never exactly like the real world. In any case, I found this part to be interesting.
For the Deep Learning Architecture, I found their introduction of the new Grace CPU for AI and its memory to GPU speed, to be intriguing, even if I may never actually interact with one of the CPUs. Additionally, I found the discussion of the increase in data parameters in large-scale models to be riveting as I can barely imagine what GPT-4 or 5 will be capable of, with that amount of data. For any of the other computer hardware they showed off, like the Super Pod Datacenter or the AI Supercomputer in the Box, I have no attachment to it, simply because I am not the target market for those, as I do not have anywhere near the money needed for their cheapest machine.
Out of their new AI models, I liked Jarvis the best, as I think its Language Translation potential to be the most useful thing out of all the AI they showed at the conference. I understand why the Merlin Recommendation Model and the Maxine Virtual Conference Model are important, but I don’t care enough about better shopping recommendations or maintaining eye contact to find those AI Models interesting. The Tao pre-trained model modification was neat though, and I think that has real potential for using pre-trained models in the real world.
Regarding the self-driving car section of the talk, I am excited for when they eventually are perfected and I am impressed by the Driving Simulator (hopefully the simulator also has bad drivers for the car to react to), but until I can ride in one, I am not that interested.
Final note, I understand that the Tech industry moves very fast, and because of that, when the next iteration of any kind of computing chip comes out, it will always be bought in massive amounts. I also understand that tech is exponential, and the next iteration of those chips will be much more powerful than what is currently on the market. But despite that knowledge, it still feels weird that they are discussing the next chip after the one they had announced one minute beforehand. It brings up memories of the Osborne Computer fiasco.
For anyone who is interested, the Osborne Computer fiasco is a legend that says that a company, Osborne Computer Corporation, announced their flagship PC and then announced the specs of the PC that was going to come out after that one. Because the specifications were much better, everyone knew to wait for the next PC, and no one bought the flagship model. The company couldn’t support itself after the flop and went bankrupt, so the next PC never came out. In reality, there were more complicated factors to the Osborne Computer Corporation’s collapse, but the legend is more fun to tell. To get back to my point, it reminds me of that story when NVIDIA announces the next iteration of the chip right after announcing the newest chip, but I know that the mechanics of the Cloud Computing Hardware market are not the same as the Personal Computer Market.
Thank you.
