All Things Engineering. Electrical, Mechanical, Software, Firmware, AI, Security and everything in between. Specialize in custom HW/FW/SW for motor control
Everyone is talking about model distillation but "software distillation" will be just as profound.
Give an AI access to existing (likely commercial, closed source) software and tell it to clone the functionality.
Entirely autonomously, it will use the software, explore the features, build a dataset and then write you your own version that is a near identical clone.
@bobollini@LainShep@LewisMenelaws Driver support is actually better than windows by far.
App support is getting close. Basically all AI development is already done on Linux now.
@davidliuxyz Yes, as long as you also "just build it"
One vastly underappreciated feature of ordering from
China is that they don't ask for a bunch of drawing/tolerance clarifications, they just make it, knowing you will change it on the next order if anything is wrong.
I used to love a web tool called StumbleUpon. Added a button to your browser that would take you to random blogs and websites. Found so much fascinating stuff that way, and it was the antithesis of todayโs winner-take-most centralised web.
@dioscuri It was fantastic until it got big and then the monetization push crushed it, every time you clicked the button it became a paid page instead of something cool and organic
@SawyerMerritt Q: If I give you a penny on day 1, and double it every day, how much do you have on day 30?
A: The entire country with Robotaxi
Keep doubling it!
@gfodor Windows running on a Linux kernel gets closer every day.
The bushiness case for maintaining a custom windows kernel is evaporating, especially with NVIDIA N1X announcements.
https://t.co/QyOv7no9Yi
@LewisMenelaws I have predicted for several years now, Windows is going to switch to a Linux kernel at some point soon.
Its the only path that makes sense and it would save MSFT a huge amount of development.
MICROSOFT SAID IT ADDED MORE DATA CENTER CAPACITY IN THE LAST 18 MONTHS THAN IN AZURE'S FIRST DECADE, PREVIEWED ITS COBALT 200 CHIP, EXPANDED WINDOWS AI APIS TO MORE PCS, AND HIGHLIGHTED DEEP AI COLLABORATION WITH NVIDIA, QUALCOMM, MEDIATEK, AND ITS OWN IN-HOUSE CHIP DEVELOPMENT.
@jon_stokes Nope. Still have a long way to go. As context and capability continue to increase with compute expansion, there will be many more steps.
Software will be 100% AI written in a few years.
@Jason The real question is why on earth Google didn't do this first.
This is what Gemini should have been explicitly trained to do in Google workspace.