30 years in tech, Cloud, Storage, HPC/AI, VCP, TOGAF, ITIL, Amateur radio. IEEE Sr Member. USAF 2A352 F-117, & Lt. Civil Air Patrol. Tweets are my own.
Had a request to attempt modeling if a nor'easter was possible. Can't validate any accuracy as my program is designed to model hurricanes. But figured I would post the results for comments or feedback. Thoughts? @ChrisMartzWX
@ncfrontiersman@NixFred Thank you we have tons to talk about. And as this is my first season, I am #4 in accuracy of intensity & track error across the 2 previous storms. So I can't complain too much.
Hey @ncfrontiersman as mentioned have my own weather stuff, mainly hurricane modeling. Call it NextWave Weather, decided to run it because of the potential for a TS in the gulf. Still some AI hallucinations to clean up but I am getting there, here is today's result. @NixFred
@gh_TechLuddite@NixFred@XSpaces@ncfrontiersman Interesting, have to circle back on this. Working with an entity where we want to take old school massive distribution like SETI, now BOINC & run it internal to harness raw HPC/AI from spare cycles. I used to work at one w/400K end user machines, lot of unused resources
@MichaelGannotti@NixFred Your much further along but we should chat. Overwatch visualizes OSINT. my ScorchedAI could index it w/merge vectors + graph so a filing, a ship, a headline, and a case note become one queryable entity graph with deterministic correlation
@MichaelGannotti@NixFred Thanks. Hey @MichaelGannotti you happen to have any of the ND 455x available in the Azure Gov locations yet? Got some customers I might need to send your way.
I am just another shmoo trying to do what I can for the community. Going to benchmark with DGX OS, then try @jdvmi00 Developer preview of omarchy-spark, GB10 is a bit different though at the bios. If it works, I may try to get a FIPS validation or STIG built out, DoD/W may be interested.
Moving our govt forward. Win for @wwt_inc w/ @Dell & @Nvidia AI Factory, named in an IDIQ contract at a fed agency combining B300's & GB300 NVL72 systems. @Google part of this deal as well. On-prem site survey begins next week while we prep for delivery. Cheers! @MichaelDell
@HotAisle@OderintMetuant Please don't be like that. You responded w/cloud on subs, I said it doesn't change quickly on those systems. You switched to modern warfare tech in Ukraine, I implied we have plenty of new tech. Two different avenues of implementation. Now can we get back to how awesome AMD is?😎
Sorry you're misinformed on world battlefield technology. Tons of tech happening in gov. Even little ol me has patents pending to enable next gen war fighters. Post quantum tech, I hope the public cant have access to for at least 10 years. Trying to figure out how to develop it to give to @DARPA
@HotAisle@OderintMetuant Don't be like that. I got a lot of respect for you and you know that. But a bunch of nuclear missiles as part of a system that runs under water is a heck of a lot different than modernizing the Pentagon or a three letter agency. Lots of advances are happening.
@HotAisle@OderintMetuant I am with you, but swapping her/sw in something like one of the legs of the nuclear triad has life threatening consequences if something goes wrong. But we have some good things in the rest of the DoW/IC community.
@HotAisle@OderintMetuant It's not about budgets or costs. it's about programs of record..you don't just change what a Virginia class submarine has for technology. those kinds of things have to be scheduled for its next dry dock upgrade, sometimes every 5-10+ yrs.
@HotAisle@OderintMetuant Not on subs.Underway for long times, w/very little bandwidth when they surface. Teacher student distillation, retrain when they get back, but even dock speed limited. all systems have to download & update. lot of sneaker net happens at port
File a formal FCC complaint. It forces ISP to write official letter in response on how they resolved it to your satisfaction. Took a couple complaints but I had 4 regional techs in my house at once working on it, never had a problem since. Presidents secretary knew me personally very quick.
LOL, They're still 'losing' the software one because y'all treat CUDA like it's the One Ring.
just hipify it. better yet, hipifly & keep your CUDA with basically zero changes. Port is trivial for over 90% of real workloads. But sure, keep paying NVIDIA ''ecosystem' tax while AMD serves bigger HBM at lower prices. The moat is real.. it's called inertia + laziness.
Hardware gap closing fast. Cope gap closing slower. 🫡