Not sure why petrochemicals and trees ie hydrocarbons are any more or less absurd than the silicon, metals, etc quarried and mined from around the world needed to store information digitally in data centers (or mobile devices).
Storing data of any kind in plastic as opposed to silicon metal seems like a meaningless distinction that only comes about from imagining that there is some disembodied, ethereal and platonic notion of digital “data” which is decoupled from any physical substrate. everything is always materialized and mediated through some complex, and probably vaguely arcane, geologically extractive process in some way.
Yep, PassiveHouse standards which typically include an extremely tight envelope which forces installation of outdoor air supply famously can get away with just a few hundred watts of heating capacity because of heat exchange on the incoming and outgoing airstreams!
Sure I was playing a little fast and loose there, but (a) the large surface area of the home (and resulting conductive transfer through the walls + convection transfer via infiltration through gaps) is directly a result of the fact that you need a significantly larger volume for humans to move around in and live in than you do to store food and (b) even if we do look directly at the volume of air, the difference is significant since at the end of the day, since for any given constant deltaT, your energy spent is still linear with mass or volume. And we are talking about roughly 2-3 orders of magnitude difference in air volume between a house and a refrigerator.
Anyways, if you write out all of the heat balance equations, you get a few W/m2 of flux on the inside wall of the home and a few W/m2 of flux on the inside faces of the fridge, assuming a typical wood frame construction in summer time and steady states all around.
So yes, of course multiplying the flux through the home’s wall by the surface area of the home results in a massive heat gain value compared to the heat gain conducted through the surface of the refrigerator, but that’s arguably precisely because of the two different volume requirements.
FWIW, in a typical apartment or single-family home, refrigeration uses a fraction of the energy that space cooling (also via a refrigeration/vapor compression cycle) requires on a warm day (and probably year round too unless in very mild climates). The psychrometric chart path is different so there are of course differences in the amount of energy required for the sensible and latent components, but the real difference is just the volume of air that needs to be dealt with.
My point being that at least from an energy and carbon perspective, lowering the space cooling demand via more effective building envelopes or increasing the space cooling supply efficiency - eg via membrane or dessicant dehumidification, better heat pumps etc) is far more impactful on a macro scale than better refrigeration.
Granted refrigeration in a warehouse eg is really also space cooling, but I’m just making the distinction between the dT=0-25F context and the dT>25F context. If I could only choose one technology to arrive at scale to improve the efficiency, it would be for the former context.
> I'm particularly annoyed by using LLMs to evaluate the output of LLMs.
Even though I largely agree with parts of what you wrote, if you squint your eyes enough you can kind of see an argument along the lines of “difficult to solve but easy to verify.”
Hey out of curiosity were there any issues with my top level comment? Seemed pretty innocuous, curious what the problem was. Feel free to email me if it’s better suited for discussion outside of post context.
> That "problem" remains unsolved because it's actually a fundamental aspect of reality. There is no natural separation between code and data. They are the same thing.
Sorry to perhaps diverge into looser analogy from your excellent, focused technical unpacking of that statement, but I think another potentially interesting thread of it would be the proof of Godel’s Incompleteness Theorem, in as much as the Godel Sentence can be - kind of - thought of as an injection attack by blurring the boundaries between expressive instruction sets (code) and the medium which carries them (which can itself become data). In other words, an escape sequence attack leverages the fact that the malicious text is operated on by a program (and hijacks the program) which is itself also encoded in the same syntactic form as the attacking text, and similarly, the Godel sentence leverages the fact that the thing which it operates on and speaks about is itself also something which can operate and speak… so to speak. Or in other words, when the data becomes code, you have a problem (or if the code can be data, you have a problem), and in the Godel Sentence, that is exactly what happens.
Hopefully that made some sense… it’s been 10 years since undergrad model theory and logic proofs…
Oh, and I guess my point in raising this was just to illustrate that it really is a pretty fundamental, deep problem of formal systems more generally that you are highlighting.
Yeah, I think I’m with you if you ultimately mean to say something like this:
“the labels are meaningless… we just have collections of complex systems that demonstrate various behaviors and properties, some in common with other systems, some behaviors that are unique to that system, sometimes through common mechanistic explanations with other systems, sometimes through wildly different mechanistic explanations, but regardless they seem to demonstrate x/y/z, and it’s useful to ask, why, how, and what the implications are of it appearing to demonstrating those properties, with both an eye towards viewing it independently of its mechanism and in light of its mechanism.”
I don’t mean to amplify a false understanding at all. I probably did not articulate myself well enough, so I’ll try again.
I think it is inevitable that some - many - people will come to the conclusion that these systems have “ethics”, “morals,” etc, even if I or you personally do not think they do. Given that many people may come to that conclusion though, regardless of if the systems do or do not “actually” have such properties, I think it is useful and even necessary to ask questions like the following: “if someone engages with this system, and comes to the conclusion that it has ethics, what sort of ethics will they be likely to believe the system has? If they come to the conclusion that it has ‘world views,’ what ‘world views’ are they likely to conclude the system has, even if other people think it’s nonsensical to say it has world views?”
> The fact that a human image comes out of a mirror has no relation what so ever with the mirror's physical attributes and functional properties. It has to do just with the fact that a man is standing in front of it.
Surely this is not quite accurate - the material properties - surface roughness, reflectivity, geometry, etc - all influence the appearance of a perceptible image of a person. Look at yourself in a dirty mirror, a new mirror, a shattered mirror, a funhouse distortion mirror, a puddle of water, a window… all of these produce different images of a person with different attendant phenomenological experiences of the person seeing their reflection. To take that a step further - the entire practice of portrait photography is predicated on the idea that the collision of different technical systems with the real world can produce different semantic experiences, and it’s the photographer’s role to tune and guide the system to produce some sort of contingent affect on the person viewing the photograph at some point in the future. No, there is no “real” person in the photograph, and yet, that photograph can still convey something of person-ness, emotion, memory, etc etc. This contingent intersection of optics, chemical reactions, lighting, posture, etc all have the capacity to transmit something through time and space to another person. It’s not just a meaningless arrangement of chemical structures on paper.
> Stop feeding the LLM with data artifacts of human thought and will imediatelly stop reflecting back anything resembling a human.
But, we are feeding it with such data artifacts and will likely continue to do so for a while, and so it seems reasonable to ask what it is “reflecting” back…
So the author’s core view is ultimately a Searle-like view: a computational, functional, syntactic rules based system cannot reproduce a mind. Plenty of people will agree, plenty of people will disagree, and the answer is probably unknowable and just comes down to whatever axioms you subscribe to in re: consciousness.
The author largely takes the view that it is more productive for us to ignore any anthropomorphic representations and focus on the more concrete, material, technical systems - I’m with them there… but only to a point. The flip side of all this is of course the idea that there is still something emergent, unplanned, and mind-like. So even if it is a stochastic system following rules, clearly the rules are complex enough (to the tune of billions of operations, with signals propagating through some sort of resonant structure, if you take a more filter impulse response like view of a sequential matmuls) to result in emergent properties. Even if we (people interested in LLMs with at least some level of knowledge of ML mathematics and systems) “know better” than to believe these systems to possess morals, ethics, feelings, personalities, etc, the vast majority of people do not have any access to meaningful understanding of the mathematical, functional representation of an LLM and will not take that view, and for all intents and purposes the systems will at least seem to have those anthropomorphic properties, and so it seems like it is in fact useful to ask questions from that lens as well.
In other words, just as it’s useful to analyze and study these things as the purely technical systems they ultimately are, it is also, probably, useful to analyze them from the qualitative, ephemeral, experiential perspective that most people engage with them from, no?
On the other hand, IP addresses have crossed into the popular lexicon in exactly this manner… it’s common enough to hear people say “what’s my “ip?” or “are there any free ips?” or what are the IPs for x/y/z”.
I agree that it sounds stupid and incorrect, but that doesn’t necessarily mean using MCP as a metonym for MCP server.
As someone who wrote my first line of code in approx 2010 and used git & GH for the first time in… 2013? it kind of amazes me to remember that Git is only 20 years old. GitHub for instance doesn’t seem surprising to me that is <20 years old, but `git` not existing before 2005 somehow always feels shocking to me. Obviously there were other alternatives (to some extent) for version control, but git just has the feeling of a tool that is timeless and so ingrained in the culture that it is hard to imagine (for me) the idea of people being software developers in the post-mainframe age without it. It feels like something that would have been born in the same era as Vim, SSH, etc (ie early 90s). This is obviously just because from the perspective of my programming consciousness beginning, it was so mature and entrenched already, but still.
I’ve never used other source control options besides git, and I sometimes wonder if I ever will!
I’ve been using Hatchet since the summer, and really do love it over celery. I’ve been using Hatchet for academic research experiments with embarrassingly parallel tasks - ie thousands of simultaneous tasks just with different inputs, each CPU bound and on the order of 10s-2min, totaling in the millions of tasks per experiment - and it’s been going great. I think the team is putting together a very promising product. Switching from a roll-my-own SQS+AWS batch system to Hatchet has made my research life so much better. Though part of that also probably comes from the forced improvements you get when re-designing a system a second time.
Although there was support for pydantic validation in v0, now that the v1 SDK has arrived, I would definitely say that the #1 distinguishing feature (at least from a dx perspective) for anyone thinking of switching from Celery or working on a greenfield project is the type safety that comes with the first class pydantic support in v1. That is a huge boon in my opinion.
Another big boon for me was that the combo of both Python and Typescript SDKs - being able to integrate things into frontend demos without having to set up a separate Python api is great.
There are a couple rough edges around asyncio/single worker concurrency IMO - for instance, choosing between 100 workers each with capacity for 8 concurrent task runs vs 800 workers each with capacity for 1 concurrent task run. In Celery it’s a little bit easier to launch a worker node which uses separate processes to handle its concurrent tasks, whereas right now with Hatchet, that’s not possible as far as I am aware, due to how asyncio is used to handle the concurrent task runs which a single worker may be processing. If most of your work is IO bound or already asyncio friendly, this does not really affect you and you can safely use eg a worker with 8x task run capacity, but if you are CPU bound there might be some cases where you would prefer the full process isolation and feel more assured that you are maximally utilizing all your compute in a given node, and right now the best way to do that is only through horizontal scaling or 1x task workers I think. Generally, if you do not have a great mental model already of how Python handles asyncio, threads, pools, etc, the right way to think about this stuff can be a little confusing IMO, but the docs on this from Hatchet have improved. In the future though, I’d love to see an option to launch a Python worker with capacity for multiple simultaneous task runs in separate processes, even if it’s just a thin wrapper around launching separate workers under the hood.
There are also a couple of rough edges in the dashboard right now, but the team has been fixing them, and coming from celery/flower or SQS, it’s already such an improved dashboard/monitoring experience that I can’t complain!
It’s hard to describe, but there is just something fun about working with Hatchet for me, compared to Celery or my previous SQS system. Almost all of the design decision just align with what I would desire, and feel natural.
There are discussions about lowering the seams (harder to generate spin and makes the same spin rates less aerodynamically effective) as well as lowering the mound.
There’s some consensus though that currently, pitching has evolved much faster than batting due to advances like Trackman and deeper understanding of the relationship between biomechanics, pitch tunneling, spinrate/flight path/movement, and so on. In conjunction with that has been a shift towards “TTO” (three true outcomes - HR/BB/K) on the offensive side, which is a statistically motivated perspective that batting for average is suboptimal. In short, you would rather have a lower BA and a higher home run rate even if it means a higher K rate, since home runs (and 2Bs) are so significantly more valuable than singles, and fly outs are also much more valuable than ground outs (or really, less bad) due to the opportunities for sac flies and the risk of double plays. TTO tho is also partly a response to the elevated pitching capabilities - velocity and spin.
This is all just to say that batters are falling behind and there’s an argument that it hurts the on-field product from an entertainment perspective since balls in play are what we ultimately watch for - if torpedo bats make it more likely that players can bat for higher averages by barreling up the ball more consistently, it will be good for the game.
Other alternative proposals include lowering the mound (famously done in the 60s), adjusting the ball (eg lower seams, which makes it harder for pitchers to generate spin and makes the same spin rates less effective), and so on.
One good (bad?) thing is that to some extent pitchers are starting to reach a biomechanical wall, evidenced by the greatly increased rates of Tommy John surgery, though that is partly also an effect of better surgical techniques and recovery times.
The biomechanics involved are insane. You are hitting a baseball-sized object (ha) moving at 90+ mph with massive break, often very late and over two axes, with something a couple of inches in diameter, and need to make decisions and react and adjust your swing path in a handful of milliseconds. And that’s just to make contact, let alone good contact, let alone contact that can find a patch of grass.
It’s the single hardest skill in competitive team sports in my opinion.
> Know what the difference between hitting .250 and .300 is? It's 25 hits. 25 hits in 500 at bats is 50 points, okay? There's 6 months in a season, that's about 25 weeks. That means if you get just one extra flare a week - just one - a gorp... you get a groundball, you get a groundball with eyes... you get a dying quail, just one more dying quail a week... and you're in Yankee Stadium.
There will always be something so beautiful about node/graph programming systems that simultaneously allow you to go back and forth between the text-based program and the graph representation of the program. That fact that Modelica lets you do this with differential equations in a declarative style, including fairly strong encapsulation capabilities, is quite powerful and fun. The only other system I’ve worked with that gave me a similar feeling of being able - at least theoretically, ha! - to model absolutely anything so long as enough attention to interfaces between systems was paid is OpenMDAO.
The main other example that comes to mind is Cycling 74s Gen for Max/MSP.
Anyways, for those who are curious, Modelica was originally developed for the automotive/aerospace industries (IIRC) but is used in all sorts of fields - for me in particular, I know about it because of its use in modeling operational energy in buildings - everything from geothermal borehole fields and mesh topology district heating/cooling systems to complicated HVAC controls and more.
OpenModelica is the main open source tool for building and running modelica programs, and there are python wrappers you can use as well!
> passing them means nothing about the ability of neural net-based systems to understand language, regardless of how much their authors designed them to test language understanding.
Does this implicitly suggest that it is impossible to quantitatively assess a system’s ability to understand language? (Using the term “system” in the broadest possible sense)
Not agreeing or disagreeing or asking with skepticism. Genuinely asking what your position is here, since it seems like your comment eventually leads to the conclusion that it is unknowable whether a system external to yourself understands language, or, if it is possible, then only in a purely qualitative way, or perhaps purely in a Stewart-style-pornographic-threshold-test - you’ll know it when you see it.
I don’t have any problem if that’s your position- it might even be mine! I’m more or less of the mindset that debating whether artificial systems can have certain labels attached to them revolving around words like “understanding,” “cognition,” “sentience” etc is generally unhelpful, and it’s much more interesting to just talk about what the actual practical capabilities and functionalities of such systems are on the one hand in a very concrete, observable, hopefully quantitative sense, and how it feels to interact with them in a purely qualitative sense on the other hand. Benchmarks can be useful in the former but not the latter.
Just curious where you fall. How would you recommend we approach the desire to understand whether such systems can “understand language” or “solve problems” etc etc… or are these questions useless in your view? Or only useful in as much as they (the benchmarks/tests etc) drive the development of new methodologies/innovations/measurable capabilities, but not in assigning qualitative properties to said systems?