jbay808

Yes, exactly. And it's the compression that makes this happen. If it was just curve-fitting then the LLM really would act like a stochastic parrot, and it wouldn't generalize; also the model would also need to be much much bigger. As it is, the models are very big but they're still tiny compared to the dataset that they compress, so the compression process forces them to approximate a solution to the inverse problem.

And yeah, that means the base model training process is solving the inverse problem of finding an approximation for the processes that generate all human records and also anything else in the dataset with a compressible pattern to discover (weather data, etc). So it's not surprisng that some kind of world model emerges out of this.

Imagine you're curve-fitting a bunch of data points on, say, the orbital motion of planets and asteroids. You get tons and tons of data on these orbital motions, and then put it into a huge black-box optimization algorithm that compresses the heck out of it. It compresess and compresses until it can't find a more compact representation, no matter how much more effort it applies. The output is a function, where you can provide an initial condition, and it gives you the predicted future orbital position at any requested time.

Of course, one thing you get out of this is a great curve-fit for your existing data, which you can interpolate to find the position of any observed planet at any desired time.

But could this function also succeed in predicting the orbital motion of objects that aren't in the dataset? If I spot a new comet, and put it into the compressed function, would I get an accurate prediction of its orbital motion, even though that object wasn't in the training data?

The answer is "it depends, but probably yes". Newton's laws of orbital motion turn out to be simple compared to the size of the training data. So if the black-box compression has done a good job, it might output that function, or a close approximation of it. With a sufficient quantity of sufficiently accurate data, it might even improve on it; random errors can't be compressed, but where the deviations between observations and Newton's law turn out not to be random but rather the influence of an unobserved gravity source, or general relativity, the black-box algorithm will likely capture that as well.

A lot of people seem to think of the training process as curve-fitting data (the "stochastic parrot" model), but I think of it more as "solving an inverse problem to approximate the unknown source that generated the training data". Machine learning has proven to be quite good at solving inverse problems, and this is just a very abstract one of them.

(A forward-problem is something like solving for the electric fields from a set of charged particles; an inverse problem, https://en.wikipedia.org/wiki/Inverse_problem, is one where you have data on the electric fields at various points and want to reconstruct the arrangement of charged particles that produced it. Or more generally, you have sampled data on the output of an unknown process, and want to reconstruct the process that produced the data).

The inverse-problem-solving happens at the ML training step. The language model itself, that comes out of that, is solving the forward-problem: it has a generative-process baked in and now it's generating new data from it. But if the training process has done a good job of compression, it will certainly be able to generate valid new ideas that aren't in the training set, because the inverse model has solved for the underlying features of the real process that generated the training data, and those features can generate additional valid outputs that it wasn't trained on.

If the proof of concept takes an hour to code up, or proving the market exists just takes a bit of googling, then sure, you can prepare that before the first meeting where you suggest the idea.

If the proof of concept requires spending a few days in the machine shop making jigs and parts, purchasing equipment, and a custom PCB, then I really hope you'll bring it up for discussion beforehand in a meeting. Ten minutes of discussion with colleagues might be as useful as several iterations of prototyping. Not so that they'll shoot it down, but because someone might say "oh yeah, we have a spare mcguffin from last year's demo that you can use, should save you lots of time."

You must have very different kinds of meetings than I do. Unless you're going into that meeting with a rehearsed PowerPoint presentation, or there's a strict agenda that doesn't allow any time for exploration, I expect to hear imperfect-ideas-in-infancy. One of the reasons we have meetings is to allow collaboration to happen. It's a format for working together.

You can fire the shot and then patch the hole at the same time, proposing solutions to the same problem you pointed out, rather than just shooting and letting one person handle defense from every attack.

Definitely impressive as a proof of concept. A lot of the other problems can be solved with iteration. There are some IP67-rated drone motors meant to both fly and run underwater (available from Westmag for example).

There might be less that can be done about the underwater drag, but if it doesn't need to go long distances underwater that's not as much of a problem. For the RF signal, it can either run autonomously underwater, or use a fibre-optic umbilical, or even convert from an umbilical to wireless when it takes to the air.

You can do a lot with chaos. One of the things it lets you do is find an unforced trajectory from the vicinity of any state to the vicinity of any other (accessible) state. Sensitivity to initial conditions means sensitivity to perturbations, which also means sensitivity to small control inputs, and this can be leveraged to your advantage.

Multibody orbits are one such chaotic system, which means you can take advantage of that chaos to redirect your space probe from one orbit to another using virtually zero fuel, as NASA did with its ISEE-3 spacecraft.

You have to account for the energy required to break the bonds of the CH4, though. This means if you burn methane the usual way you get (CH4 + 2O2 --> CO2 + 2H2O + 803 kJ/mol); if you burn it with an ideal zero-emissions reaction, you get (CH4 + O2 --> C + 2H2O + 409 kJ/mol), or just a little more than half the energy from the same gas.

Your accounting works if someone else does the pyrolysis for you and you're left with just the H2 and C at the end, but mine includes the energy consumed by the pyrolysis step that breaks the methane molecule (albeit neglecting any thermodynamic losses, which there will be several -- for example you need to recapture the heat carried away by the hot carbon atoms). On the other hand, you can hardly wish for a better feedstock for CVD diamond production...

Interesting to see this on the HN front page. On the subject of methane pyrolysis, it turns out if you look at the Gibbs free energy calculation, about half of the energy of methane combustion is released from the formation of water, and the other half from the formation of carbon dioxide. That suggests that if you can be efficient with conserving the heat of pyrolysis, you can make a methane power plant that starts with a pyrolysis step to separate out the carbon atoms in an oxygen-free environment, and then burn the remaining hydrogen to power the cycle, and the end result would be a zero-emissions natural gas power plant. It would require twice as much gas to run, but if you can find a good value-added use for the carbon, it could potentially still be cost effective.

This would probably be much more efficient than doing pyrolysis to extract the hydrogen for use in electricity generation somewhere else, because you don't lose the substantial stored heat energy in the process of cooling that hydrogen back down.

And I can't help but wonder if fossil fuel companies might suddenly start endorsing aggressive zero-emissions targets if there's a way for this to double the demand for their products, rather than eliminating it.

It's mainly the laser itself that is the expensive part. If you only care about resolution it's easy, you just need a single-mode laser. But if you care about accuracy it's very difficult, because then the wavelength needs to be stable, and that requires a much more expensive laser. Most people looking for an interferometer are interested in accuracy, unless they're just measuring vibrations.

Thanks, it should be fixed now!

This one still keeps me up at night, especially the figure on the 6th page.

https://web.archive.org/web/20180513182952/http://burro.case...

The short summary of this hypothesis is that the ocean develops hypoxic zones, anaerobic bacteria boom, and eventually the ocean starts releasing masses of poisonous H2S gas that wipes out most life on land (and strips the ozone layer for good measure).

They speculate that this might have been a mechanism behind the "great dying" at the end of the Permian. I'm sure the thinking has advanced in the last 20 years, but whenever people ask what the worst-case scenario for global warming could be, my mind drifts back to this.

I disagree with the assertion that "VLMs don't actually see - they rely on memorized knowledge instead of visual analysis". If that were really true, there's no way they would have scored as high as 17%. I think what this shows is that they over-weight their prior knowledge, or equivalently, they don't put enough weight on the possibility that they are being given a trick question. They are clearly biased, but they do see.

But I think it's not very different from what people do. If directly asked to count how many legs a lion has, we're alert to it being a trick question so we'll actually do the work of counting, but if that image were instead just displayed in an advertisement on the side of a bus, I doubt most people would even notice that there was anything unusual about the lion. That doesn't mean that humans don't actually see, it just means that we incorporate our priors as part of visual processing.

If I were given five seconds to glance at the picture of a lion and then asked if there was anything unusual about it, I doubt I would notice that it had a fifth leg.

If I were asked to count the number of legs, I would notice right away of course, but that's mainly because it would alert me to the fact that I'm in a psychology experiment, and so the number of legs is almost certainly not the usual four. Even then, I'd still have to look twice to make sure I hadn't miscounted the first time.

This might be a great alternative to beryllium copper for the spring contact element in high-current electrical connectors.

Can you say more precisely what you mean?

I mean that needing to scan the full context of tokens before the nth is inherent to the problem of sorting. Transformers do scan that input, which is good; it's not surprising that they're up to the task. But pairwise numeral correlations will not do the job.

As for avoiding certain cases, that could be done to some extent. But remember that the untrained transformer has no preconception of numbers or ordering (it doesn't use the hardware ALU or integer data type) so there has to be enough data in the training set to learn 0<1<2<3<4<5<6, etc.

> I don’t really understand what you’re testing for?

For this hypothesis: The intelligence illusion is in the mind of the user and not in the LLM itself.

And yes, the notion was provided by the training data. It indeed had to learn that notion from the data, rather than parrot memorized lists or excerpts from the training set, because the problem space is too vast and the training set too small to brute force it.

The output lists were sorted in ascending order, the same way that I generated them for the training data. The sortedness is directly verifiable without me reading between the lines to infer something that isn't really there.

It might seem like you could sort with just pairwise correlations, but on closer analysis, you cannot. Generating the next correct token requires correctly weighing the entire context window.

I was interested in this question so I trained NanoGPT from scratch to sort lists of random numbers. It didn't take long to succeed with arbitrary reliability, even given only an infinitesimal fraction of the space of random and sorted lists as training data. Since I can evaluate the correctness of a sort arbitrarily, I could be certain that I wasn't projecting my own beliefs onto its response, and reading more into the output than was actually there.

That settled this question for me.

Yes, this is how frequency-doubled lasers work (eg. 532 nm green laser pointers, which are generated as a harmonic from a 1064 nm Nd:YVO4 laser by a nonlinear KTP crystal).

That's a great question! I have no idea. At low frequencies it should be very easy, because the sound is just a pressure measurement, so you can compare against a calibrated pressure reference. So the main challenge is measuring the high-frequency amplitude and phase response. If I had to do this, I'd probably set up a speaker in a long box with standing-wave resonance modes, and put both the microphone-under-test and a laser interferometer at an antinode to measure the change in refractive index that occurs with air pressure. A photodiode should have a flat frequency response out to well beyond 20 kHz, so that would do well as a calibration. But this is probably overkill for audible frequencies.

This project might be exactly what you're looking for:

https://github.com/peterdsharpe/AeroSandbox

This is correct, although in my experience the distribution is often more like a lognormal than a Gaussian: https://thesearesystems.substack.com/p/task-estimation-conqu...

I'd expect Chebyshev polynomials to be much faster and easier to work with than splines, certainly, and probably Fourier series as well. (Especially if there aren't trig instructions in hardware, because then each sine or cosine is itself a chebyshev polynomial to evaluate).

One reason you can get away with this sometimes is that ADCs themselves act like a low-pass RC filter. The ADC input is itself a capacitor that gets charged to the voltage that needs to be measured, and it has an appreciable resistance (which can be supplemented with an additional input resistor). Sometimes that's all the low-pass filtering you need to prevent aliasing, and the rest can often be done in the digital domain.

But if the signal you're measuring is very weak (or poorly matched to the ADC range), you probably need some kind of amplifier for it anyway, and in that case you may as well make a filter out of it at the same time to maximize the signal-to-noise ratio.