527 karma · joined January 26, 2021
For a bare sphere, it is about 10 kg for plutonium and 50 kg for uranium.
The jewelry is the remaining 0.8% by mass, and it is split roughly equally between the natural and synthetic stones by mass, but with about 80% of the revenue going to the natural stones.
Here is a very good video showing how large poly-crystalline industrial diamonds are made in the USA: https://www.youtube.com/watch?v=6o5RprIJmfA
China has their own, slightly different flavor of this machine, the cubic press. These machines are manufactured in thousands and cost about half a million USD each. They are used to produce both industrial and jewelry quality diamonds: https://www.youtube.com/shorts/cED0TjwKUDM https://www.youtube.com/watch?v=5cnEVb7aPfM
The original machine was invented by a guy at General Electric: https://en.wikipedia.org/wiki/Tracy_Hall
1. Thought will be liberated from the confines of mortal bodies. 2. Technology will be able to create and modify things on the molecular level, limited only by imagination (and the laws of physics of course).
Spirits will be roaming the cloud, developing in ways unseen before, and interfacing with the "real process" through whatever shapes suit the purpose.
Of course, learning from the treasure of design solution accumulated through the billions of years of biological evolution will be a large step towards this. Extending this heritage will be the next step.
What the authors have developed, is transistors which could work at a very high temperature, but also work at extremely low temperature. That is quite rare.
The capacitance of the wires themselves is typically an order of magnitude greater than that of the storage capacitors. So when the memory is read, the wires are first precharged to some standard voltage. Then the desired row of storage capacitors is connected to the wires, and the charges from the storage capacitors spread onto the wires, changing their voltages very slightly. These voltage deviations from the standard value are amplified by the "sense amplifiers". The amplifiers are sort of like flip-flops. Once they start in a state which "tilts" slightly to "zero" or slightly to "one", they go all the way to the full magnitude zeros and ones. This not only amplifies the signal, but also automatically brings the voltages on the wires and the still connected to them capacitors to the full magnitude, thus "refreshing" the data. The row is disconnected, and the next read cycle can start for some other row.
In the video, an array of 4x5 capacitors and the associated with them switches was fabricated. The capacitors in the video are several hundred times larger (12400 fF) than typical capacitors in a 64 Kbit DRAM (about 50 fF). I assume this is done so that in the later episodes the author could implement the readout electronics outside of the chip.
This guy is not exactly a regular person. He is a pretty unique case of a talented semiconductor engineer who has a home lab for side hassles. It is not a low effort thing. He runs the equipment 24/7, scrubs all the surfaces in the lab daily to keep it clean.
Still, with the lab and all of the equipment already at hand, it cost him several weeks of work to produce this demonstration of transistors and capacitors, which kind of work, but are still long ways from a "completely complete" 20 bit DRAM chip.
Unfortunately it is simply too much work for one person to maintain a viable semiconductor fabrication process, even when it is done semi-professionally.
Much more impressive are the modifications to the microscope, transforming it into an improvised lithography machine, and the home made plasma etching machine, cobbled together from surplus components.
Of course, the whole thing, starting from the clean room, is extremely impressive -- Intel started their business in a much simpler facility.
https://www.youtube.com/watch?v=Rwbyl7ywfhk&list=PLLnAFJxOjz...
Hydrogen mixed with air or with oxygen produces an ear piercing supersonic detonation, exceedingly loud and unpleasant. Not recommended for demonstrations.
One of the larger episodes was in 2012 in Armenia, where thousands of balloons exploded during a meeting, injuring 154 people, of which 4 seriously (the video is of poor quality): https://www.youtube.com/watch?v=jWEm2sS7Dw8
A smaller, more recent episode in India: https://www.youtube.com/watch?v=FH5JwHeKnZo
Heat conductivity, on the other hand, is an order of magnitude higher for helium, compared to argon, because its atoms are moving faster due to their lower mass.
When the gas is used for cooling, heat conductivity is important because it determines the conductivity through the boundary layer near surface, where the velocity of the flow drops to zero at the surface itself, and all the heat transport is through conduction rather than advection.
Here is some detailed information about low cost units: https://github.com/kaiaai/awesome-2d-lidars/blob/main/README...
The source in the ASML machine produces something like 300-500W of light. With an Xray tube this would then require an electron beam with 50 MW of power. When focused into a microscopic dot on the target this would not work for any duration of time. Even if it did, the cooling and getting rid of unwanted wavelengths would have been very difficult.
A light bulb does not work because it is not hot enough. I suppose some kind of RF driven plasma could be hot enough, but considering that the source needs to be microscopic in size for focusing reasons, it is not clear how one could focus the RF energy on it without also ruining the hardware.
So, they use a microscopic plasma discharge which is heated by the focused laser. It "only" requires a few hundred kilowatts of electricity to power and cool the source itself.
Later, improved units based on the same principle became ubiquitous in Chinese robot vacuums [2]. Such LIDARs, and similarly looking more conventional time-of-flight units are sold for anywhere between $20-$200, depending on the details of the design.
[1] https://scholar.google.com/scholar?q=%22A+Low-Cost+Laser+Dis... [2] https://github.com/kaiaai/awesome-2d-lidars/blob/main/README...
Taalas of course builds base chips that are already closely tailored for a particular type of models. They aim to generate the final chips with the model weights baked into ROMs in two months after the weights become available. They hope that the hardware will be profitable for at least some customers, even if the model is only good enough for a year. Assuming they do get superior speed and energy efficiency, this may be a good idea.
However, [1] provides the following description: "Taalas’ density is also helped by an innovation which stores a 4-bit model parameter and does multiplication on a single transistor, Bajic said (he declined to give further details but confirmed that compute is still fully digital)."
[1] https://www.eetimes.com/taalas-specializes-to-extremes-for-e...
https://patents.google.com/patent/US20240092508A1/en
They have also included a way to disconnect the stud from the leg afterwards, such that the deck can be tidied up conveniently after the rocket had been removed. This is a neat idea -- the damage to the deck should very localized, and the rocket gets secured quickly and without putting human welders at risk.
For over a decade, "Mythic AI" was making accelerator chips with analog multipliers based on research by Laura Fick and coworkers. They raised $165M and produced actual hardware, but at the end of 2022 have almost gone bankrupt and since then there has been very little heard from them.
Much earlier, the legendary chip designers Federico Faggin and Carver Mead founded Synaptics with an idea to make neuromorphic chips which would be fast and power efficient by harnessing analog computation. Carver Mead published a book on that in 1989: "Analog VLSI and Neural Systems", but making working chips turned to be too hard, and Synaptics successfully pivoted to touchpads and later many other types of hardware.
Of course, the concept can be traced to an even older and still more legendary Frank Rosenblatt's "Perceptron" -- the original machine learning system from 1950s. It implemented the weights of the neural network as variable resistors that were adjusted by little motors during training. Multiplication was simply input voltage times conductivity of the resistor producing the current -- which is what all the newer system are also trying to use.
The data travels as the differential voltage in each of the twisted pairs, and is transmitted magnetically by the transformer to the secondary winding. The power is applied between different pairs, and in each pair appears as a common mode voltage. This is all stopped by the transformer, and in devices designed to support PoE, the PoE circuits tap the mid-point of the primary windings to access the supplied voltage.
So at a first glance, it seems that if 48 volts is applied between the twisted pairs to a non-PoE device, this voltage would simply be blocked by the transformer. But since there is a widespread concern about this, there must be more to the story -- maybe somebody who actually worked with these circuits can explain why this is more complicated than it seems at first?
Edit: Found an answer. It seems that at least some of the designs of non-PoE Ethernet jacks terminate the common mode signals to a common ground though 75 Ohm resistors. In this case, if the voltage were applied between the twisted pairs, the resistors would dissipate far too much power and would burn out. So there is definitely a concern with the dumb PoE injectors and at least some non-PoE devices. https://electronics.stackexchange.com/questions/459169/how-c...
So, many people, including Searle, wanted to push back on reading too much into what the program was doing. This was a completely justified reaction -- ELIZA simply lacked the complexity which is presumably required to implement anything resembling flexible understanding of conversation.
That was the setting. In his original (in)famous article, Searle started with a great question, which went something like: "What is required for a machine to understand anything?"
Unfortunately, instead of trying to sketch out what might be required for understanding, and what kinds of machines would have such facilities (which of course is very hard even now), he went into dazzling the readers with a "shocking" but a rather irrelevant story. This is how stage magicians operate -- they distract a member of the audience with some glaring nonsense, while stuffing their pockets with pigeons and handkerchiefs. That is what Searle did in his article -- "if a Turing Machine were implemented by a living person, the person would not understand a bit of the program that they were running! Oh my God! So shocking!" And yet this distracted just about everyone from the original question. Even now philosophers have two hundred different types of answers to Searle's article!
Although one could and should have explained that ELIZA could not "think" or "understand" -- which was Searle's original motivation, this of course doesn't imply any kind of fundamental principle that no machine could ever think or understand -- after all, many people agree that biological brains are extremely complex, but nevertheless governed by the ordinary physics "machines".
Searle himself was rather evasive regarding what exactly he wanted to say in this regard -- from what I understand, his position has evolved considerably over the years in response to criticism, but he avoided stating this clearly. In later years he was willing to admit that brains were machines, and that such machines could think and understand, but somehow he still believed that man-made computers could never implement a virtual brain.
https://openai.com/index/whisper/
Such approach dates back to 1940s, when people were trained to read the speech from spectrograms. There is a 1947 book "Visible Speech" by Potter, Kopp, and Green describing these experiments. Here is a more slightly recent 1988 review of the subject: "Formalizing Knowledge Used in Spectrogram Reading"
https://www.tek.com/en/products/oscilloscopes https://www.keysight.com/us/en/catalog/key-34771/infiniivisi...
"Sampling" oscilloscopes are a much less common product -- they are useful for analyzing signals that are too fast to digitize in the ordinary way. They typically sample at a very slow repetition rate -- some hundreds of kilohertz, but each sampling aperture can be exceptionally short, allowing to record signals to 100 GHz frequency.
I think some of the short range depth cameras (Kinect v2 was one) use time-of-flight technique, and could in principle be reconfigured to perform a similar demonstration, though it would not be as "homemade" and cool as the system built for the Youtube video.
For each laser pulse, one microsecond of the received signal was digitized with the sample rate of 2 billion samples per second, producing a vector of light intensity indexed by time.
A large number of vectors were stored, each tagged by the pixel XY coordinates which were read out from the mirror position encoders. In post-processing, this accumulated 3D block of numbers was sliced time-wise into 2D frames, making the sequence of frames for the clip.
But in principle, a LIDAR could be reconfigured for the purposes of such demonstration.
If one wants to build the circuit from scratch, then specifically for such applications there exist very inexpensive time-to-digital converter chips. For example, Texas Instruments TDC7200 costs just a few dollars and has time uncertainty of some tens of picoseconds.