Cheese Ends, 20230918

Welcome to another addition of cheese ends.

 

(Picture from here.)

 

Water on the Moon

 

The moon has hematite.

 

What’s interesting about this is hematite is a form of rust—a product of iron interacting with oxygen and water. The moon is airless. How did it get there?

 

Chandrayaan-1 has a Moon Mineralogy Mapper instrument (adorably known as M3.) that detects the spectra that reflects off minerals. It detected the spectra of minerals on the poles of the Moon. In so doing, scientist Shuai Li found the signature of iron oxide in the form of hematite. Without air, hematite on the Moon is... unexpected. (Cue Spock raising his eyebrow.)

 

The model proposed by LI and others was that oxygen from Earth’s upper atmosphere was drawn out by the Earth’s trailing magnetic fields, the magnetotail. Then, when the orbit of the Moon intersected this area, oxygen was deposited, driving the formation of hematite. However, hydrogen from the same source was also deposited. Hydrogen inhibits hematite formation. It turns out the magnetotail inhibits hydrogen during certain phases of the Moons orbit, allowing oxygen to be deposited directly. 

 

But there also needs to be water. It’s considered that fast moving dust particles could dislodge water particles could release water deposited on the surface, freeing them to interact with the oxygen and the iron to form hematite. Or, possibly, the energy of the protons of the solar wind could have a role. More research was needed.

 

This was back in 2020.

 

In September 2023, the same team discovered water formation in the same magnetotail where the oxygen was coming from. The high energy electrons of the magnetotail allow formation of water molecules directly, without any contribution from the solar wind. When the Moon passes into the magnetotail, it is shielding from the solar wind but not from solar photons, giving an the magnetotail an opportunity to deposit water on the Moon’s surface.

 

Chandrayaan-3

 

The big news recently is India’s landing of a surface probe and rover on the south pol of the Moon. This is terrific for a few reasons. For one thing, human beings actually landed a probe on the south pole: something that has never been done. For another, this brings the number of countries that have successfully landed any working probe on the Moon’s surface to four: China, USA, Russia, and India.

 

Spreading the wealth is a good thing.

 

Chandrayaan-3 recently completed is 14-day mission—the length of a lunar day—and the data is still being digested. I looked for more information and found scraps so I’ll delay a deeper discussion until there is more data forthcoming.

 

China Expands its Deep Space Network

 

It takes big, sensitive radio telescopes to keep in contact with probes and human operations in space. Radio isn’t that great at carrying information compared to higher frequencies and the transmissions are limited by the equipment sent. 

 

China has been using its Very Long Baseline Interferometry (VLBI) network to both track its material and to investigate black holes, the dynamics of our galaxies, and the like. The VLBI technique tracks receipt of signals in time. In this way, a virtual telescope of the size of the separation between the telescopes can be achieved. 

 

I want to talk about this for a moment. Not the achievements of India and China directly, but the way we’re stovepiping the process. Each country has its own space process. There’s a fair amount of cooperation that can be helped or hindered by the country of origin. Russia is now cozying up to China and North Korea using its space program as the proverbial carrot. China is largely going its own way. India breaks both ways. Europe mostly cooperates with the USA but not always.

 

Science is based on complex competitive/cooperative relationships. Scientists and teams compete with one another to gain information and cooperate to achieve those goals at the same time. This process utterly depends on the free flow of information. Restrict that flow in the name of national prestige and science itself is restricted.

 

There are also economic issues involved. It took the cooperation of many countries to build the ISS and the Large Hadron Collider. When the USA tried to go it alone with the Superconducting Super Collider, we failed. The cost was too high. We couldn’t get cooperation from other countries because it was promoted as an American product.

 

There’s a limit to how much a single country—USA, China, or any other—can do. No single country is responsible for the gross domestic product of the world. The world GDP is estimated as 105 trillion dollars for 2023. The US component is about 27 trillion and China’s about 19 trillion. As technology becomes more evenly dispersed, that world GDP is going to increase and both the US and China relative contribution will lessen. The cost of the LHC  was about 5.1 billion dollars. That’s not even visible in the world GDP. 

 

True space exploration, the founding of settlements across the solar system, cannot be achieved by a single country. It’s only a way of exporting conflict and failure.

 

Okay. Enough soap boxing.

 

Planet of Iron

 

Gliese 367 is a red dwarf star about 31 light years from Earth. It has been named Ananuca. GL 367 b—or Tahay—is an exoplanet that is orbiting close to it. It has an orbital period of only 7.7 hours. It’s so close it’s tidally locked and its atmosphere has been stripped away. What’s left is a rock that only presents one scarred face to its sun at all times.

 

What’s interesting about it is that it is very dense—nearly twice as dense as Earth—as if it were essentially just an iron core orbiting its sun. 

 

Two hypotheses suggest themselves. It’s possible that Tahay was once a significantly larger planet with an appropriately larger iron core but some cataclysm ripped the outer covering off. Or, Tahay was born in a particularly iron rich section of the initial proto-planetary disk. Planetary origin models don’t handle either scenario all that well.

 

Summer of 2023 Hottest on Record

 

Yeah. Things are getting hotter. More and more energy is being driven into the land, sea, and air. More bigger and stronger storms, deeper droughts, heavier floods, happened. More are coming. When you put more and more energy into a system, that energy gets expressed. I can’t believe this is a surprise to anybody.

 

This is one of those things that have to be done together. Me: I’d like the luxury to send people elsewhere in the solar system rather than being force into it because we burnt our planet to the ground.

 

It amazes me this has become so political. It’s like standing in the crosswalk of a busy street arguing the existence of the car coming to run us down. 

 

State of the Farm: September, 2023. The Problem of Scale.

Be careful what you wish for.

 

(Yes, yes. Another State of the Farm entry. I’m in Wisconsin. Sue me.)

 

Some years ago, I discussed the problem of abundance. At the time, I was concerned with production from tree products such as chestnuts and Cornelian Cherries. That particular year we had good crops from several of our fruit trees and thought pretty good of ourselves. But, then, these were tree products that have their own issues and fecundity.

 

While we could do some things such as spraying and fertilizing, there wasn’t all that much we could do to help the trees. Weeding? Not an issue. Pests? Sometimes but for many years we haven’t had much of a problem. 

 

In the garden, however, we had much more control and, therefore, felt the failures much more keenly.

 

Over the last few years, we’ve started attempting to address the problems of producing calories at scale. By this, I mean that we are trying to make a dent in the sustenance of the household. Trying to take the garden to a fun project that is good in the summer to an effort that actually makes a dent in the grocery bill.

 

This year we put everything we learned to use and now it’s time to evaluate how we did.

 

We put our caloric faith in beans, squash, and potatoes. Other crops came and went and a couple—I’m looking at you, sugar beets—are still experimental. But those three are intended to feed us.

 

We learned a lot from beans but I can’t say we were a rousing success. We planted bush beans, pinto beans (which are a kind of bush bean), runner beans, and pole beans. We also had record-breaking heat waves and, as I said in a previous entry, found out that beans don’t set fruit in high temperatures. When the temperatures began to drop, we started to get beans and we’re letting them dry some on the vine before we determine the yield. I’m not hopeful.

 

In retrospect, I think I planted too densely. While the Great Bean Tunnel looks cool, it’s not really all that productive. It was, however, a nice place to sit in the heat.

 

We had similar problems with the squash. The zucchini never really set fruit—we’re still trying to find out why. Other people we know had tons of the stuff but we didn’t get much. We had a lot of squash germination problems that we’ll need to address. Germination problems all around. We do have some squash but, again, I don’t think we handled scale well there either.

 

Potatoes were a fair success story. We have about a 3/1 return on investment. Not as good as I would have preferred but significantly better than breaking even. We planted a white and a red potato. (I forget the varieties at the moment.) The reds did well but the whites didn’t. We got many tiny potatoes. I understand that this is in style at the moment but I would rather have had more literal poundage than pretty plate decorations.

 

Other crops did better. Amazingly, we were able to grow two cabbages—usually, they’re eaten to the bone. And we had some cold crops that did well. Basil did well and the sugar beets appear—without yet being harvested—to have done quite well. Melons are always questionable.

 

Interestingly, the melons and cucumbers both seemed to have something that looked like a squash borer problem. All of a sudden, a whole stalk of cucumber would die off. I never found the culprit. This was mostly because of the density of planting and the random nature of the melon/cucumber growth. Every stalk crossed every other one on the trellis. It looked nice but it made tracing back the problem difficult.

 

If I have learned any lesson this year, it was that I planted too densely. I wanted to grow everything. But, while it was cool to grow a couple of cabbages, each cabbage was a circle three feet in diameter and produced a single “fruit.” It was space I could have used for more staples. I think I need to limit my experiments and turn more space over to calorie crops.

 

I planted nine different kinds of melons very close together. When I got fruit,  I couldn’t tell where it was coming from. The runner beans and pole beans were also planted too densely. Both layered up to the point that some beans were shadowed by their neighbors and didn’t produce as well.

 

It was ambition, pure and simple. I blame society.

 

Many pickles were canned. Lots of tomatoes were sacrificed to the vegetable gods to form sauce. Much applesauce and dried apples were made. Ten gallons of Cornelian Cherry wine is percolating even as we speak. If you can hear me talking or are speaking to me right now, I’d suggest visiting a professional.

 

So: more planning next year. This year, raise a glass and pass the potatoes.

Revisiting "AI"

 

Picture from here.

 

I discussed a while back my opinion on the stampede to bring AI into the kitchen, living room, workplace, and bedroom that seems to be going on. 

 

My opinion isn’t all that different though I’m now a bit more informed. I believed I referred to ChatGPT (and others) as a stochastic parrot—that is, a device that is able to reliably generate convincing language but does not understand the meaning of the language it is processing. 

 

ChatGPT is not, exactly, such a device. It is a language prediction mechanism in the sense that given its broad training it can respond to a statement (or “prompt”) in a manner consistent with how its model of the language suggests is the desired response. This is quite a bit more sophisticated than a stochastic parrot. I am not saying large language models (LLMs) are conscious or have any understanding of the language they generate in the human sense of that word. I am saying they are extremely adept and sophisticated in mapping how language works.

 

This is not language at the sentence or word level—though ChatGPT maps that, too—but a much broader and deeper map.

 

Consider a regular visual map—like what we used to use on road trips when I was a child. It’s a piece of paper with coordinates on it and imagery. If you have the right x/y (east/west, north/south) coordinates, you can find a city. That’s a two-dimensional map. You can find a three-dimensional map of the solar system here. However, since you have to tell the map when you’re looking at the solar system, it is, in effect, a four-dimensional map. With those four coordinates, you can place a planet geographically.

 

However, to go forward with that metaphor, each of the planets influence the other and they’re all influenced by the sun. Light comes from the sun, for example. If we added that to the map, that might come in as a fifth dimension—a separate coordinate. You can see how this gets complex pretty quickly. While computers can work with multidimensional math easily, humans have more trouble. We want to actually understand such things—and we’re pretty much embedded in a four-dimensional brain. 

 

Physicists came up with the idea of phase space. This is a space where all physical quantities are represented as different dimensions and a given state is described as a coordinate set in those dimensions. It’s not hard to see the utility of such a description—and it’s not much of a leap that something like that level of complexity would be needed to describe all the intricacies of language. That’s not all of what goes on inside of ChatGPT but a complex multi-dimensional representation of language is part of it.

 

There’s a good description how LLMs work here. I’m not going to reproduce that. Timothy Lee and Sean Trott, the authors, did a much better job than I ever could. So, you should go read it. Wikipedia has a broad discussion of LLMs here.

 

In place of what I described as phase space, LLMs use a word vector. Word vectors mathematically represent a word in all its dimensions. Like the coordinate systems I was describing before, and like phase space, you can determine something approximating distance between words by referring to them within a complex coordinate system. If you look at Washington, DC, and Baltimore, MD, and get their x/y coordinates, it’s not a difficult mathematical operation to determine that their close to one another. Similarly, if you look at Baltimore and Boston, MA, you can tell Boston is farther away from Baltimore than Washington. Seattle is farther away yet. 

 

Using word vectors, you can develop maps of words that are close to one another in use. Or far away. Or build nets describing sets of words. Computers don’t balk at numbers of dimensions—LLMs use hundreds or thousands. GPT-3 uses over twelve thousand. I suspect the number of dimensions will only increase as more computing power is applied to training.

 

Ah, yes. Training.

 

One of the interesting things about LLMs is that they are not programmed per se. They analyze reams of material on their own—the training. Again, Lee and Trott discuss this in digestible detail. I’m not going to go into it here.

 

The next level of LLMs is the emergence of word predictions from the raw word vectors. Language has syntax and grammar—it is organized into meaningful packets. So is computer language code. When the LLM is trained on a language, that underlying structure is embedded in the dimensional relationships between the word vectors. LLM processing takes input and applies it to transformers. The input to a transformer results in a more meaningful result. That result can be further input into another transformer. The final result set is the predicted output from the input prompt. (Again, refer to Lee and Trott.)

 

It should come as no surprise to anyone that a powerful engine such as an LLM is trained on a great deal of language input that it would return sophisticated language as a result.

 

But there’s more to this—and this is the important bit.

 

GPT-3 recently (reported in July, 2023) that it had scored highly in tests of reasoning by analogy. Reasoning by analogy is something SAT and other college entry exams test. You’ve seen this. The test presents evidence (Given a letter sequence, abcde, eabcd, deabc, what is the rule? Or, what is the next entry in the sequence? That’s reasoning by analogy.)

 

Now, it’s interesting that GPT-3 failed in extracting the analogy from text but it did extract it from prompts. Is this evidence of analogic reasoning?

 

For my part—and the article I linked to above suggests the same—I think it indicates that reason by analogy is embedded in the language training of the LLM. By this, I’m suggesting that what we are seeing is not intelligence in the LLM but intelligence embedded in the training material as represented by the LLM.

 

I think this is borne out by the LLMs being used to determine protein folding. A team at Meta thought that though the “L” in LLMs was intended for language, it really just signifies a packet of data that can be analyzed. Reconfiguring the LLM to use protein chemistry as training input instead of language seemed a productive path.

 

It is, of course, far more complex than I have stated. My point is that the LLM produced plausible protein sequences similar to the way it produced plausible language constructs. Let’s remember that the capacity of the LLM to retain data is far greater than any human and its ability to analyze multi-dimensional space is far superior. By representing data as digestible material to the LLM, the inherent intelligence imbued in protein systems that are the product of a billion years of evolution is amenable to analysis. 

 

This does not mean the LLM understands protein chemistry anymore than it understands language—in the human sense. What it does analyze the underlying structure beneath the training material and, given a catalytic prompt, presents plausible combinations that conform to that underlying structure.

 

Given this point of view, LLMs become more interesting. They are not conscious. They do not comprehend what they are doing. But they are potentially powerful tools  to express the underlying sophistication and intelligence of the material they’re given.

 

That lack of comprehension shows up regularly—anyone using an LLM without oversight is just asking for trouble. AtomicBlender—a YouTube channel I follow—asked ChatGPT to design a nuclear reactor of the future and it gave surprising—if impractical—results.

 

I think that is going to be the main problem of using LLMs: the output must be verified. In the case of protein chemistry, a solution candidate would have to be verified experimentally and be assured it didn’t violate existing patents. LLM produced original material cannot be taken at face value.

 

LLMs are a powerful tool. The current situation reminds me of something from the Larry Niven, Protector: Intelligence is a tool that is not always used intelligently.