A hub motor for each leg adds some weight, but you get a rolling mode on flat terrain with great range.
At the same time you keep obstacle climbing mode with stationary wheels, and can use the leg actuators as shock absorbers and obstacle avoidance on a bumpy path. Best of both worlds, with a new software challenge.
The distinction is rather that the device was made next door, from raw materials, with them possibly watching. Certainly acknowledging the craftsmanship, but still while understanding your tools.
Today, labs are filled with expensive machines and you are not able to peek inside.
You need something? Only from a catalog, made in a mystical factory, without you knowing what's inside.
This abstraction speeds up your process (the tool you bought is fully qualified for what you plan to do) but also detaches you from the low level inner workings. Kids are fascinated if you take an everyday object apart with them (but maybe only if it was already broken)
Taking apart modern day “advanced” objects is way less fulfilling than those from the past. Miniaturization/integration and the increased complexity puts things beyond what even an adult can easily and visually grasp, let alone a kid. It’s mostly “black box” chips now.
Those advanced lab tools were built by many very skilled people. We’re past the time where a single person could hold it all in their head. One man could build a modern lathe back when they were a new thing, one man can’t build a modern lithography machine today.
I wanted to say that lathes have been around for over 2000 years and for most of that time a single well trained man could build one. In fact I’d wager that most lathes and indeed maybe even every part of that lathe up until the industrial revolution were built by a person working alone.
The advanced machines of today are no longer within reach for almost any single well trained man. Now it takes an army of people with non-overlapping skills and knowledge.
200 years ago a basic component was wire or a metal sheet, you’d get them “pre made”. Today the basic component is a chip with millions of transistors and thousands of lines of firmware code that do 99% of the job.
At least you can understand every part of a pencil even without having one in front of you, what it does, and a bit of experimentation would get you an absolutely terrible but technically functioning pencil. Make the tiny leap to a pen and you already lost everyone.
Could build a lathe from parts they could purchase, but not from scratch. The simple ballpoint pen is an absolute marvel of precision engineering that we take utterly for granted. That ball in the tip? Not something you could knock out in a weekend by hand. Even wire itself as a basic component is freaking impossible to feasibly do from scratch compared to going down to the hardware store and spending $5 for a spool. That I can also get a microcontroller devkit that's impossible for me to recreate from a local store is mind-blowing, if we stop to think about it.
2000 years ago the person who was capable enough to think of and build a lathe didn’t go to order the pieces. They modeled the raw materials (wood, metal, stone) into what they needed. But let me give you a less debatable example. Watch makers or even the people who built the Antikythera mechanism didn’t go to the cog and sprocket store to get the parts. They manufactured the components from the raw material or as close to it as possible. They probably built their own tools to for the longest time.
And when I said pen I meant fountain pen. It’s counterintuitive but probably more people today can draw a conceptual diagram of a ballpoint than a fountain pen. Ballpoints are a very simple design limited only by the very tight tolerances need to make them work.
But that was my whole point, complexity got so high that most things around us cannot be built without specialized tools, specialized building blocks, and specialized knowledge that takes years to gather. And this needs to be repeated for almost every object. Different tools, building blocks, and knowledge.
A highschooler today knows more chemistry than Mendeleev but that doesn’t help because the complexity in each field or product increases faster than one person’s ability to keep up with learning.
The linked article is a high level announcement that new sensors were installed, but no experiments yet.
I would have loved if they shared some technical details, e. g. which sensors, how does it record, frequency range and sample rate, time synchronization, offline capabilities. Maybe even a photo of the sensor? (I couldn't find any)
I'm surprised there are no security researchers that would pick up on this.
Take the same prompt and all incoming mails and run again through various existing models, even the simpler local ones. He now has a serious cross section of prompt injection ideas. This is a publication I would like to read!
For privacy reasons I understand the corpus might not get published. But for a research collaboration and safeguards (don't send automatic answers from each model you try)... why not?
I guess the decoder is more than the 208 bytes that this page uses..
But maybe you can misuse this and store a session ID / cookie in a favicon (give everyone a unique one) and survive some cookie cleanup and evade privacy restrictions?
Maybe you can still make it that the favicon looks like an image a little to not raise suspicion?
Favicons seem to be cached across private browsing sessions. Oh no
This post spends a lot of attention on how to fasten, glue and adapt a specific existing keyboard and monitor from a 1U rack mounted console. If I would need a similar serial console, it will end up differently - just because my parts are different.
> Tindie has not paid me for the last month. They take your money for your order and then require me to ship product to you, but they do not pay me. Temporarily, please place your order through the tattlersolutions.com website. See below
If you were to take 500 computers with older 1080 GPUs, you might have enough compute/ram equivalent to an H200 GPU for training such a model. Maybe take 10000.
But if those machines are spread over 10000 homes, wired with residential internet service, training a large model will not get anywhere.
You go from "data in the same HBM memory chip" at 4.8TB/s or "data in adjacent GPU" with NVlink at 1.2 TB/s down to 25 MBit/s upload speed. Accessing the next piece of data is going to be about a Million times slower.
At the same time you will heat a thousand times more, for a Million times longer.
You need to train independently and merge rarely. The problem is the merge step. Weights are too entangled, you are not going to get an improvement commensurate to the effort. Otherwise, everyone would do it. It is an open research problem.
You can switch a motor without permanent magnets to "idle mode".
I understand in Tesla dual motor configurations, the front motor is without magnets. The excitation field will be turned on when you need extra power, but at crusing speed it does not cause extra "drag".
From one teardown I've seen, they even went so far to use cheaper and less efficient IGBTs for the front drive, and more efficient SiC Mosfets for the rear motor (in the same vehicle!). If you need extra acceleration briefly, lower efficiency can be accepted.
Well.. The automatic part comes from the camera directing the settings mostly.
The lens would be motorized focus/aperture.
For motion picture cinematography, I've seen remote controlled focus anyway. I don't see why you could not have a good motor built I to the lens and remote control it. If the external motor focus is quick and precise enough, then the internal motors should be as well.
A hub motor for each leg adds some weight, but you get a rolling mode on flat terrain with great range.
At the same time you keep obstacle climbing mode with stationary wheels, and can use the leg actuators as shock absorbers and obstacle avoidance on a bumpy path. Best of both worlds, with a new software challenge.
reply