bloaf

I'm not a fan of the way grey hydrogen was written off: That hydrogen is already being produced today by several different refinery processes, and then burned in a furnace because no one else wants it.

So the right way to handle the carbon accounting isn't to assume that all the CO2 produced by the refinery processes count against the hydrogen produced, but rather that the energy that refineries get from burning the hydrogen would be replaced by them burning natural gas instead.

The per-kg energy value of burning H2 is ~2.5x the value of natural gas (refineries generally use LHV for this accounting). But each kg of natural gas that gets burned produces ~2.8 kg of CO2 (because burning replaces the puny hydrogen with relatively larger oxygen atoms).

2.5*2.8 = 7kg of CO2 per kg of H2 taken out of the refinery. Which isn't as big a difference from the 10kg reported in the article as I expected when I set about writing this comment.

You think companies are all deliberately leaving big money on the table by making hourglass clothes as an oopsie?

They're doing it because people are buying clothes based on superficial appearance, and most people prefer the aesthetics of the hourglass shape.

Rectangular clothing doesn't sell as well because it doesn't look as good on a mannequin even if it fits better.

Expecting mass-market, lowest-common-denominator products to be tailored to your special circumstance is the issue.

Normalize going to a tailor, instead of grumbling about how you aren't benefiting enough from the sweatshops mass retailers are running.

It would have prevented both this 7zip attach and the recent notepad++ one.

I've started using winget to install my apps for exactly this reason. I can't keep track of every url for every piece of software.

At my company I saw a team of devs pay for a special purpose "query optimized" database with "exabyte capability" to handle... their totally ordinary HR data.

I queried said database... it was slow.

I looked to see what indexes they had set up... there were none.

That team should have just used postgres and spent all the time and money they poured into this fancy database tech on finding someone who knew even a little bit about database design to help them.

This thing was done in 2 weeks. In the orgs I've worked in, you'd be lucky to get HR approval to create a job posting within 2 weeks.

"I have no tools because I've destroyed my tools with my tools" is a phrase I think to myself at least weekly.

We all know there has been a replication crisis across many different disciplines of science. I think that the set of things we actually know about nutrition and health is a lot smaller than the experts think.

However, the problem is that the public has also come to that conclusion. The public has gone on to decide "that means my incredibly weakly-evidenced idea is just as good as the expert opinions" which does not follow and is often disastrously wrong.

So I'm also sympathetic to the idea that the saturated fat picture is more complex than a blanket ban suggests. But I know better than to treat things like Brad's arguments as anything other than "interesting hypothesis" as opposed to "something we actually know about nutrition."

> Nowhere else can you get scientific computing, a GUI toolkit, a high level embedded software environment, a HiL/SiL toolkit, a model based simulation environment, a plotting and visualization toolkit and so much more in a single cohesive package. Nobody else has any offering that comes even close.

Mathematica does. Arguably Mathematica is even more cohesive because it's not split up into "feature sold separately" packages.

I have a computer running fedora with a 5080 GPU. The newer games all work fine (e.g. cyberpunk 2077, BG3). The really old games mostly work fine (e.g. system shock 2, original Deus Ex) . The in-between games... are really hit-or-miss (e.g. I can't get witcher 1/2 or Deus Ex Human Revolution to run).

> The harms engendered by underestimating LLM capabilities are largely that people won't use the LLMs.

Speculative fiction about superintelligences aside, an obvious harm to underestimating the LLM's capabilities is that we could effectively be enslaving moral agents if we fail to correctly classify them as such.

> When you have a thought, are you "predicting the next thing"—can you confidently classify all mental activity that you experience as "predicting the next thing"?

So notice that my original claim was "prediction is fundamental to our ability to act with intent" and now your demand is to prove that "prediction is fundamental to all mental activity."

That's a subtle but dishonest rhetorical shift to make me have to defend a much broader claim, which I have no desire to do.

> Language and society constrains the way we use words, but when you speak, are you "predicting"?

Yes, and necessarily so. One of the main objections that dualists use to argue that our mental processes must be immaterial is this [0]:

* If our mental processes are physical, then there cannot be an ultimate metaphysical truth-of-the-matter about the meaning of those processes.

* If there is no ultimate metaphysical truth-of-the-matter about what those processes mean, then everything they do and produce are similarly devoid of meaning.

* Asserting a non-dualist mind therefore implies your words are meaningless, a self-defeating assertion.

The simple answer to this dualist argument is precisely captured by this concept of prediction. There is no need to assert some kind of underlying magical meaning to be able to communicate. Instead, we need only say that in the relevant circumstances, our minds are capable of predicting what impact words will have on the receiver and choosing them accordingly. Since we humans don't have access to each other's minds, we must not learn these impacts from some kind of psychic mind-to-mind sense, but simply from observing the impacts of the words we choose on other parties; something that LLMs are currently (at least somewhat) capable of observing.

[0] https://www.newdualism.org/papers/E.Feser/Feser-acpq_2013.pd...

If you read the above link you will see that they spell out 3 problems with our understanding of thought:

Consciousness, intentionality, and rationality.

Of these, I believe prediction is only necessary for intentionality, but it does have some roles to play in consciousness and rationality.

And the big players have built a bunch of workflows which embed many other elements besides just "predictions" into their AI product. Things like web search, to incorporating feedback from code testing, to feeding outputs back into future iterations. Who is to say that one or more of these additions has pushed the ensemble across the threshold and into "real actual thinking."

The near-religious fervor which people insist that "its just prediction" makes me want to respond with some religious allusions of my own:

> Who is this that wrappeth up sentences in unskillful words? Gird up thy loins like a man: I will ask thee, and answer thou me. Where wast thou when I laid up the foundations of the earth? tell me if thou hast understanding. Who hath laid the measures thereof, if thou knowest? or who hath stretched the line upon it?

The point is that (as far as I know) we simply don't know the necessary or sufficient conditions for "thinking" in the first place, let alone "human thinking." Eventually we will most likely arrive at a scientific consensus, but as of right now we don't have the terms nailed down well enough to claim the kind of certainty I see from AI-detractors.

Everyone is out here acting like "predicting the next thing" is somehow fundamentally irrelevant to "human thinking" and it is simply not the case.

What does it mean to say that we humans act with intent? It means that we have some expectation or prediction about how our actions will effect the next thing, and choose our actions based on how much we like that effect. The ability to predict is fundamental to our ability to act intentionally.

So in my mind: even if you grant all the AI-naysayer's complaints about how LLMs aren't "actually" thinking, you can still believe that they will end up being a component in a system which actually "does" think.

I'll make the following observation:

The contra-positive of "All LLMs are not thinking like humans" is "No humans are thinking like LLMs"

And I do not believe we actually understand human thinking well enough to make that assertion.

Indeed, it is my deep suspicion that we will eventually achieve AGI not by totally abandoning today's LLMs for some other paradigm, but rather embedding them in a loop with the right persistence mechanisms.

As someone who regularly looks up things I read "a while back," her experience is very common and insanely frustrating.

There always do exist magic combinations of words that you can put into google that will find the thing you're looking for. But the search space doesn't feel differentiable in a mathematical sense: you can't iteratively improve your terms because you either hit on a combo that works, or you get the same wrong results as you saw for your past 10 searches.

On android, KDE Connect felt great for basic copy/pasting/file transfer between devices.

https://kdeconnect.kde.org/

On iOS, KDE Connect feels like its running a potato sack race with both arms tied behind its back.

So tcl handles it somewhat more elegantly, I think. It also has a set function, but does not require any special quoting because it uses the $ prefix to denote "get the value of this symbol":

    set a 5
    puts $a  #prints 5

and of course because it is modeled as a function (albeit an impure one) you can pass arguments to it as normal:

    set a b
    set $a 5  #equivalent to set b 5
    puts $b   #prints 5

of course, everything in tcl is a string, so this works too lol

    set a 5
    set $a b
    puts $5  #prints b

> Rye also has left to right flow so it adds left-set-word. In Rye all assigned words with set-words are constants and they are used by default. So we also need a "mod-word", that is the double colon that you noticed, and left-mod-word

So I would assume that the :i is actually constant within the loop body scope. That is, the loop function is doing something like this:

; i is not assigned in this scope

evaluate {1 :i, prns i}

evaluate {2 :i, prns i}

evaluate {3 :i, prns i}

; i is still not assigned in this scope

But it sounds like you're telling me that :i would actually escape the scope of the loop body and so it needs to be modifiable or else the loop will break.

Certainly:

I can tell the rye devs like the idea of everything being a function. But in their very first "language basics" section, they introduce assignment not as a function call, but as some kind of magic that happens when you have colons in a name.

So when we get to the "looping" section, it is the first time we have seen a colon-having-name outside the context of assignment:

> loop 3 { ::i , prns i }

And it is explained that the above line of code is "injecting" values for the code block to "pick up".

But right away this begs a number of questions:

* Why the double-colon? I would assume each loop-body-evaluation happens in its own scope, and that we're not creating global variables, so a single colon (:i) should be sufficient, right?

* What are we doing, conceptually? Is the ::i meant to be "a function which when given a value modifies its enclosing scope to include the symbol i" or "an unassigned symbol which the loop function will use to do something akin to term-rewriting with?"

* Do we really need a symbol at all, or could we just have a the point-free loop "loop 3 {prns}"?

* If we can't have the point free thing, is it because somehow the injected value would end up "to the left" of prns, if so, why would we want that?

* If we're doing something more like term rewriting, why isn't the symbol given as a separate argument from the body?

I checked the Rye homepage, and it has "dialects" which allow more familiar math infix notation. In that sense it is very tcl-y, tcl technically being a lisp.

Looking at their rather confusing looping mechanisms, they probably could benefit from being a little more tcl-y, since tcl has some of the best looping semantics I've worked with.

Most radio stations are simulcast online, so strictly speaking you can get the same content without an fm tuner

https://radio.garden

I think it is important to remember that "decades" here means <20 years. Remember that in 2004 it was considered sufficiently impossible that basically no one had a car that could be reliably controlled by a computer, let alone driven by computer alone:

https://en.wikipedia.org/wiki/DARPA_Grand_Challenge_(2004)

I also think that most job domains are not actually more nuanced or complex than driving, at least from a raw information perspective. Indeed, I would argue that driving is something like a worst-case scenario when it comes to tasks:

* It requires many different inputs, at high sampling rates, continuously (at the very least, video, sound, and car state)

* It requires loose adherence to laws in the sense that there are many scenarios where the safest and most "human" thing to do is technically illegal.

* It requires understanding of driving culture to avoid making decisions that confuse/disorient/anger other drivers, and anticipating other drivers' intents (although this can be somewhat faked with sufficiently fast reaction times)

* It must function in a wide range of environments: there is no "standard" environment

If we compare driving to other widespread-but-low-wage jobs (e.g. food prep, receptionists, cleaners) there are generally far more relaxed requirements:

* Rules may be unbreakable as opposed to situational, e.g. the cook time for burgers is always the same.

* Input requirements may be far lower. e.g. an AI receptionist could likely function with audio and a barcode scanner.

* Cultural cues/expectations drive fewer behaviors. e.g. an AI janitor just needs to achieve a defined level of cleanliness, not gauge people's intent in real-time.

* Operating environments are more standardized. All these jobs operate indoors with decent lighting.

You're not thinking like a salesman. The value of vague is that you can make your product sound like it is the fulfillment of that prediction in ways that you can't do if the prediction is too concrete.

No one is quite sure what will constitute "sufficiently like a dude in a ditch with a rifle" to transform warfare, and that's precisely why the AI startups are flocking to defense spending like the seagulls to the dump.

So I think the author is right regarding current sales pitches, but also wrong about the long term impact. There is nothing on the market today that would be anything but an incremental improvement in an already existing capability.

However

The thing everyone knows is coming is fully artificial boots-on-the-ground, and this will be a categorical revolution on par with the use of airplanes in warfare. (note: "boots" could still be drones).

Like the author points out "A battalion commander in the Ukrainian Army stated to me a few weeks ago that even in the Summer of 2025, the single most decisive factor in his unit’s ability to hold and take ground is the number of infantry".

The instant you have a factory that churns out the equivalent of a "dude with a rifle in a dugout", you've suddenly changed warfare in all the non-tactical ways the author was talking about.

The politics of war becomes different: your battlefield losses may no longer be measured in lives, no one comes home with PTSD, and no collateral damage is due to soldiers' malice.

The doctrines of war become different. Suicide charges/runbys are fine, there's no need to worry about PoWs, and friendly fire is potentially a thing of the past.

There are definitely sites which block firefox, even though they work fine in firefox. Most of the time, the block can be bypassed with simple user-agent spoofing.

I don't think its nearly as cut-and-dry as that. Even if you tried to make tests to differentiate world-model from non-world-model, all you'd end up concluding is:

If the AI has a world model, its world-model doesn't have features that allow it to do what I tested for.

This is great. For a long time I've had a gut feeling that there must exist a synthesis between pure functional, array, and data-oriented programming that gives you the best of all worlds:

Type safety, concise-to-the-point-of-terse code (with the types helping humans read it), and very high performance.

And this feels like a step in the right direction.

> The Church's understanding of morality draws heavily from natural law theory.

This is true, although I do want to draw a bit of a distinction between the churches understanding-qua-official-teaching, and understanding-qua-what-actual-catholic-officials-believe. I often see very devout people look at something like CCC 1956:

> The natural law, present in the heart of each man and established by reason, is universal in its precepts and its authority extends to all men. It expresses the dignity of the person and determines the basis for his fundamental rights and duties

and come away with "Our moral rights and duties derive from our dignity via natural law" which isn't quite right, but nevertheless drives their behavior.

Its just like how the church finally, in 1822, explicitly allowed heliocentric books to be published. Technically, the church never officially asserted geocentrism as a doctrine and so heliocentric books should have been fine, but in practice, the chief censors were actually prohibiting them from being published because it was the common view among officials at the time that the church had in fact officially condemned heliocentrism in the Galileo case.

> I would also say that "postulate" is not the right term, as the Church is not postulating. It accepts this as true.

Yes, the better word for me to use would have been axiom; I was muddling my mathematical terms a bit.

> I have no idea what you mean here. The intellect and will are held to be immaterial faculties, making human souls intellectual

I am saying that because of the belief that "you can't explain that physically" where that = "abstraction", we've entered "God of the gaps" territory.

Now its true, I have read philosophical arguments that abstraction (or in the case of Ed Feser's argument, Incompossibility) is fundamentally impossible to do physically. And indeed, if those arguments succeeded we would be out of the woods. But I've universally found the philosophy to be very weak; to the point that even an amateur philosopher like myself can see that there are real-actual logical flaws, or that they rely on what appear to be extremely weak premises.

> Concerns have nothing to do with some kind of conceptual threat to the "specialness" of human beings. AI, on this account, simply cannot reason; if it could, then it could, but it cannot.

This is precisely the worry. This is a falsifiable prediction of Catholic theology: the instant there exists an AI which can actually reason, Catholic theology will have been falsified.

Now no doubt the Catholic philosophers will respond to such an eventuality by simultaneously claim that the machine isn't "doing it right" and that "our other accounts of Catholic theology are better anyway", but real credibility-damage will be done to Catholic theology.