Now when it comes A.I. it’s all well and good to be neutral and proscribe caution, and say there’s lots of good and bad on “both sides”. But let’s also consider physically, de facto the tangible artifacts of this pursuit. This is why I love EETimes (Engineers can be hostage to trends, and fashions as anyone else, but there’s also the tendency to want to measure and quantify as well). In this article the new AI Supercomputer “Inflection” is announced along with it’s particulars (NVidia GPUs, and lots of ’em), along with the attendant costs to acquire, much less RUN the thing. A relevant pull-quote:
When finished, Inflection’s installation will be 22,000 Nvidia H100 GPUs, making it both the largest AI cluster and one of the largest computing clusters in the world. All for a chatbot. Nvidia’s own AI supercomputer, Eos, a 4,600-GPU monster, is still in the bring-up phase, but will be dwarfed by Inflection’s cluster.
I’m a gen-Xer. So I grew up reading Douglas Adams at an impressionable age. The reason I say that is there’s a number of side/tangent stories throughout A Hitchiker’s Guide to the Galaxy. And the one most relevant to this is the idea that computers beget bigger computers and so on. From Deep Thought sprung the answer: 42. And from Deep Thought the Earth was created to find and understand the actual meaning of the original question, What is the meaning of Life the Universe and Everything.
Inflection is not as big as the Earth. But it is BIG. So this supercomputer is 5X bigger than the computer NVidia itself (who actually designs and builds the underlying GPU) uses for training and inference. In the article, the author estimates roughly the costs alone for the hardware: $800Million USD. Followed by costs to run the cluster for one day: $700,000 USD. So one has to ask, how the hell do you charge customers money, and make a profit having sunk all that capital into building and operating a behemoth like Inflection? Do you ever get to that point? When do you make money?
Or is it more like a Cold War mentality? Where competitors in the AI space attempt to cow all comers into submission. Meaning that having an asset at your disposal makes competitors less likely to “attack” because they cannot scale up to that size? And more so, costs of Nvidia GPUs are not going down the longer they manufacture them, Nvidia has a monopoly on it’s GPU technology, they’re never going to let that go (ever). So you won’t see anything like price competition for AI Supercomputers, ever. It’s just going and going, growing and floating on Venture Capital and investors, hoping like hell there’s a pay-off, sometime, somewheres in teh futures.
Folks are really trying to optimize the way they do writing to the web these days. Not least of which are the growing number of static website generators. Today just now @kftiz was outlining a new workflow for her blog using the Eleventy static site generator and hosting via Reclaim Hosting (instead of Github). So far #webmentions are the only really missing piece of the puzzle. ‘Cuz who doesn’t want to know who is reacting, reading and writing in response to one’s own writing. It’s a Read/Write Web after all. Let’s see to that. Nobody is shouting down a well.
May not always say it or brag about it buy I never have spent a penny of my own money on Spotify.com. However, I know full well that I am now the product of Spotify.com. But I don’t care ‘cuz one thing I’ve learned once I finally got off my butt and started LISTENING to all the recommendations coming out of their algorithmic suggestions engine. I spent a number of late afternoons evenings in Winter and Fall some years (I think 2015 was when I started?) picking my favorite songs from albums I once owned. And like a dutiful A.I. powered robot, Discover Weekly would come back with 30 tracks EVERY week. Without fail.
I spent the first couple years not paying any attention because I thought recommendation engines were dumb and ONLY promoted what record companies wanted (a little like commercial radio payola amirite?!) So I ignored it, but my old standby Apple iTunes was slowwly transmogrifying into Apple Music and also recommending music as well. I tried that, liked it and took Spotify much more seriously. And wouldn’t you know it, it got better and better the more data I fed to it. And again I wasted probably all of 2015-17 ignoring suggestions. But I finally around 2017-18 got serious about listening to all 30 Discover Weekly tracks to start liking/disliking stuff. Again, algorithmically, Spotify is dialed in, and the mark of a great algorithmic suggesetion engine is SERENDIPITY.
A lot, over 75 to 80% of the recommendations are bands I never heard of or never new existed. And probably don’t exisst today ‘cuz back catalog and all that. But what’s really cool and special is when the bands and solo artists still are around, producing stuff and publishing it up to Spotify. That’s how I’ve bumped into odd little historical outfits (Yellow Magic Orchestra) and better known but obscure tracks (Nilsson). But today I learned that Japanese Breakfast/Michelle Zauner is amazing/fantastic, but in particular I had to share the track as delivered to me as #16 of 30/from the album Soft Sounds from Another Planet. Diving Woman track #1 from the album. What can I say?
Take Throwing Muses, and move it ahead to the present roughly, but just the most kickin’ bassline one can imagine with a single guitar riff played over top. All I know is, haven’t heard the same similar before, and now I want to listen to all of the Japanese Breakfast catalog. And I owe it all to Spotify.com. I learned one thing in this journey since 2015 is,
1.) I may “think” I know about all the music that’s out there, but I am wrong
2.) I know this because for the 8 years since 2015 Spotify has never/EVER made a single repeat on Discover Weekly
3.) I have found more obscure, interesting, and dare I say it “enjoyable” stuff that I know could never have been suggested, shared with me by a single recorded music nerd.
I would need a network? Nay and ARMY of devoted recorded music nerds to suggest the tracks (over 8,400 and counting) that I have liked since I’ve started listening to Discover Weekly on Spotify.com. So even though I have to occasionally listen to ads from Geico, Grammerly and Doritos, I don’t care. What I’ve gotten in return for me data feels like so much more than what I sacrificed. So here’s my praise, #fanboi celebration off bumping into an artist I wasn’t familiar with (or I should say AS familiar with, I had liked 1 track from Japanese Breakfast about 2 weeks ago). There’s so much music out there, I’ll never get to listen to all of it, but with Spotify, at least I’ll FIND, or discover the stuff that’s being made that I like and is being published/released.
At one time back around 1996 or so, browsers were being updated an released at such a torrid pace. You had to practically download and update 2x per month to get all the features. Not least of which were all the multimedia plug-ins to get the latest viewers for all the multimedia being produced.
I say this because I saw on Mastodon someone had been trying to get a web-browser working on an old computer, showing an old webpage for a University “Campus Wide Information System” thats what some IT orgs called their websites back in teh day. Because sometimes things were made available via gopher:// links and sometimes through http:// links, it was an interesting wild-cat, frontier days atmosphere.
The biggest barrier to getting an old web-browser working an old OS to display an old webpage was the need for https:// and the support libraries on the OS and the browser pre-cluded some choicies of web browser. For instance you can get a web browser from Google in 2002 because Chrome didn’t exist. But Netscape Navigator Did! But the https:// and SSL certificates and their underlying OS support libraries (things like OpenSSL) don’t run against Netscape Navigator 4.0 very well, if at all. And no easy fixes to civilian types like myself or artists attempting to recreate an old experience like that.
But back to my original paragraph, about the rate of change, and flurry of activity as someone in EdTech say in 1996 days, was that Netscape Navigator released updates fast and furious to add new functions constantly. New plug-ins were released by various and sundry outfits trying to get their new media tech adopted and market dominant. And darned if Microsoft wasn’t trying to get a browser into people’s hands that would compete with Netscape. So much so, they had to build it into Windows as part of the OS to get people to start using it regularly. I haven’t seen before or since this level of the new, new thing hitting FTP servers (at the time) and amount of work being done by developers for these desktop apps. And it wasn’t even for security exploits, or anything it was just to have NEW things you could do on the web. <sigh!> A bit of nostalgia over the web that was.
“Ultimately, to figure out what we really need to worry about, we need better AI literacy among the general public and especially policy-makers. We need better transparency on how these large AI systems work, how they are trained, and how they are evaluated. We need independent evaluation, rather than relying on the unreproducible, “just trust…
Despite the popular adage about everything on the internet being there forever, every day pages of information and sometimes entire websites are lost to the sands of time. With the imminent shutdown of the DPReview website, nearly 25 years of reviews and specifications of cameras and related content are at risk of vanishing. Also lost…
Having DPReview around in the dizzying days of newer cheaper digital cams, and each one claiming superiority was a godsend. Nowhere else were people using sensitometry, resolution charts, and color charts to characterize each one. In a word DPReview was the Consumer Reports for nascent, burgeoning digital photographers. And they were beyond reproach.
Chugga chugga chugga goes the 12 bar space age (bachelor pad) blues. In the same way a pot of your granny’s soup comes to be more than the sum of its secretive parts, the far-out music bubbles and squelches and fizzes and farts in all the right places, all gnarly, knotted wood Fender fuzz bass […]
This is the track that Craig is attempting to compare Denim to in his essay regarding the influences and lifting done by Stereolab of sometimes obscure, and out of public view music producers. They do really.
Cliff is given a series of prompts by his buddy Norm at the bar (Cheers TV series)
Full credit to CogDogBlog for this writing prompt and his attempt to nail ChatGPT with a known quantity (Fluid Dynamics and Volcanology). Cliff Claven on the other hand just like most other attempts at prompting ChatGPT does escape the same level of scrutiny. But that doesn’t make the rejoinder, “…what else are you gonna do with it?!” any less pathetic. Considering the money/effort poured into ChatGPT by OpenAI. They spent millions and all we got was this lousy meme.
Tony Hirst brief take on the spectrum of cheating, but also how that might be harnessed to measure student performance and mastery. The question isn’t “if” but WHEN are the students going to escalate up to the higher/negative end of the spectrum. Diploma mill, contract cheating, are all very intentional acts. But there’s a long slope before that happens, and so given how broad the slope is, how does one measure learning? Applausee to Tony’s measured and reasonable response to the hub-bub surrounding ChatGPT, OpenAI and all the machine learning thingz.
2022 goodbye, hope that’s the last we have to say that.
No end of year review here trying to track-back and account for what it IS I actually DO here. No this is just a pole, a fence post, hammered into the ground. This is the fence line between last year and now. Things happened, some were successful. Others were not (to borrow from Sec. Don Rumsfeld). I know a lot of people who are retiring. I hope to do that too. Not close, not by a long shot. I hope to travel overseas (let’s hope!) and finally declare Covid travel restrictions null and void. That’s my wish for this year. Selfish as it is.
Carpet Bomberz Inc. 