Absolutely. They hooked up an LM and asked it to talk like it's thinking. But LMs like GPT are token predictors, and purely language models. They have no mental model, no intentionality, and no agency. They don't think.
This is pure anthropomorphization. But so it always is with pop sci articles about AI.
It's quite an odd setup. If we presuppose the "agent" is smart enough to knowingly cheat, would it then also not be smart enough to knowingly lie?
All I really get out of this experiment is that there are weights in there that encode the fact that it's doing an invalid move. The rules of chess are in there. With that knowledge it's not surprising that the most likely text generated when doing an invalid move is an explanation for the invalid move. It would be more surprising if it completely ignored it.
It's not really cheating, it's weighing the possibility of there being an invalid move at this position, conditioned by the prompt, higher than there being a valid move. There's no planning, it's all statistics.
You could create a non-intelligent chess playing program that cheats. It’s not about the scratchpad. It’s trying to answer a question if a language model, given an opportunity, could circumvent the rules over failing the task.
> could circumvent the rules over failing the task.
or the whole thing is just a reflection of the rules being incorrectly specified. As others have noted, minor variations in how rules are described can lead to wildly different possible outcomes. We might want to label an LLM's behavior as "circumventing", but that may be because our understanding of what the rules allow and disallow is incorrect (at least compared to the LLM's "understanding").
I suspect that this commonplace notion about the depth of our own mental models is being overly generous to ourselves. AI has a long way to go with working memory, but not as far as portrayed here.
I'm afraid the author has it backwards. This style of writing didn't start as a font from some company.
This is just how you used to be taught to do lettering in a drafting class. The straight lines and simple shapes make it easy to do lettering with pencil or a pen, which is also why all the lines are the same width and all the line caps are rounded.
Later on these turned into stencils, which let users do lettering quickly by tracing the pen inside the stencil - and then turned into fonts for use in print. But all that came later. What we have here is just lettering people learned in drafting classes, and which was later used to make stencils and templates.
> This style of writing didn't start as a font from some company.
TFA doesn't suggest that, at all. A font is a concrete instantiation of a writing style, and TFA is about the history of one such concrete instance - not the general style it's an instance of. Also the connection to drafting and lettering stencils is discussed in some detail midway through.
(Also more generally: kind of amazing to imagine reading an article of this depth, that mentions years of obsessive research and links to the author's 1200-page book on the history of typing, and thinking: "yeah this guy probably doesn't know about drafting".)
Of course I read the details on stencils and patterns in the text. But you misunderstand what I'm saying.
What I'm saying is that there are many "concrete instantiations" of drafting lettering style, that all look basically the same, because they all came from the same source, and Gorton is just one of them. So what we're seeing in elevators and on plaques is not "Gorton" specifically. While in contrast we do see Helvetica specifically on NYC subway signs, Johnston in the London tube, etc.
Too fine of a point? Perhaps. And also, it doesn't take away from the quality of the essay which is a delightful romp through the history of draftman's lettering showing up in all sorts of forgotten utilitarian places.
(But I've got to ask - what's with the ad hominem at the end? We should be above that.)
> What I'm saying is that there are many "concrete instantiations" of drafting lettering style, that all look basically the same, because they all came from the same source, and Gorton is just one of them.
Open a couple of old drafting lettering guides (e.g. ones linked in sibling comments or TFA), and look closely. They'll obviously have a similar overall vibe, but there'll be tons of variations - differently shaped 3/4/7, where the curves start and stop on letters like CJGS569, whether the various corners are pointed or rounded or flattened, etc.
If your premise above is true, then we'd expect to see similar sorts of variations in the various fonts derived from drafting style, and TFA's entire point is that the author has collected hundreds of cases where we don't. Check his photos - they show the same font with the same idiosyncrasies, the off-balance G, two flattened points on the 4, the slightly asymmetric 8, etc. TFA is about the ubiquity of that specific set of letter shapes (modulo some variations that he discusses), not just of lettering that's generally in the drafting style.
(Also: in the best of faith and not meant as shade, you might want to look up "ad-hominem".)
> If your premise above is true, then we'd expect to see similar sorts of variations in the various fonts derived from drafting style
If you look at the photos in the article, there are a lot of variations! For example, in the article, if you look at the first two sets of photos of keyboards, you see a variety of shapes, especially visible with the 6s/9s, 0s, Rs, Ss, etc. And then in the next set of photos (the ones with a selection of plaques), you again see a collection of various letter shapes - look at the varying shapes of Gs, Ss, etc. This repeats throughout, when you look at the random assortments of plates and signage.
Later on, after he discusses ANSI and DIN standards, the author goes on to say:
> In the regulatory space, the U.S. military canonized Gorton in 1968 as a standard called MIL-SPEC-33558 for aircraft and other equipment dials, cancelled it in 1998… then brought it back again in 2007.
Except that the specimens he shows right below, of ANSI Y14 and MS 33558 (and whatever the third one is), are very different from Gorton and even from each other - just look at those letter forms. Which makes sense, as their lineage is _not_ from Gorton, but from traditional lettering.
So that's what I mean - it's not that Gorton _specifically_ is everywhere, it's just that draftman's lettering style is everywhere, and in many variants, including the very popular Gorton one.
I saw the argument more as: Gorton begat Leroy and that became the defacto technical drawing standard through the widespread adoption of drawing templates. The differences between Leroy and the standards you mentioned seem very small to me.
Playing cards use this style of lettering. Not sure how far back that goes but I kind of doubt they all derive from Gorton's specific engraving machines.
> And here's A TEXT-BOOK OF FREE-HAND LETTERING, part of the TECHNICAL DRAWING SERIES, first published in 1895:
Called out in the article - "but I know simple technical writing standards existed already, and likely influenced the appearance of the newfangled routing font. [photograph captioned: From a 1895 “Free-hand lettering” book by Frank T. Daniels]"
I know it was a long essay and I skimmed most of it myself. But the author definitely mentions this and even has a picture from the 1895 book you linked to.
After reading such a beautiful piece about an interesting topic and researched for hundreds of miles on end, I can rest assured that somebody on HN knows better. No doubts, no nothing... "just" knows it better.
Edit: The above might sound too harsh, sorry. You're probably right, but after reading such a beautiful piece, your reaction was really a buzzkill for me.
I learned that style of writing in school (German: Normschrift). I think this is where the origins of the font lie, but having read the article I don't see how the author is wrong here.
This is just a thing they failed to emphasize — maybe — considering the credentials, because it was blaringly obvious to them.
This feels more like a detrctive story about figuring out the origins of one concrete manifestation of a font and not a text about where that family of fonts comes from.
The article goes into almost tedious detail about where it came from, and whether it constitutes a "proper" font. I think you may have skimmed too fast.
While I can appreciate the parent commenter's criticism, I think the author tried to head it off with this concession of their chosen naming convention:
> In the end, I’m sticking with Gorton for the whole branch since that feels the most well-known name, but I feel ill-equipped to make that call for everyone. You might choose to call it Gorton, Leroy, TT&H, Taylor-Hobson, or one of the many other names. (Just, ideally, not Linetica.)
Since the author was concentrated on the particular letterforms that seemed more consistent in the lineage observed, the usage makes sense. But that naming, even with the acknowledgement that it likely comes from the standards of drafting of the day:
> I don’t know how this first proto-Gorton was designed – unfortunately, Taylor, Taylor & Hobson’s history seems sparse and despite personal travels to U.K. archives, I haven’t found anything interesting – but I know simple technical writing standards existed already, and likely influenced the appearance of the newfangled routing font.
> This was perhaps the first modern pantograph engraver, and perhaps even the arrival of a concept of an engraving font – the first time technical writing was able to be replicated consistently via the aid of the machine.
But it also seems a reasonable critique of the article that it's mislabeling to call the MIL-SPEC-33558 and ANSI Y14.2M or even the WWII equipment lettering "Gorton" simply by visual similarity without evidence to show ancestry to the specific engraving machines, dies, or letter sets of Gorton/TTH/etc. And that is also done throughout both with direct evidence and without.
That's an assumption, I read it just fine. :) I just disagree with how the author keeps talking about draftsman's lettering as if it were some company's font (e.g. Gorton and Leroy) rather than a commonly taught community standard.
They do later talk about the real origin in 1894 in England, with TT&H creating it due to the constraints of their self-built lettering machine to engrave tiny letters on their products
As I mention in a previous comment [1], this style of hand lettering was common in textbooks prior to 1894. From 1883, for example, we find this specimen in Standard Lettering, published by the Columbia School of Drafting:
And again, the article explicitly mentions that, with a picture of a similar book:
> I don’t know how this first proto-Gorton was designed – unfortunately, Taylor, Taylor & Hobson’s history seems sparse and despite personal travels to U.K. archives, I haven’t found anything interesting – but I know simple technical writing standards existed already, and likely influenced the appearance of the newfangled routing font.
The issue is that the author presents the entire set of typefaces that are similar to Gorton as derived from Gorton without presenting evidence to rule out the obvious alternative lineage: that, just like genuine Gorton, they too were derived from the various regional single-stroke letterforms that draftsman everywhere were taught and used. Excellent draftsman’s examples abounded and would have been so much more common than genuine Gorton and its genuine ancestors that it’s hard to believe that regional companies marketing their own type engraving machines would have had to copy Gorton rather than local examples of the draftsman’s art that were considered superior.
The article has a link to the licensing agreement between Taylor Hobson and Gorton, links to other posts explaining how Leroy bought Gorton machines, an interactive comparison where you can see the similarity of the letters, and dozens of photos and scans of docs where you can compare them yourself, too. I would say that is a lot of evidence presented.
But that evidence is not persuasive that the set of fonts that the author calls Gorton are actually derived from the lineage he presents (see the diagram captioned “The Gorton quasisuperfamily”), rather than from freehand lettering that would have been much more widely used at the time. Remember that the letterforms themselves could not qualify for legal protection in the United States. So none of the licensing agreements offered as evidence were needed to acquire the fonts. So we can conclude that they were executed to acquire the machines and the patents behind them for the purpose of introducing similar machines in a new market. The machine designs and methods of production were the hard part. The fonts were comparatively trivial to create and incorporate into a machine, whatever its design.
The author's comparison of the fonts actually argues against them being of the claimed lineage. Consider the many differences between the Taylor, Taylor & Hobson machine’s fonts and the Gorton machine’s fonts. If Gorton had a license to use the Hobson machine designs, which they did, they could have simply copied the TT&H fonts verbatim. But they clearly did not. Why not? I think it's likely that they simply preferred a different design, one closer to the letterforms that were more commonly used by draftsmen in the American market. In other words, the Gorton reference design was not the TT&H font design.
At least, that's my best guess based on the evidence presented.
This quote is where the author seems to reach the conclusion they want, that is Gorton being the Proto-Indo-European of these drafting letterforms:
> Each of these reappearances made small changes to the shapes of some letters. Leroy’s ampersand was a departure from Gorton’s. Others softened the middle of the digit 3, and Wrico got rid of its distinctive shape altogether. Sometimes the tail of the Q got straightened, the other times K cleaned up. Punctuation – commas, quotes, question marks – was almost always redone. But even without hunting down the proof confirming the purchase of a Gorton’s pantograph or a Leroy template set as a starting point, the lineage of its lines was obvious. (The remixes riffed off of Gorton Condensed or the normal, squareish edition… and at times both. The extended version – not that popular to begin with – was often skipped.)
As a child, I have learned lettering based on the German DIN standards (DIN 16, DIN 17, DIN 1451).
While these standards date from around 1930, they were based on much older lettering textbooks. The oldest that I have seen was from 1871, and these German textbooks did not differ much from the American textbooks quoted above.
I just read the article in full, and don't understand your point, either.
Maybe you could explain what part of the comment you replied to you think is either wrong or already covered by the article, so that either we can realise what we failed to notice in our first reading or so someone can explain to you why you're wrong in thinking that comment isn't correct in the context of having read the full article :)
I'm sure it started as an oral tradition based on legibility. The communicating new ideas about physical objects on paper is difficult and I was taught that is was a lesson on clarity. Codifying best practices is what cultures do as they evolve.
I think it's worth calling a font when there were likely cases of the font being transmitted, by copying from one source to another. E.G. Gorton being made by machines licensed from another company, then the output of those machines being used for lettering guides.
This was also pointed out in the article, but the point at which it became a physical standard, and not a lettering style, was with the Gorton engravers.
(The tail of the Q is a tell -- not a straight line, not a simple shape, but identical in lettering systems based on the engravers.)
Yeah -- my dad's lettering template had letterforms like this. As the article mentions, the Simplex characters from the Hershey set, used in CAD applications, are based on this font, probably for consistency with how drafting was traditionally done.
I would just add that Mathematica-style term rewriting (e.g. analytical integration, or equation solving) is done with _semantics-preserving_ symbolic solvers, which are hand-made and human-reviewed to guarantee correctness.
LLM style pattern matching and rewriting does not preserve semantics, except accidentally due to an overwhelming amount of examples.
Planning and reasoning have long standing meaning in AI, based on explicit knowledge representation and inference. It's a subspace of AI that predates LLMs by decades.
But LLMs don't do that kind of planning and reasoning, and CoT is very clearly not a reasoning technique in the AI sense.
Also, pros: world-class architecture, fun music scene, great museums and of course The Art Institute, top universities and everything that comes with that, great restaurants from every cuisine under the sun, very immigrant-friendly city, lovely tree-lined streets with block parties, I could go on and on. Oh, and living is affordable and people are Midwest Nice.
Cons: it's the great flatness - you have to drive for hours to see hills. Tech industry is ok but very b2b and fintech focused and not a lot of consumer or entertainment stuff. Also as a very liberal city it's a favorite target of culture warrior types.
Tie: winter. Yeah, it gets cold and snowy. But also, it's "just weather". It's the same every year and people deal with it pretty easily.
As a life long Chicagoan I really think you're underselling how bad the winter is. Combined with the flatness it means that most people just don't spend any time at all outside december-april. I think this is why Chicago has such a strong drinking culture, there's just not much else to do in the winter, certainly not 5 months worth of stuff.
I've only visited Chicago once. It was nice, but I'd never thought architecture would be a selling point. Do you have examples of where I should look, the next time I'm there?
1. We can observe how the state machine gets generated, first just a jumble of locations in a hub and spokes topology (no correlations), then some correlations start happening pairwise, making a kind of a beads on a string topology, and then finally the mental model snaps marvelously to two completely separate paths that meet at ends. It's amazing to see these mental models get formed in vivo out of initial unstructured perceptions.
2. In addition to standard HMM modeling, authors find that a "biologically plausible recurrent neural network (RNN) trained using Hebbian learning" can mimic some of this (but not exactly). But more interestingly, they find that LSTMs or transformers cannot. Which makes sense structurally, but it's a good reminder for those who believe the anthropomorphic hype that transformers have memory or other such (they don't :) ).
Couldn't memoryless neural networks still possibly learn the Next-State function of a Finite-State machine? Depending on the training algorithm. Especially if the eventual usage of such networks is to be called over and over again to generate the next token; conceptually this to me seems analogous to the process of finitely unrolling a while loop or a computer pipeline.
This is pure anthropomorphization. But so it always is with pop sci articles about AI.