pkoird

I had a question related to diplomacy and border-protection. If say this plan goes through and say US reports a citizen from China to El Salvador for whatever reasons. What diplomatic consequences would there be between China and El Salvador? What if said Chinese national suffered harm or even died in El Salvador prisons, would China be allowed to hold El Salvador accountable? Or US?

I have always believed that the role of a bureaucracy is to make changes slow. In the far extreme, it leads to red-taping, corruption, and obscurity and presumably that's why one's first response would always be to advocate for more efficiency and cuts. However, there's a reason that virtually every bureaucracy in this planet is slow. Not one, not two, but all of them are not designed, rather devolve over time to such a state. The reason is that any change you make at such a level has tremendous consequences flowing downstream to every single level so you need time to plan, review, and document (for accountability) such changes. Plainly, that's not happening right now. I just wish that people who were doing this would stop and think for a while on the difference between the consequence of tanking a social media company versus tanking a government. I sure would be glad to be proved wrong and see it all work out though.

I remember the adage of the google of the yore, "don't be evil". It would be exciting to see what the community at large thinks on whether google upheld said principle and consequently, whether it could be believed to make further claims on "principles".

The answer is Perpetuation. Only those "structures" that learn to perpetuate can exist for a significant timeframe given the intrinsic randomness and the vast scale of universe. Be it a cell, or a cluster of cells, or an organism in its entirety, no matter how "intelligent" it is, if it does not figure out how to perpetuate, it will certainly die off due to entropy. So the most meaningful of all rewards, as far as it relates to AGI, is decidedly perpetuation* because any agent optimizing for otherwise will cease to exist.

*Note that improvement is also a perpetuation as it helps perpetuate in a changing world.

> "Petro's finding out that tweets have consequences. He's not (facing) a U.S. counterpart that looks at Colombia through a strategic lens, as a key ally, but as a country to make an example of," said Sergio Guzman, director of consultancy Colombia Risk Analysis.

Economic equilibriums are often hinged upon the fact that other players are "rational" and play "strategically". Of course, if you do forego such strategic principles (and rather impetuously), then you compel other rational players (enemies and allies alike) to rethink their stance on your rationality and seek other alternatives.

While I do consider Chatbots to be the low-hanging fruit of LLM research, I think the hype is partially justified by the fact that they weren't this good even 2-3 years ago. If you are willing to forgive their current status in face of the progress made, you should be able to see what they may hold in your five year future.

I bought a used Samsung Galaxy S21 Ultra phone and it's astonishing how future proof it was to begin with. Even now, the phone barely stutters, the battery life holds, the camera is surprisingly good and I was recently able to play GTA San Andreas (Netflix version) on it with surprising smoothness. I haven't tried but I believe the phone won't have any trouble emulating lot of consoles from 2010-2020 era as well. A really good purchase I must say.

So basically a before:2023 filter?

They say its different than neural networks in the text but honestly, it just sounds a lot like monte carlo dropouts.

Are you suggesting that two people in a shared inertial space when venture into other intertial frames but ultimately end up in a same frame later will agree on how much time has passed in between?

Because I lowekey believe this to be true but don't have means to prove it.

I don't think the recent H1B debacle spread outside the boundary of tech very much simply because a) There is now an increasing abundance of domestic tech workers, and b) The layoffs haven't been so severe outside as much as in tech.

Fields like Architecture, Agriculture, Forestry, Medicine, etc still need a lot of skilled workers that may not be available domestically. So I'm increasingly skeptical whenever someone claims that the solution to H1B is auctioning or a significant increase in wages as other fields outside of tech may not have the resource to bid for talents. US already has a diversity based immigration system called DV, maybe a diversity based system w.r.t. jobs is warranted?

Cool game. You should do a wordle like version with daily puzzles.

I'd love to read a semi-technical book on everything that we've learned about what works and what does not on LLMs.

Futures are a financial instrument that are much more volatile than Stocks (or even Options).

We already have a device for artificial reasoning, it's called Prolog.

Looks like a job for LLMs?

The difference you are pointing to is the acausal vs causal description. In Simulink (Causal), you have to describe what causes any changes in the system. For example, to model Ohm's law, you'd write something like I=V/R. Which translates to the current obtained when some potential difference is applied across a resistor. So I is a function of V in this respect and you can say V causes I in R.

In Acausal systems like modelica, it does not matter whether you write V=IR or any alternate formulation. You simply provide the description and it is up to the compiler to obtain the relevant set of equations that can be fed into the ODE solver. The advantage being that you spend less time fitting your problem into the causal framework required by Simulink or alternates where, when you change your model, you end up doing significant rewrites.

TL;DR: Causal descriptions take more time to arrange as compared to Acausal.

Obligatory mention to Openmodelica[1] which is an open source implementation based on the Modelica language. While I haven't used it yet, I was planning on exploring some of the features this holiday.

[1] https://openmodelica.org/

I'd argue for the converse: Computer Science is Mathematics

Ahh, so maybe we should first focus on formalizing Lean itself?

I'll say this again, any sufficiently advanced LLM is indistinguishable from Prolog.

I should think so. I forget the source but I seem to have once heard (to my satisfaction) that all global diplomatic maneuvers can essentially be explained in terms of kids playing in a playground.

Well, strictly speaking, it is suitable for places where all actors make choices in ways we can predict. So even if an actor was choosing, say, uniformly at random (and we knew this fact), we'd be able to incorporate it into analysis to get some expected behavior.

> Photographs of the night sky are compatible with all theories of the solar system in human history (including, eg., stars are angels). There is no summary of these photographs which gives information about the world over and above just summarising patterns in the night sky.

This is blatantly incorrect. Keep in mind that much of modern physics has been invented via observation. Kepler's law and ultimately the law of Gravitation and General Relativity came from these "photographs" of the night sky.

If you are talking about the fact that these theories only ever summarize what we see and maybe there's something else behind the scenes that's going on, then this becomes a different discussion.

KANs have learnable activations based on splines parameterized on few variables. You can specify a prior over those variables, effectively establishing a prior over your activation function.

Kolmogorov Arnold nets might have an answer for you!

I am not sure I understand. Sin(x) approaches x only when x approaches 0. When else does the universe does a bad job with this approximation?

Unpopular opinion but I really liked Microsoft Academic instead until they canned it, sadly.

Here's an idea, use a (portable) ultrasound emitter device that sends a pulse every set second and use an earpiece receiver that produces equivalent acoustics in audible range. The setup may reduce "tongue-fatigue' and the ultrasound will travel farther and can reflect off smaller objects. Would be an interesting experiment if nothing else.

Curious to how long the frozen structure can "survive". I wonder if it's a good idea to freeze one such frog and thaw it centuries later (an amphibian time-traveler!)