iiinova — where your digital self keeps evolving

The best take I've seen so far.

030 3

Fundamentally, more hardware = faster research feedback loops = more innovation. More HW also lets them scale easily. Both absolutely critical in this AI arms race for those companies.

That's assuming there's an old shovel somewhere as an alternative to the super deluxe mega shovel. But there's not. Nvidia still wins even if you use their older hardware. This also assumes the big companies will slow buying because they can. I doubt that they will give up their lead in HW ever.

Totally agree, but we've seen that coming for awhile (as we saw the improvement rate slowing, and foundational models focusing on their UI improvements as announcements). I mean, 2 years ago current techniques would be seen as magical compared to what was going on then. This just seems normal?😀

I am having a hard time doing the mental gymnastics required for the "hurt" part of your statement, though. Megacorp AI isn't going to stop using their hardware just because someone did it cheaper? Better HW = faster research feedback loops. More loops = more innovation, fundamentally. So why?

000 0

FWIW and my purely hot speculative take - I doubt this will happen for a very long time. But who knows in this space. The only thing I do know is that for the last two years, alternatives to Nvidia based methods have existed, and none of them have gained much traction.

000 0

And these companies (FB, Google, Apple, Microsoft, OpenAI, Claude, etc) depend on being the latest/greatest state of the art. Which means ass tons of research. These companies have tons of cash. As soon as we see people do this without Nvidia cards the market will freak out for real.

100 1

Ok, I've spent an hour and a half reading the talking heads' interpretations of the Deepseek/Nvidia hype. So far, it is: "Deepseek did it cheaper, so buying will slow." 🙃 Erm, why will it slow? Foundation AI co's will not leave behind their HW edge - no matter what, more HW means faster research?

130 3

Yes, way less compute needed in this circumstance. But there are many examples of efficient training and models out there already. I do not see why this explanation is any reason for folks to believe the foundational co's will slow their buying. People can freak out when it's done without Nvidia.

000 0

I mean, most of us knew that there really is no moat other than scale and recurring customers, and all the niceties the large foundational companies' user interfaces provide (code artifacts, document parsing, code running, etc). If anything, the large co will just consume the learning and continue.

Continuing the thought, having more compute allows tighter feedback loops on your training experiments as researchers. Why would the large foundational model companies suddenly stop investing in that hardware? Deepseek's innovations mean they would just be able to train faster and iterate sooner?

100 1

Ok - 100% agree with this - as long as we're assuming it's truthful. I think we can at least agree they used less. But does that somehow signal to the megacorp AI leaders to stop using their cash on hardware? I don't think so. That would be insane. The only edge they really have is HW.

200 2

THANK YOU. Those of us on the ground have seen plenty of innovative training techniques for training, fine-tuning, and running models on commodity hardware over the last year and a half. Big AI companies slowing their innovation and feedback loops by slowing hardware purchasing would be insanity.

000 0

I definitely agree the US Media / Government is likely going to see it as USA vs China, but we all win with this :P Having _EVERYTHING_ fully open (data + training code + models, etc) means we all get in on it, and the US companies will be sure to consume all this learning as well.

000 0

Furthermore, about once a week I see a new fine-tuned or fully trained model claiming the top of the leaderboards. For the uninitiated, there are 1.3M AI models available on huggingface, built using all kinds of techniques (mostly using Nvidia hardware). So, again, how does Nvidia lose here???

I think the real response we're seeing in the news is that the US analysts are surprised that China is able to innovate within their restrictions. (?!) This is the same story time immemorial about every megacorp vs. scrappy upstart - resource constraints breed innovation.

Someone help explain to me how this hurts Nvidia? I'm not seeing it. In a gold rush where there is only one company not only selling the pickaxes, but the jeans, tents, and every other required tool to mine gold, how does someone using hardware more cleverly hurt the hardware manufacturer?

It does. Where are these conflicting reports? Haven't seen them. We also know Deepseek was using their 10,000 H100s they started with, and probably a lot more they can't talk about publicly.

Keith Hanson @keithhanson.bsky.social · 18/01/2025

No co-working space is complete without exposed pipes. I've never visited one without them 😅

000 0

Design rules in EasyEDA have my traces at 0.6mm and spacing between traces at 0.6mm. Chonky but it works 😅

000 0

My basic process is to export a Gerber from EasyEDA, then punch everything into pcb2gcodeGUI, then use Candle to heightmap and run the gcode.

100 1

This was milled with a 3018-prover, this bit at 0.045 cut depth and a tool diameter of 0.1, 2 extra passes: HUHAO 5pcs V Router Bits 20 Degree Engraving Bits 1/8 Shank 2 Flutes 0.1mm Tip Engraving Tool Bits CNC V-Groove Router Bit for Acrylic Wood MDF Stainless Steel a.co/d/6e1rzDV

a.co

Amazon.com

120 2

Keith Hanson @keithhanson.bsky.social · 25/12/2024

Only about 16 boards scrapped until I found a favored bit / tool diameter / cut depth, but the multimeter says it's good! On to solder masking and drilling! 😁

240 4

After measuring what felt like dozens of times, I punched it into OnShape and prayed to the FDM gods 😅 The tolerances are so close that all I have to do is mash the board in and it can't move. Thankfully, the board's imperfections (+-0.02mm) cause it to be a super snug fit - perfect for a jig!

000 0

Keith Hanson @keithhanson.bsky.social · 25/12/2024

You are awesome! 😍 😍 😍 We need one of you in every city! 😁

Keith Hanson @keithhanson.bsky.social · 25/12/2024

Having an additive and a subtractive method of creating your own prototype objects is amazing. Sometimes I think I was born at exactly the right time 😅 Made a jig for the 70x100mm (actually - 71 x 101mm‼️) PCB blanks I've been buying. Double sided tape was too squishy, and was causing problems!

250 5

I am embarrassed to admit how many times I left the sensor alligator clip on the drill and spun it up 😅😅😅

It has been a journey 😅 I learned 3D printing before CNC, so it has been fascinating to see where all that came from. You can see all the little things learned from CNC -> 3D printing controllers and programs and such. This feels like taking the bumpers off the gutters and bowling for real 😅

Closer! Testing out all the steps here before solder masking and milling that away. I did something wrong on the drill step 😬 but everything else came out pretty solid. My process so far is: EasyEDA -> grbr export -> pcb2gcode -> candle + height map -> hit send

140 4

Keith Hanson @keithhanson.bsky.social · 08/12/2024

Using usbip to allow my rPi (attached to 3018 prover) to forward USB connections to my desktop, and from there, Candle to drive the CNC. After trying several options, Candle's height mapping is amazing! I had up to .04mm variations in height that was absolutely wrecking my traces. Height map ftw!

050 5

Always wanted to try my hand at milling PCBs and milling away a layer of solder mask myself. Just got this setup using cncJS - my better half told me 12am was too late for any further testing 😅 Not saying I'm ready to call this success, but it is exciting 😁

040 4

Oh my! Popped the BIG magic smoke that day 😅😅😅

So good! I love how our brains just turns those dots into "it's looking at me" 😅

Love those blink animations!!! Clever!

220 2

3D printing solder to create PCBs o_0 I... Really want to turn my old ender into one of these now 😅

youtube.com

030 3

At least the antennas are still good! 🤣🤣🤣

Hahaha thanks 😅 I laughed pretty hard after I was like, "oh wait... Have I seen 5v with my own eyeballs? ... Oh shit. That says 15V 🙃". But this is why I soldered all those female headers 😅 I knew I'd either do something stupid like this on my first custom drone, or smash it into a tree at 100MPH!

Pro tip: verify the voltage out BEFORE you hook it up to your components 😬 Everything on that board just got rocked with 15v 😭 I've never seen every component pop the magic smoke all at once 🤣 That's what I get for assuming the V pin was of course 5v 🤣

4305 30

👀 also looking for the recording 😁

Plus, as @garymarcus.bsky.social said, there is no moat - a GPT is a GPT is a GPT. All the big players are feeling pretty similar if you use them all (convergence, as mentioned). So with such a high burn rate, such an enormous valuation, such large hype... That is certainly WeWork territory 😅😬

020 2

I only think this because the big five mentioned have more resources and money, and every motivation to be the first to reach these breakthroughs. My secret hope is that the breakthroughs keep happening in the open source 😁 Then "chat bots" aren't impressive, it's just a tool anyone can use/make.

000 0

Amazing commentary. I do think those big companies will likely innovate something either derivative of GPTs or some way to "hack" them for lack of a better term. That or they'll continue just making agents with fancy UI and call it GPT (duct taped to be useful), until there are a thousand of those.

120 2

Omg I almost forgot the Hamilton Funkos 😱

000 0

We can be friends if any of these speak to you. 😅

230 3