Are you aware of HIP? It's officially supported and, for code that avoids obscure features of CUDA like inline PTX, it's pretty much a find-and-replace to get a working build:
This is for access to compatible 250kW Supercharges. There are a lot of them, but it is a subset of the total Supercharger network. [1]
If you check the Find Us map and select "Superchargers Open to NACS" then you will see which ones are compatible and their associated 'Charging Fees for NACS EVs' rates.
The main problem with the Groq LPUs is, they don't have any HBM on them at all. Just a miniscule (230 MiB) [0] amount of ultra-fast SRAM (20x faster than HBM3, just to be clear). Which means you need ~256 LPUs (4 full server racks of compute, each unit on the rack contains 8x LPUs and there are 8x of those units on a single rack) just to serve a single model [1] where as you can get a single H200 (1/256 of the server rack density) and serve these models reasonably well.
It might work well if you have a single model with lots of customers, but as soon as you need more than a single model and a lot of finetunes/high rank LoRAs etc., these won't be usable. Or for any on-prem deployment since the main advantage is consolidating people to use the same model, together.
In case you don't know about it, make sure to check out SponsorBlock, which optionally, automatically skips parts you don't want to see like for example sponsors (but not only). You can use this awesome DB for what you want to do.
And also Invidious.
And also their combination.
User of both, I'm unaffected by the recent adblocking issues on YouTube and I can still subscribe to channels, reliably. Without any Google account. (by the way, a simple regular RSS feed reader would do, since YouTube provides RSS feeds for each channel, but Invidious is a really convenient, specialized UI for this, without the notification / algorithm issues that seem to plague YouTube wrt this, but I digress)
All these issues are already solved by these projects. I guess one could consider contributing to them (financially or with code for instance). The official YouTube frontend actually don't need no love, others already achieve what we want.
Piped, NewPipe and FreeTube are also projects to look into (I loved NewPipe when I had a smartphone, Piped looks very good too and FreeTube looks interesting but I know less about them).
QEMU 6.2 introduces full hypervisor support for M1 Macs. I've installed ARM Ubuntu Server in a VM on my MacBook, and wrote a blog post explaining it. It uses libvirt to start the VM in headless mode.
Chia pre-mined 6 years worth of block rewards. Proof-of-space may be the closest thing we have to a viable proof-of-work replacement but this instantiation of it feels like a get rich quick scheme for its founders.
Edit: My calculation was based on the initial reward schedule of 64 units per block, but when you take into account the fact that it ramps down to 4 units per block shortly thereafter, it's actually a premine of 21 years: https://www.coindesk.com/5-takeaways-from-chia-networks-new-...
It's amazing to me that there are distinct, concentric rings on all of our keyboards that are very analogous to the concentric rings you find in trees revealing their age. The oldest keys are in the center, newest are along the edge. (I count about 5 rings: https://breckyunits.com/how-old-are-these-keys.html)
(Maybe someone could take this idea and make a nice tree ring visual of the keyboard?)
Obama started putting US citizens on the kill list in 2010 and once they started killing them barely anyone complained so it doesn't surprise me they just assume they can keep doing it.
Not sure why link a tweet, what you're looking for is [1]. While we're at it, take a look at MIT's other courses at [2]. They are amazing. I wish all schools'd put up their classes online for free.
Some of the most interesting excerpts (although it's worth reading in its entirety):
> My path in technology started at Facebook where I was the first Director of Monetization. [...] we sought to mine as much attention as humanly possible and turn into historically unprecedented profits. We took a page from Big Tobacco’s playbook, working to make our offering addictive at the outset.
> Tobacco companies [...] added sugar and menthol to cigarettes so you could hold the smoke in your lungs for longer periods. At Facebook, we added status updates, photo tagging, and likes, which made status and reputation primary and laid the groundwork for a teenage mental health crisis.
> Allowing for misinformation, conspiracy theories, and fake news to flourish were like Big Tobacco’s bronchodilators, which allowed the cigarette smoke to cover more surface area of the lungs.
> Tobacco companies added ammonia to cigarettes to increase the speed with which nicotine traveled to the brain. Extreme, incendiary content—think shocking images, graphic videos, and headlines that incite outrage—sowed tribalism and division. And this result has been unprecedented engagement -- and profits. Facebook’s ability to deliver this incendiary content to the right person, at the right time, in the exact right way... that is their ammonia.
> The algorithm maximizes your attention by hitting you repeatedly with content that triggers your strongest emotions — it aims to provoke, shock, and enrage. All the while, the technology is getting smarter and better at provoking a response from you. [...] This is not by accident. It’s an algorithmically optimized playbook to maximize user attention -- and profits.
> When it comes to misinformation, these companies hide behind the First Amendment and say they stand for free speech. At the same time, their algorithms continually choose whose voice is actually heard. In truth, it is not free speech they revere. Instead, Facebook and their cohorts worship at the altar of engagement and cast all other concerns aside, raising the voices of division, anger, hate and misinformation to drown out the voices of truth, justice, morality, and peace.
I highly recommend Lex Fridman's recent podcast with David Patterson [1] for anyone interested in learning about the history of RISC, computer architecture, and also interesting predictions re: Moore's Law.
Nigel Smart's Cryptography Made Simple is a great book which covers elliptic curve cryptography amougst many other topics; despite its name it's a very technical book, but it's easily accessible to anyone with a CS/EE/Maths/Physics degree.
> The Brookings Institution thinks that the US can shut China out of high-end semiconductor production indefinitely. That seems like a pretty foolish conception to me.
U.S. had done that to Japanese Semiconductor industry in the 80s and 90s[1], which was why South Korean semiconductor industries flourished and Samsung became the king of Memory chips (it used to be Toshiba etc.) And today it's Huawei. Who's gonna be the next in the future?
We’re developing open source industrial machines that can be made for a fraction of commercial costs, and sharing our designs online for free. The goal of Open Source Ecology is to create an open source economy. [...]
OSE (Open Source Ecology) is currently developing a set of open source blueprints for the Global Village Construction Set (GVCS) – a set of the 50 most important machines that it takes for modern life to exist – everything from a tractor, to an oven, to a circuit maker. In the process of creating the GVCS, OSE intends to develop a modular, scalable platform for documenting and developing open source, libre hardware – including blueprints for both physical artifacts and for related open enterprises.
The current practical implementation of the GVCS is a life size LEGO set of powerful, self-replicating production tools for distributed production. The Set includes fabrication and automated machines that make other machines. Through the GVCS, OSE intends to build not individual machines – but machine construction systems that can be used to build any machine whatsoever. Because new machines can be built from existing machines, the GVCS is intended to be a kernel for building infrastructures of modern civilization.
To me as a Swiss citizen, born in Switzerland, living in Switzerland, it's an impertinence that every time I open a bank account with a Swiss bank I have to fill out at least 1 form with information about my relation to the US, about my travels in the US and whether I might have to pay taxes there.
As a side note, the only other country besides the US that collects taxes based on citizenship and not based on residence is Eritrea. The tax policy of Eritrea is described as extortion by a lot of western governments and media.
I thought I was reading my own blog post :) We use very similar tech stack at my current company:
- Ansible for provisioning
- Python/Django for website/api
- VueJS for frontend(where needed, some pages are simple Django templates)
- Celery for background work
- uWSGI and Nginx as servers with AWS Load balancer
- Elasticsearch for search
- Redis for caching
- Postgres with Postgis as main datastore
- Datadog for monitoring
- Cloudflare for DNS
Some differences as I am working with a team:
- We do use multiple branches and git tags for releases. Feature branches are also common as multiple devs maybe working on different features.
- We use Gitlab-CI a lot for testing and auto-deployment(ansible script can be called from our machine as well)
- Terraform for infrastructure provisioning. We have stopped provisioning any AWS service by console. Once the service is provisioned by terraform, ansible takes over.
I have tinkered with Docker, Hashicorp Packer but this setup has been dead simple to reason and scale reasonably well.
I update this list from time to time. It overwhelmingly supports the premise that most of the founders of successful (or otherwise well-known) tech companies are typically in their 30s or older (and not ~18-22):
Paul Graham (31, Viaweb); Jan Koum (33, WhatsApp); Brian Acton (37, WhatsApp); Ev Williams (34, Twitter); Jack Dorsey (33, Square); Elon Musk (32, Tesla | 31, SpaceX | 27, PayPal); Garrett Camp (30, Uber); Travis Kalanick (32, Uber); Brian Chesky (27, Airbnb); Adam Neumann (31, WeWork); Reed Hastings (37, Netflix); Reid Hoffman (36, LinkedIn); Jack Ma (35, Alibaba); Jeff Bezos (30, Amazon); Jerry Sanders (33, AMD); Marc Benioff (35, Salesforce); Ross Perot (32, EDS); Peter Norton (39, Norton); Larry Ellison (33, Oracle); Mitch Kapor (32, Lotus); Leonard Bosack (32, Cisco); Sandy Lerner (29, Cisco); Gordon Moore (39, Intel); Mark Cuban (37, Broadcast.com); Scott Cook (31, Intuit); Nolan Bushnell (29, Atari); Paul Galvin (33, Motorola); Irwin Jacobs (52, Qualcomm); David Duffield (46, PeopleSoft | 64, Workday); Aneel Bhusri (39, Workday); Thomas Siebel (41, Siebel Systems); John McAfee (42, McAfee); Gary Hendrix (32, Symantec); Scott McNealy (28, Sun); Pierre Omidyar (28, eBay); Rich Barton (29, Expedia | 38, Zillow); Jim Clark (38, SGI | 49, Netscape); Charles Wang (32, CA); David Packard (27, HP); Craig Newmark (43, Craigslist); John Warnock (42, Adobe); Robert Noyce (30, Fairchild | 41, Intel); Rod Canion (37, Compaq); Jen-Hsun Huang (30, nVidia); James Goodnight (33, SAS); John Sall (28, SAS); Eli Harari (41, SanDisk); Sanjay Mehrotra (28, SanDisk); Al Shugart (48, Seagate); Finis Conner (34, Seagate); Henry Samueli (37, Broadcom); Henry Nicholas (32, Broadcom); Charles Brewer (36, Mindspring); William Shockley (45, Shockley); Ron Rivest (35, RSA); Adi Shamir (30, RSA); John Walker (32, Autodesk); Halsey Minor (30, CNet); David Filo (28, Yahoo); Jeremy Stoppelman (27, Yelp); Eric Lefkofsky (39, Groupon); Andrew Mason (29, Groupon); Markus Persson (30, Mojang); David Hitz (28, NetApp); Brian Lee (28, Legalzoom); Demis Hassabis (34, DeepMind); Tim Westergren (35, Pandora); Martin Lorentzon (37, Spotify); Ashar Aziz (44, FireEye); Kevin O'Connor (36, DoubleClick); Ben Silbermann (28, Pinterest); Evan Sharp (28, Pinterest); Steve Kirsch (38, Infoseek); Stephen Kaufer (36, TripAdvisor); Michael McNeilly (28, Applied Materials); Eugene McDermott (52, Texas Instruments); Richard Egan (43, EMC); Gary Kildall (32, Digital Research); Hasso Plattner (28, SAP); Robert Glaser (32, Real Networks); Patrick Byrne (37, Overstock.com); Marc Lore (33, Diapers.com); Ed Iacobucci (36, Citrix Systems); Ray Noorda (55, Novell); Tom Leighton (42, Akamai); Daniel Lewin (28, Akamai); Diane Greene (43, VMWare); Mendel Rosenblum (36, VMWare); Michael Mauldin (35, Lycos); Tom Anderson (33, MySpace); Chris DeWolfe (37, MySpace); Mark Pincus (41, Zynga); Caterina Fake (34, Flickr); Stewart Butterfield (31, Flickr | 36, Slack); Kevin Systrom (27, Instagram); Adi Tatarko (37, Houzz); Brian Armstrong (29, Coinbase); Pradeep Sindhu (43, Juniper); Peter Thiel (31, PayPal | 37, Palantir); Jay Walker (42, Priceline.com); Bill Coleman (48, BEA Systems); Evan Goldberg (35, NetSuite); Fred Luddy (48, ServiceNow); Michael Baum (41, Splunk); Nir Zuk (33, Palo Alto Networks); David Sacks (36, Yammer); Jack Smith (28, Hotmail); Sabeer Bhatia (28, Hotmail); Chad Hurley (28, YouTube); Andy Rubin (37, Danger | 41, Android); Rodney Brooks (36, iRobot); Jeff Hawkins (35, Palm); Tom Gosner (39, DocuSign); Niklas Zennström (37, Skype); Janus Friis (27, Skype); George Kurtz (40, CrowdStrike); Trip Hawkins (28, EA); Gabe Newell (33, Valve); David Bohnett (38, Geocities); Bill Gross (40, GoTo.com/Overture); Subrah Iyar (38, WebEx); Eric Yuan (41, Zoom); Min Zhu (47, WebEx); Bob Parsons (47, GoDaddy); Wilfred Corrigan (43, LSI); Joe Parkinson (33, Micron); Aart J. de Geus (32, Synopsys); Patrick Byrne (37, Overstock); Matthew Prince (34, Cloudflare); Ben Uretsky (28, DigitalOcean); Tom Preston-Werner (28, GitHub); Louis Borders (48, Webvan); John Moores (36, BMC Software); Vivek Ranadivé (40, Tibco); Pony Ma (27, Tencent); Robin Li (32, Baidu); Liu Qiangdong (29, JD.com); Lei Jun (40, Xiaomi); Ren Zhengfei (38, Huawei); Arkady Volozh (36, Yandex); Hiroshi Mikitani (34, Rakuten); Morris Chang (56, Taiwan Semi); Cheng Wei (29, Didi Chuxing); James Liang (29, Ctrip); Zhang Yiming (29, ByteDance);
I read years ago in comp.risks about a similar story. A guy in 1979(!) requested a personalized plate "SAILING", with second choice "BOATING". He didn't want a customized plate if he couldn't get those, so for his third choice he put down "NO PLATE". Of course, he ended up with "NO PLATE". He ended up getting 2500 parking tickets, since cars with no plate had "NO PLATE" written on the ticket.
One book that changed me was reading Master and Margarita in Russian for the first time.
It was the first book I started reading I could not put down until the end. Gained a lot of appreciation for literature at that time.
The other book that I enjoyed and changed me was ‘The Wisdom of Insecurity’ by Alan Watts. I was a fan of Alan Watts works through his lectures already and it was wonderful to hear his ideas in writing for the first time.
I wish everyone read or watched Alan Watts lectures and books. The world would be a much nicer place if that was the case.
My favorite quote is by him:
‘We thought of life by analogy with a journey, a pilgrimage, which had a serious purpose at the end, and the thing was to get to that end, success or whatever it is, maybe heaven after you’re dead. But we missed the point the whole way along. It was a musical thing and you were supposed to sing or to dance while the music was being played.’
I have one connected to my amplifier and speakers — it's the only thing connected, I don't need any other audio source.
I run miniDLNA[1] which shares all my music, and BubbleUPnP[2] on my phone to browse the music and cast it to the Chromecast Audio.
It's also possible with a Python library[3]; I used this for a while as an alarm clock.
About every 2-3 months I have to reboot the Chromecast, but it's otherwise fine. It doesn't do gapless playback, but I think that's partly because miniDLNA is running on an ultra-low-power ARM board, i.e. is slow.
https://github.com/ROCm/HIP
Don't believe me? Include this at the top of your CUDA code, build with hipcc, and see what happens:
https://gitlab.com/StanfordLegion/legion/-/blob/master/runti...
It's incomplete because I'm lazy but you can see most things are just a single #ifdef away in the implementation.