stoniejohnson

So since a model is fine tuned via RLHF my point doesn't stand?

Genuine question; it would be interesting if some other mechanism was at play here.

First off, thank you for the resource, I'll dig into Wood's research + checkout this specific podcast.

I think what would be interesting to study (and is likely covered by Wood outside of this soundbite) is not the steady state of someone with good self-control, but the transition period to having good self-control.

I don't know much, but this isn't surprising based on the little I know.

Transformers predict the next token.

If your question is at the end of the prompt, the start of an answer is a more likely next token than if the question is at the beginning of the prompt followed by a ton of other relevant, but non-question-forming tokens.

Still, if you had to put the question at the beginning of your prompt, a transformer is more likely to give an answer than an RNN.

bruh your OS scheduler wants a word with you

What is hard about things like this is it ultimately boils down to self control.

Sure, by creating the friction tunnel, you are less likely to get distracted online, which makes self control easier.

But the thing is we are smart and computers are fast; meaning most software-based friction tunnels are quick to undo.

I did something similar to what the OP described by solely using the Mullvad Browser, which is in permanent incognito mode.

The moment this introduced friction in some nuanced way that I had not foreseen, I just switched back to my main browser as I could not think of a way to handle this edge case.

For me at least I think the friction tunnel needs to be inherently physical. Using an old computer (with a new one not in the same room) is a much better friction tunnel than changing your browser.

This seems like a very literal interpretation of "spatial computing"; I don't think anyone will be physically manipulating ASTs with any regularity.

> We've only seen marketing footage at this point.

Agreed that it is too early to judge. I was basing my opinion mainly off of tweets from people who used it hands on at WWDC, but still that is a biased group.

I think for each use case how you interact with the headset over a significant period of time will be different. For productivity, you still will use a keyboard. For a video game, you may be standing and moving.

But, no matter what, the AR objects should be touch-able, reactive to your sight, etc. When you want to interact with it using your hands/eyes, it should be seamless, and that is what Apple nailed here.

>"Maybe it was getting the weight off my head. Maybe it was that when I was finally able to see again with my own eyes—looking at my notebook, or my hands or the other people in the room—I realized I was no longer staring at the world through a screen. Even Apple’s crisp 23 million pixels can’t replace standard vision. Maybe that’s the long-term challenge, getting us to forget that difference."

This sounds like a negative take from the author, but it makes me incredibly bullish on this reality-first AR/VR approach (sPatIaL cOMpuTiNg).

Apple nailed the important stuff. As time goes on the hardware will get lighter, screens will get better, etc.

"The cause is algorithmic. This is the algorithm.

EDIT: The post doesn't follow the algorithm. But still, it's algorithmic, stop being conspiratorial."

Look man, in my opinion, in the current political climate this is clearly on the short list of possible bad takes for this tech scandal (hate the phrase, but it fits).

That is why this thread even exists in the first place.

That was my whole point.

You calling me "shamefully delusional" is definitely an ad hominem no matter how many words you try to stuff in my mouth.

See my response to Tremon but resorting to an ad hominem based on an (incorrect) assumption of how I judged the situation is both rude and a waste of time.

I don't appreciate the insult, especially saying I have a "low mind" whatever that means.

I am not racist. I wouldn't wager that this was race related. But I wouldn't say that someone seeing the situation, and then making the assumption that is it is race related, is out of the realm of possibility.

I would think that would be an unfortunate, but frequent, take away if the average person evaluated things at surface level.

You can disagree with me here, maybe with a Gallup poll for example, without insulting me.

Just because this is a sensitive topic doesn't mean you get to throw around insults and look down on another person.

Humans engage in tribalism and racism regularly.

Thus someone might assume something of that sort is at play here, though it might be an unreasonable assumption for reasons x, y, and z.

It might not be reasonable, but it is likely.

I agree with your points.

There was some interview where a smart dude basically said we became enamored with the scientific method and created these strict, quantitative laws/rules/measures. Now we are trying to squeeze reality into these derivations, which is backwards.

Are there fundamental phenomenon that undergird consciousness that we are entirely unaware of? Yeah probably.

You're creating a false narrative man. Him creating tinygrad makes sense given he created a self-driving car company which leverages deep learning (and now uses tinygrad in prod, go figure).

It's not like he decided to hop from self-driving cars to 'AI' because fads changed.

Consciousness does not emerge; that's a trick your brain plays on you.

Really your conscious experience is quite empty/vacuous.

Let's take vision, for example. (this is based on reading a handful of papers without a background in neurophysiology, so mountain of salt etc)

Your vision isn't some continuous, perfect experience. Meaning, if you look at an old smartphone, you can really see the pixels, and you can see the jagged breakdown of the UI if you kinda squint and look close. This reveals the magic trick that it's really just a grid of pixels.

Same with your conscious experience!

Your vision, if you pay attention to the sensory content rather than what it represents, is grainy. It has constituent "pixels".

What are those? This is where I am guessing, because sure we have retinas, which would be an obvious choice, but there are also retinotopic maps in the back of your head in the visual cortex (https://elifesciences.org/articles/40224).

These create essentially a visual homunculus of your first person POV. It is entirely made up of cells, that get more dense as they get closer to mapping the fovea.

Back to my early point about vision being grainy. The point there is that vision has artifacts of it's medium. It isn't some perfect magical substance that gives a continuous, smooth qualitative experience, rather it is grainy, and likely a physical configuration of neurons.

What neurons?

Well, for vision, I think a conscious person's first person POV is identical to the retinotopic map.

When you look around, those neurons are changing.

If someone asked, in physical space, where my first person POV, made up "qualia", really is? I think you would have to point to this retinotopic map in the back of my head. There is nothing additional needed. Neurons can seem to be a 3-D a video game experience if configured in this convoluted way in a biological organism's nervous system.

TLDR: Those grainy movements of your vision? Those are neurons dancing. Your first person, visual perspective is entirely physical.

EDIT:

If they further inquired: well, okay, those neurons are identical to the first person POV that this biological organism has. But why is it appearing to them? Why aren't they a P-Zombie?

Well this neural homunculus (retinotopic map) is in the visual cortex which is connected to the parietal cortex (involved in spatial awareness and attention) and the temporal cortex (involved in object recognition and memory).

So whatever is happening, is a coordination of this visual homunculus with all the other processes corresponding to our other features/senses, with the goal of having a cohesive experience for an agent, you.

So these neurons are made up of a larger global process, and thus you can sit and look at this screen, but also hear yourself reading this sentence in your head as you vocal chords have minute movement, and then stop and think of something. This kind of ties back to my first comment, but the point I'm making here is if you phenomenologically investigate one of the most salient soups of qualia, vision, it becomes pretty clearly it is entirely physical. So the onus falls back onto you, what extra stuff is there?

I don't think it is accurate to think of it as the "output" of information processing, rather I think it is the actual act of diverse, yet integrated information processing.

Simply put, different parts of the brain do different things; there is some level of isolation of function, e.g. Broca's area for speech, the visual cortex for sight, etc.

Conscious experience clearly relies on several disparate parts of the brain (e.g. sight, sound, touch and smell can form a cohesive experience).

The integration of these parts is the act of consciousness, or put in another way, when the brain is highly integrated in a waking state of perception, it is conscious.

That's what the "P Zombie" thought experiment misses, imo. We are not some Rube Goldberg machine of pulleys and levers, as the analogy would lead us to believe. We are also not a series of pulleys and levers with some qualia painted on top.

We are, instead, the global process that arises when a biological agent's neural system is highly integrated with the goal of providing the organism a sense of self and direction.

It sounds pretty grandiose, but it is still fundamentally empty, much like the a system of pulleys and levers. Where do your thoughts come from? What do you actually do? Not much.

Is that not also available in Pandas?

https://pandas.pydata.org/docs/reference/api/pandas.DataFram...

You have to use GPT4, and specifically click the browsing beta option.

Am I the only one that feels like this article really doesn't say anything at all?

Sure most of the work in the computing world is done by scripts (written by humans).

Sure if an AI can magically write/manage all those scripts that would be great.

I don't really see anything beyond someone creating an artificial divide in computing that I don't really see as relevant ("batch" vs using a literal app with an AI feature).

Of course most computing doesn't require a person to be interacting with it.

I'm being genuine here; please clarify if I'm just lacking context.

Sadly I agree as a another user of USPS informed delivery.

What about Claude from Anthropic?

The fact that it is learning online in fractions of a second to prevent falling (I would say) is pretty magic!

Via simulation + meta-learning they can make a blind robot walk over arbitrary terrains solely via proprioception.

They add a visual component, which with 30 minutes of real world data, through a technique called Cross-Modal Supervision (CMS), is able to improve walking performance significantly!

They can even distort its vision (turn the camera) and it learns to adapt!

Paper: https://arxiv.org/abs/2211.03785

Advisor Presentation: https://www.youtube.com/watch?v=f6fDpKDxpL0

Note: This is not my work; I originally found the presentation above.

It's a joke. Jokes are fine. Jokes about anything are fine, as along as they are jokes.

You didn't think this joke was funny, and that's fine. I thought it was funny, and that's fine.

Calling it a "Nazi-ism", is also fine, but if that means he shouldn't be allowed to say it, that is not fine.

See the following picture, it kinda captures the point:

https://pbs.twimg.com/media/FrrVOoFXoAEOWYe?format=jpg&name=...

I think the research discussed here is as significant as the advancements in LLMs.

Great free-form comedy, thanks for elevating this discussion with it, hitting you with an upvote

What search engine have you switched to?

Maybe I've just been tuned to use Google effectively, but Bing/DuckDuckGo results are horrid except for very simple things, while Google gets me exactly what I want in the first ~5 results.