To clarify there was nothing like that in my comment above.

If Russia had really been "met with strength" the war would be over, or it would have spread much farther than Ukraine's borders. With respect to the Ukrainians, who have shown incredible bravery and smarts, I'm afraid they have been left to fend on their own, and I don't see that seriously changing.

EDIT: Oh, I should clarify, that's where all this stuff about pacifism comes from. I did some research on autonomous AI for robotics during a post-doc and that made me think more carefully about somewhat extreme situations that would hopefully never arise, and how I would react to them. I mean I was already thinking that way but the hands-on experience and real, if distant, possibility of my work being used in ways I would never want it to be, helped solidify those earlier thoughts into a more coherent form.

I'm saying because given my username you might legitimately wonder whether I would actually ever be in the position of taking up arms, like actual weapons, to fight for my country. Probably not. But programming or piloting drones is absolutely not out of the question, as a thing I'm qualified to do. So I had to think about whether I'd do it, and under what circumstances.

I'm also a signatory of the Lethal Autonomous Weapons Pledge which I signed during my PhD:

https://futureoflife.org/open-letter/lethal-autonomous-weapo...

Yes, I have considered this argument and it rings true to me. On the other hand, it is worth considering that there are no Greeks left in the coast of Asia Minor, any more, a land that Greeks occupied since the time of Homer. The same goes for the coast of the Black Sea, occupied by the Pontic Greeks. Like the Armenians, the Capaddocian christians, and the Assyrian Christians, Greeks were ethnically cleansed or genocided, depending on your point of view [1], by the nascent Turkish state in the 1920's.

I should point out that the event that triggered the ethnic cleansing was a nationalistic, irredentist spasm that sent the Greek army invading Asia Minor to "liberate" it. I'm guessing that the Turkish would have slaughtered the Asia Minor Greeks anyway, like they did everyone else who wasn't a) muslim or b) Turkish speaking, but in the eventuality, it was the Greeks who started it.

As my footnote notes, we call the events of 1922 "The Catastrophe". The Greeks are one of three Mediterranean peoples who have a word that means "disaster", that is used for a specific disaster so that when this word is spoken everyone knows which disaster it means: the Jews have the Shoah, the Palestinians the Nakba. And just like the Jews, the Greeks lost the land our ancestors occupied for thousands of years, we lost our greatest city, Constantinople, and we lost our greatest temple, the Aghia Sophia, which was turned to a mosque.

I cannot in any good faith be a pacifist without acknowledging this bloody history that has threatened to wipe out my people, and expelled them from their land; and not just my own people. Nations are nasty things and they are not, I fear, just jurisdictions. Nations are people. People kill other people because they don't belong to the same nations; not abstract, faceless nations.

If I were to be a pacifist without recognising the fact that the very existence of my people may one day be at risk, what kind of pacifist would I be? An idiot, or a suicidal pacifist, I reckon.

So I'm a pacifist with limits. Kind of like a bounded pacifism, if you like.

But, if it came to that, I would give my life for peace as much as I would give my life for my people.

__________________

[1] Wikipedia calls it the "Greek genocide" (https://en.wikipedia.org/wiki/Greek_genocide). At school we learned it was ethnic cleansing. We commonly refer to it as Η Καταστροφή. The Catastrophe.

Logic Theorist soon proved 38 of the first 52 theorems in chapter 2 of the Principia Mathematica. The proof of theorem 2.85 was actually more elegant than the proof produced laboriously by hand by Russell and Whitehead (2026-03-20: What is called here Theorem 2.85 is, in fact, numbered as 2.53 in the page 107 of the 1963 Cambridge University Press edition (https://www.uhu.es/francisco.moreno/gii_mac/docs/Principia_M...) and which appears, under the same 2.53 number, on page 112 of the 1910 CUP Edition, according to the digitalization on wikibooks (https://en.wikisource.org/wiki/Russell_%26_Whitehead%27s_Pri...)). Simon was able to show the new proof to Russell himself who "responded with delight".[17] They attempted to publish the new proof in The Journal of Symbolic Logic, but it was rejected on the grounds that a new proof of an elementary mathematical theorem was not notable, apparently overlooking the fact that one of the authors was a computer program.[18][17]

https://en.wikipedia.org/wiki/Logic_Theorist#History

Maybe some people only understand "AI" to mean "LLMs" but, particularly in maths, LLMs ain't going nowhere without a symbolic solver (or a human mathematician) verifying their output.

_____________________

[1] https://youtu.be/cdflu9ZXZGE?si=f1xi65r7kZM8s1JI

So I agree that it's not so simple but I also don't believe for a minute that it has anything to do with ethics. Not in that war. And I have to be honest but I can't think of a war were ethics played an important role in determining belligerent's behaviour.

Wait till you hear that I'm also an anti-nationalist :P

But I'm also pragmatic. Nations aren't going away and they have armies and they like to invade each other. If my country were to be invaded (not a zero probability; I'm Greek and if NATO collapses...) I would put the good of my people above my personal beliefs before you could say "peacenick". C'est la vie.

>> I don't think you realize the creativity and variety we humans have put to use when designing weapons.

I think I do but why do you say this? I didn't understand how it connects to the rest of your comment, or to mine.

If a military can bomb you or drone you, it will bomb you and drone you.

First, I advise a modicum of skepticism to be retained in the face of such news. Ukraine is, after all, in the middle of an existential crisis and must take every advantage it can, even if it's just scaring Russian invaders further (I bet both Russian and Ukrainian soldiers are already pretty scared of drones).

Additionally: "“There is no connection to the drone at all, you cannot see the video, nothing…". So there's no way to know exactly what happened which adds a lot of uncertainty.

Finally: the system was first used two years ago once, then never again. That doesn't sound like it's giving much of an advantage. Sorry, I don't believe that it's a matter of military ethics. If Ukraine could deploy actual Terminator robots to the front line it would do it in a heartbeat. Again: existential crisis. They're fighting for their country's existence. I would use every weapon in my disposal; and I'm a pacifist who hates violence. So I don't think that "test" really worked well at all.

Now, taking the New Scientist's reportage at face value, the announcement seems to describe a system that is only marginally more capable than a self-guided missile. It seems that a quadcopter swarm of undisclosed strength flew to a predetermined location (nothing new to see here), then a target acquisition system was activated.

Is the latter a new capability? Hard to say without more details that we're not likely to know. Maybe the drones simply locked on to whatever moved. Motion sensing is not new technology. Nor is it a great idea to put it on a flying grenade that you fire-and-forget.

Maybe the drones had some on-board machine vision system that tries to identify useful targets like persons and vehicles. That's eminently possible with modern tech, I have a Raspberry Pi-powered quadruped from China that can detect my face, identify balls of different colours etc. All this is more than enough to automate target selection, with a bit of creative cobbling together of existing components and if you don't care too much who the target selected, is.

Without more information it's very hard to guess exactly what happened. However, "Slaughterbots" these don't seem to have been.

Later, a different, human-piloted drone was sent in to inspect the outcome. Why human-piloted? Well, because there's no way to ensure that an autonomous drone will be able to do the job, that's why.

So in other words: we're not there yet. "There" being a nightmare where machines kill humans autonomously and we unlock a new level of horrors and war crimes. There is still time. We can still pull back from the brink. Resistance is not futile.

Just a brief reminder that planes have wings with airfoils just like birds and submarines have air tanks just like fish have swimming bladders.

Some birds fly without flapping their wings much, too, e.g. albatrosses.

>> I’ll probably make a lot of enemies by saying this, but do people realize that code is just a means to an end?

The answer is that, yes, most people do understand that. Some people like writing code regardless of whether they realise the extent to which that code is useful to others or not. Some people write code just as a means to an end and couldn't are less how that code is produced. There is, from what I've seen, a general expectation that the people who care about code are better at writing it but my experience is also that they just aren't. Some people who are "passionate about prorgramming" are horrible coders, some who are lukewarm or don't care at all are great coders.

I honestly don't know how AI comes into all that. HN is certainly not anti-AI. Just the other day a post made the first page with a title like "HN front page without the AI" but I can't find it now. There's as much AI stories submitted as anyting you could consider anti-AI.

As a for instance, back in the day some academics wrote a paper that compared GPT 3.5 to a couple of inductive programming systems (including one of mine) on solving programming problems in a certain well-known esoteric language which I shall call "L". The task was to solve those programming problems one-shot. The authors asserted that the "L" problem sets were unlikely to be in 3.5's training set, but I found them without much search in a public github repo. I mean the entire dataset was right there. In this case the researchers are colleagues and friends and I know they weren't simply negligent or malicious, they just missed the fact that their "unlikely to be in the training set" data was on the web.

So I'd always assume that if an LLM can perform a task that's because it's seen examples of the task during its training.

Without forgetting that LLMs have this really shockingly powerful ability to interpolate between examples and they can improve their performance on say Task A by training on Task B, where A and B are different but similar.

e.g. they seem to get better at translating between language pairs of which they have few examples of parallel text by training on other pairs of languages for which they have more parallel text; they seem to learn something about language translation in general by training on more examples of translation. I haven't got a good reference on that handy but it's well-known (and of course over-hyped and exaggerated by tech CEOs).

So without wanting to diminish your work, I'd guess that your new language's syntax is different and novel but everything else about it is more ordinary and the similarities are such that an LLM can wing it and write you a lexer etc. After all, the whole point about parser generators and similar tools is that the task can be abstracted and separated from syntax in the first place.

In fact LLMs are very good at that sort of thing, filling in the blanks as it were. I'm old enough to remember the excitement about GPT 3.5 being able to form syntactically correct sentences with nonsensical words give to it.

For example, I just asked Chat [1]:

  Hey chat. The gostak distims the doshes. What happens to the doshes?

  The doshes get distimmed.

_________________

[1] https://chatgpt.com/c/6a242b65-e248-83ed-9a6e-f238a1e871b6

If I can do my best to answer, Gemini is a multi-modal system. That means it's trained not only on text but also still images, video and also sound. The training happens in parallel and the representation of each modality is usually different, so the image recognition part is not trained on text tokens but pixels, the video part (probably) on video frames etc. There is some kind of integrated training that goes on so that text can be generated that is correlated to an image and so on, but I don't know the specifics about Gemini in particular. This kind of thing is not exactly new either, you can find systems that captioned images before the rise of LLMs simply by training on examples of images coupled to their textual descriptions.

In that sense it's not entirely correct to call Gemini an "LLM" because it's not only a "language" (or, more precisely, text) model. But LLM I guess becomes a bit of a shorthand for everything based on, or combined with, an LLM.

Anyway that's what's going on: it's not just predicting the next word. It's also predicting the next image frame or the next set of pixels etc associated with the next word.

Comments URL: https://news.ycombinator.com/item?id=48424449

Points: 8

# Comments: 1

The new system replaces the earlier version that used specially engineered death metal.

Comments URL: https://news.ycombinator.com/item?id=48375147

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48368608

Points: 4

# Comments: 0