jimmah (@jamesdouma) 's Twitter Profile
jimmah

@jamesdouma

deep learning dork

ID: 24405165

calendar_today14-03-2009 18:10:06

9,9K Tweet

49,49K Followers

248 Following

jimmah (@jamesdouma) 's Twitter Profile Photo

Eugene Wigner: “It was a burden growing up so close to him. I often wonder if my horrific inferiority complex, which not even the Nobel Prize has diminished in the slightest, is a product of having known von Neumann for the better part of my life.”

jimmah (@jamesdouma) 's Twitter Profile Photo

Ashok on Q324 earnings call: “Like you mentioned, Elon, we already made a 100x improvement with 12.5 from starting of this year and then with v13 release we expect to be 1000x from the beginning from January this year.” At that time both Elon & Ashok predicted Q2/3 2025 for

jimmah (@jamesdouma) 's Twitter Profile Photo

Nice. Glad to see more people thinking through this carefully. Some comments: 1) The functional TOPS of the silicon is dominated by the NNU, and (conceptually) modest architectural changes to the NNU have a big impact on TOPS without much power or area change. For example -

jimmah (@jamesdouma) 's Twitter Profile Photo

While hype is a real thing, it’s also true that deep learning continues to outperform simple extrapolations of existing trends, often by orders of magnitude.

Elon Musk (@elonmusk) 's Twitter Profile Photo

Sawyer Merritt The Samsung fab in Texas will make AI6, not AI5, and the decision regarding using HBM has not been made yet. As the ratio of total parameters to active (frequently used) parameters increases, the cost per unit of useful compute and compute per Watt may favor relatively

Elon Musk (@elonmusk) 's Twitter Profile Photo

Sawyer Merritt You can fit more total RAM on the board if you use “normal” memory than high-bandwidth memory and it is super cheap. Maybe high-bandwidth memory is still the right choice, but using HBM isn’t the slam dunk many people think it is.

jimmah (@jamesdouma) 's Twitter Profile Photo

If Tesla’s Robotaxi efforts continue to go well we could see both US road accidents and US road emissions halved in less than 10 years.