News for hackers. Stuff that matters.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @10:41AM (#32343534)
I have a phrase I'd like to coin in contrast to AI. "Artificial Bullshit". AB.

I of course mean "bullshit" in the highly technical sense defined by Frankfurt [1]. The defining feature that separates a bullshitter from a liar is that a liar knows and understands the truth and intentionally misrepresents the matters of fact to further their aims, whereas a bullshitter is wholly unconcerned with the truth of the matters they are discussing and is only interested in the social game aspect of the conversation. Bullshit is far more insidious than a lie, for bullshit can (and often does) turn out to be coincident with the truth. When that happens the bullshitter goes undetected and is free to infect our understanding with more bullshit made up on the spot.

DallE generates the images it thinks you want to see. It is wholly unconcerned with the actual objects rendered that are the ostensible focus of the prompt. In other words, its bullshitting you. It was only trained on how to get your approval, not to understand the mechanics of the world it is drawing. In other words, we've trained a machine to have daddy issues.

A profoundly interesting question (to me) is if there's a way to rig a system of "social game reasoning" into ordinary logical reasoning. Can we construct a Turing Tarpit out of a reasoning system with no true/false semantics, a system only designed to model people liking/disliking what you say? If the answer is yes, then maybe a system like Dalle will unexpectedly gain real understanding of what it is drawing. If not, systems like Dalle will always be Artificial Bullshit.

[1] http://www2.csudh.edu/ccauthen/576f12/frankfurt__harry_-_on_...

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by ___rubidium___ on Thursday August 04, 2022 @11:12AM (#32343968)
I think you're right, but I would qualify that the AI is bullshitting in the same way that a child's drawing of a stick figure, house, and smiling sun is bullshit designed to get approval. The AI is giving symbols--very visually stunning ones, to be sure, but symbols nonetheless--of what it is prompted to create, just like a child learns that "circle with lines coming out of it" is a symbol that can be read as "sun" and praised by adults.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by notahacker on Thursday August 04, 2022 @01:59PM (#32346178)
I don't think the child's drawing of a stick figure, house and smiling sun is just "bullshit designed to get approval" though. I think they care deeply about the concept of mummy, and how mummy relates to various things especially them and have strong emotional impulses towards mummy. Learning that a stick figure is an symbolic representation of mummy that might meet with approval is a trivially small part of that intrinsic interest in mummy, not the extent of their engagement with the concept of mummy. And if you tell them you don't approve of their drawing, they'll burst into tears and possibly have a completely different attitude towards drawing and/or you, not just iterate on the bullshit. Similarly, the kid's concept of mummy being happy or sad is a bit more nuanced than which way up the crescent under the two dots should be....
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @12:01PM (#32344635)
The reason for my final question at the end is this: can AI emerge out of AB? The child eventually grows up and (hopefully) stops bullshitting everything. But, do they actually stop bullshitting, or is the human ability to reason itself somehow built out of bullshit?
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by ___rubidium___ on Thursday August 04, 2022 @12:24PM (#32344963)
That's a very good question. The child eventually learns to stop using symbols and instead begins to copy the shapes, negative spaces, curves, etc. in front of them and makes an assemblage of otherwise-meaningless shapes, lines, curves which is interpreted by viewers as a representation of the real thing. Does AI have the potential to do that? I don't know.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IshKebab on Thursday August 04, 2022 @02:07PM (#32346287)
Nobody grows out of your definition of bullshit - doing things that they think others want to see.

But that definition is ... well bullshit. Bullshitting is a deliberate deceptive act. Children aren't being deliberately deceptive when they come up with nonsense answers to questions they don't understand.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by thaw13579 on Thursday August 04, 2022 @01:09PM (#32345541)
To me, Dall-E seems analogous to a film production team that produces visual imagery reflecting a script written by a screenwriter. By the above reasoning, would that team would be producing "bullshit"? I think most people would think not, because the goal isn't to communicate objective truth about the world, rather something plausible, interesting, entertaining, etc. (unless it is a documentary).

I also think distinguishing bullshit from lying depends heavily on internal mental thoughts, goals, and intentions. Isn't talking about Dall-E this way personification and ascribing some level of consciousness?

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by croes on Thursday August 04, 2022 @02:53PM (#32346849)
The production knows was it puts together Dall-E seems not
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by thaw13579 on Thursday August 04, 2022 @04:37PM (#32348020)
Exactly my second point. Producing "bullshit" requires some internal thoughts and intentions, as the parent put it "DallE generates the images it thinks you want to see" (emphasis mine). A very big claim, which is interesting to consider, but probably premature.

However, I do think that Dall-E is able to learn complex high-order statistical associations, i.e. beyond just juxtaposing and visually blending objects. For a recent example, this post with a prompt "ring side photo of battlebots vs conor mcgregor":

https://twitter.com/weirddalle/status/1554534469129871365

What is amazing here is the excessive blood and gore. That feature can't be found in any individual battlebot or MMA match, but it is exactly what you would expect from robots fighting a person. Pretty amazing, and I wonder at what point we could consider this analytical reasoning.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by tomc1985 on Friday August 05, 2022 @01:01AM (#32351911)
I wonder if it made the association from being shown images with various kinds of fights, a large portion of which may have been bloody.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by leereeves on Thursday August 04, 2022 @10:58AM (#32343779)
> It is wholly unconcerned with the actual objects rendered that are the ostensible focus of the prompt.

I disagree. To the extent that the training data are images of actual objects, recreating images of actual objects is the only thing DALL-E cares about.

If we define "caring" about something as changing behavior to cause that to happen, then a neural network doesn't "care" about inference at all, because inference never changes the network's behavior.

It also doesn't know or care about your approval. It only cares about minimizing the loss function.

(But now that you bring this up, I think it would be really interesting to create a network that, after training initially on training data, began interacting with people and continued training to maximize approval.)

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by mjburgess on Thursday August 04, 2022 @11:11AM (#32343947)
Images of objects arent objects and measure only their 2D geometrical properties projected onto the image plane. These properties are correlated across images, inexplicably, since the machine has no access to the physical properties of those objects, nor the laws of physics.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by leereeves on Thursday August 04, 2022 @11:20AM (#32344095)
It's hypothesized (and, I think, likely) that the machine learns some of those hidden/latent properties, though, obviously, only to the extent that they are useful to recreate the images.

In any case, it does at least care about the appearance of the actual objects. So I think it would be fair to say that there are aspects of the actual objects that the network doesn't care about, but there are also aspects that it cares very much about. Thus it's not "wholly unconcerned with the actual objects".

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @11:47AM (#32344461)
My gripe with this is that it has no understanding of what the objects are beyond "the thing that's supposed to be output". Its a textbook case of the Chinese room argument [1]. Sure it can follow translation rules, even infer some sophisticated rules and patterns about the slips of paper that are passed in, but it still has no idea what any of it means.

https://en.wikipedia.org/wiki/Chinese_room

Dalle is very concerned with the 2d shapes of things we call objects, and has correctly inferred some rules about those shapes, but it neither knows nor cares about the things we call objects and how the shapes it has learned are representations of them. It doesn't do reasoning about the round peg fitting in the round hole. It just glues a few pegs and holes together in a way that's mostly consistent with our expectations and says "is this what you wanted"?

Its a student that cares about passing the test, not learning the material.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by leereeves on Thursday August 04, 2022 @12:34PM (#32345105)
Isn't that true of all of us? We all care selectively about certain attributes of objects.

I care that my car is quiet and has comfortable seats, I don't care (or know) what material the muffler is made of, but somewhere there is an engineer who cared about that.

A road designer cares what the tires are made of and how much it weighs, but doesn't care what color the paint is.

An AI recreating an image of my car would care what color the paint is, but not how comfortable the seats are.

I think I see what you're describing - the AI has a very limited scope and doesn't know or care about most of the things we do - but I think that's just a limitation of our current small models and limited training data, not an inherent limitation of neural networks.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @01:22PM (#32345705)
You may not know what material the muffler is made of, but you know its made of a material. You "care about the objects themselves" in the sense of having this framework of objects, all of which have certain properties and relations to each other which form a predictive model you have about the world. You expect everything to have a position in x,y,z, to have a size, shape, mass, hardness etc. You might not care about a particular property of a particular object in a particular context, but you fundamentally think in the semantics of objects. You know what to expect when the heavy hard object meets the brittle lite object.

DallE doesn't really have a system of the semantics of objects. It doesn't know why it would be useless for a building to have all the doors on the 2nd level. Its not even clear that DallE makes use of discrete "objects" in its limited understanding.

Here's an example from budget DallE

https://ibb.co/B2B5zky

It understood the shape of "stick figure" and "boobs", but had no understanding of what a stick figure is meant to represent and thus where it should place the boobs. The results are hilarious. I'm not sure which I like more, the guy with a boob casually walking down his arm, or the lady with a boob head that's shrugging with uncertainty.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by MauranKilom on Thursday August 04, 2022 @11:26AM (#32344178)
> It's hypothesized (and, I think, likely) that the machine learns some of those hidden/latent properties, though, obviously, only to the extent that they are useful to recreate the images.

Notably, this precisely describes humans too. We don't know the "true" properties of anything we interact with. We just have models - some more sophisticated than others - but only to the extent that we care for reasoning about the objects. From "This stone is too heavy to lift." to "Oh yeah Becky is always late."

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by icehawk on Thursday August 04, 2022 @05:38PM (#32348626)
Now explain how that's different than anything you look at.

- Your visual system only has access to the 2D geometrical properties projected on your retina. The properties are correlated across images, inexplicably. (I certainly cannot explain how chairs are, in a fashion that includes all chairs I've encountered and excludes anything I've encountered that is not a chair)

- Any other interaction is also a correlation.

- Humans don't have access to the laws of physics, just reasonable approximations in certain contexts.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by Barrin92 on Thursday August 04, 2022 @05:53PM (#32348770)
>Now explain how that's different than anything you look at.

For starters you just don't look at things, you're embedded in the world. You have sensory input far beyond visual information, you also have something akin to cybernetic feedback in response to your mechanical actions in the world, and DALL-E has not.

In fact DALL-E doesn't even have access to visual information in the same sense you have, which is to a large extent biochemical and analog, not digital.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by mjburgess on Friday August 05, 2022 @04:05AM (#32353089)
It's exhausting to have people go around repeating this pseudoscience. I hope it's over soon.

We have accurate causal models of our own bodies, and are directly causally active in the world so we can create experiments which enable us to discover its properties.

We do not live in a schizophrenic nightmare, forever trapped just to correlate pixel patterns. The claim that we do must, imo, now strongly be labelled as pseudoscience.

This is the new homeopathy; and AI/ML based on it is a scientific fraud.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by Earw0rm on Thursday August 04, 2022 @11:54AM (#32344541)
Where this gets interesting is when the AI attempts to create animations or moving images - this requires a much deeper understanding of objects and their physical properties, even to create a highly stylised or "drawn" animation.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by dougmwne on Thursday August 04, 2022 @12:11PM (#32344786)
I reached essentially the same conclusion after playing with GTP-3 for a while. It spins out mountains of convincing and impressive bullshit, but you can’t actually trust anything it says because it is disconnected from right, wrong, correct and incorrect. Reading too much of what it outputs is dangerous because it basically is feeding white noise into your perception and experience of the world.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by acoard on Thursday August 04, 2022 @11:54AM (#32344542)
I'm a big fan of Frankfurt's "On Bullshit", and love the reference.

I think there's one significant distinction between a normal human bullshitter that Frankfurt originally envisioned, and the AI practicing Artificial Bullshit. The bullshitter knows there is truth and intentionally disregards it; whereas the AI is blind to the concept. I guess this is "mens rea" in a sense, the human is conscious of their guilt (even if they're apathetic towards it), whereas DALL-E is just a tool that does what it is programmed to do.

I do like this application of "bullshit" though, and will keep it in mind going forward.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @12:37PM (#32345156)
Interesting, but I have two follow ups

1. What are the implications of intentionally disregarding the existence of truth vs being blind to the concept? How does this distinction you made manifest?

2. Are you sure all humans actually believe in the concept of truth, or could it be the case that some people genuinely function on a principle "there is no truth, only power". Is it possible to think "truth" and "current dominant narrative" are one in the same?

I've certainly had a ton of luck with Bullshit in Diplomacy. As Russia, I offered a plan that involved France helping me take Munich and I would repay by supporting him against the English invasion. Did I intend to actually follow through, or was this a cunning lie? Neither. It was bullshit that got me into Munich. I myself didn't know because (in game) I don't believe in the concept of truth. Everything I say is true and none of it is true. Its all true in the sense that it might happen and I put some weight on it, none of it is true in the sense that there is no branch of the future game tree privileged as "the truth". Some truths have more weight than others, but there is no underlying absolute truth value that must exist yet I choose to ignore. Eventually the order box forces me to pick a truth out of the many I have told. But prior to being forced, it didn't exist.

Is it possible to think in this way all the time about everything? Maybe.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by samstave on Thursday August 04, 2022 @01:52PM (#32346072)
>conscious of their guilt

This hits it for me:

Consciousness is kinda "being aware of the fact that you have choices for available actions, and what the impact of these actions/non-actions will have on either yourself, your environment, the object of your action, or impact on others.

Intelligence is being aware of the inputs and knowing the (non)available list of actions to take.

Intelligence acts on stimuli/input/data?

Consciousness is awareness of one's own actions from intelligence, others acts from their standpoint of intelligence or consciousness...

A yin/yang, subjective/objective sort of duality that Humans make. (thought v emotion)

Dogs are both intelligent and conscious. They know guilt when they are shamed or happiness when praised for intelligent actions..

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by thfuran on Thursday August 04, 2022 @02:28PM (#32346537)
>Bullshit is far more insidious than a lie, for bullshit can (and often does) turn out to be coincident with the truth. When that happens the bullshitter goes undetected and is free to infect our understanding with more bullshit made up on the spot.

If the bullshit is turning out to be true, what's the issue with more of it? If it's not true but still believed and so causing problems, what's the practical difference between it and an undetected lie that makes it more insidious?

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by IIAOPSW on Thursday August 04, 2022 @02:46PM (#32346762)
A lie is anti-correlated with the truth. Bullshit is uncorrelated with the truth. The problem with trusting a coin flip because it happened to be right this time is that now you trust the coin a source of information, when really its a source of entropy. The lie will be found out just as soon as the truth comes out, whereas the bullshitter might get enough lucky flips in a row that people start taking their advice seriously.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by adamsmith143 on Thursday August 04, 2022 @02:56PM (#32346890)
>In other words, its bullshitting you. It was only trained on how to get your approval, not to understand the mechanics of the world it is drawing.

A system can learn to do all kinds of interesting things by trying to optimize getting rewards.

See: https://www.deepmind.com/publications/reward-is-enough

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by visarga on Thursday August 04, 2022 @04:11PM (#32347699)
What you call bullshit I call imagination. Both humans and AI need it. Humans use imagination to plan ahead. AlphaGo was generating moves to plan ahead.

Dall-E and GPT-3 are not being used as agents, they are just tool AIs. They have a narrow task - generating images and text. Agents on the other hand need to learn how to act in the environment, while learning to understand the world at the same time.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by elefanten on Thursday August 04, 2022 @05:16PM (#32348395)
How did you get "imagination" out of gp's comment? They described something quite clearly and it's not reasonably summarized as "imagination".

Per gp, bullshit is cynically self-interested pontificating. It's performance. Maybe you could say that the bullshit produced is imaginative, sometimes. But it has nothing to do with "imagination" as a simulation-like capability used for planning and learning.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by TremendousJudge on Thursday August 04, 2022 @03:05PM (#32346982)
As far as I know, the human brain is just a "social game reasoning" optimizer, that we try (and fail) to use to do actual logical reasoning. The zillion cognitive biases we have are the clue: we don't do logic, we have biases and sometimes stumble upon logic.
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by gfodor on Thursday August 04, 2022 @01:24PM (#32345723)
DALL-E either doesn't generate images you want to see, or if it does, it does a bad job, because it generates many images you don't want to see.

In other words, the claim you've set up is basically unfalsifiable, given that thre's no way to form strong counterevidence from its outputs. (I would argue that if there was, we'd already have it in the vast majority of outputs that aren't images people want.)

If I were to refine what you're saying, is that DALL-E is constrained to generating images that make sense to the human visual system in a coherent way. This constraint is a far cry from what you need to be able to lift it up to claim it is "bullshitting" though, since this constraint is at a very low level in terms of constraining outputs.

Re:Is DALL-E 2 ‘gluing things together’ without understandin... by tomc1985 on Friday August 05, 2022 @12:45AM (#32351788)
Stop anthropomorphisizing computers! They hate that!
Re:Is DALL-E 2 ‘gluing things together’ without understandin... by garyrob on Thursday August 04, 2022 @12:10PM (#32344772)
I agree, but I disagree about one aspect. For the most part, humans don't use reason all that much or all that deeply. We usually use intuitive thinking, and there is research showing that immediate intuitive responses are often better than the result of long thinking. More negatively, is QAnon belief, or even Trump election claim belief, about reason? Or is it about associations between words and concepts, especially when those concepts are believed in by the people the believer tends to trust and associate with?

In other words, the takeaway here may not be that GPT-3 spews bullshit. It may be that most of the time, human "thinking" is a less-nuanced, biological version of GPT-3.