angusturner

Why though? I mean, why can't we make piecemeal reform that forces billionaires to accept 4% return on capital instead of 5% as a small step toward a more equal society?

There's no meaningful sense in which this is true. Having the keys on your computer is neither necessary nor sufficient to establish ownership.

Only functioning legal institutions can establish that, which points to the the main flaw in the entire crypto project: without the institutional trust that btc presumes to replace, society would collapse. And what good are your tokens then?

Early in what sense though? How many trillions have been invested at this point.

"The truth is no online database will replace your daily newspaper, no CD-ROM can take the place of a competent teacher and no computer network will change the way government works"

This sentence seems pretty accurate to me? Like it does undersell the scale of change, but you still can't replace a good human teacher, and people still read books and newspapers despite never ending predictions they are done for

Okay, better answer than I expected to be honest. TIL.

I guess I should refine my question to be about text to media specifically.

Likewise for the orcs I guess. Maybe its just going to be a matter of degree and any test I come up with have exceptions and counter examples.

The redeeming feature of AI right now is that people can discern if you've relied on it too much.

Are we speeding towards a future where most people are effectively wireheaded?

To keep some perspective, I encourage people to name one single cultural artifact (picture, video, game, music, book, etc) that was improved, or made substantially possible, by AI. Ideally something that has moved you or provoked a lasting response.

Then try to think of examples where Generative AI has betrayed a lack of creativity or care from the creator, and thus left a sour or ambivalent impression.

Personally, I can't do the former and the latter is depressingly easy.

The main purpose of wealth tax is not redistribution, but curbing the tide of power concentration.

If large fortunes keep doing 5-7% per year compounding, outpacing the avg. growth rate of economy, wages etc. then we keep ending up with Musk-like figures with such outsized amounts of power and influence that it basically destabilizes democracy altogether.

You can't get rid of these people because they just have too much influence -- i mean, literally buying entire social media platforms. C'mon.

Its really just a question about whether you want society to continue functioning with some semblance of fairness and stability.

The phrase "your success" is doing a lot of heavy lifting here.

You understand that billionaires don't make money from labor right? They get their money from rents and the return on capital. Roughly 5-7% per year, compounding, risk free, for perpetuity.

Is this concerning? Can you explain your point a bit more clearly, for the benefit of people like me that don't want individuals to have more power (whether its in "stocks" or "assets") than entire nations?

I don't understand the problem.

Besides any argument about fairness, wealth inequality is highly destabilizing and historically has only been corrected by revolutions, wars, crises etc.

If a few billionaires need to sell stocks to avoid such an outcome, that seems quite desirable to me.

I have thought about it.

I'm encouraging others to do the same and be honest with themselves about the direction we're headed.

I'm glad you feel this way.

I can't personally relate, and I think its kind of telling that I'm being down-voted for simply asking people to engage with this thought experiment.

Without getting into a larger argument about whether this is the correct approach to building AGI, I think its reasonable to ask people to engage with how this technology gets used in practice, besides its nominal purpose of "building AGI".

The predominant way people engage with this technology right now is through slop content in their instagram feeds.

(Btw I say all this as someone that has worked on generative models for 10 years).

Software might be the exception, but I think some skepticism is still warranted.

When software is hard there's a kind of filter; a good piece of software means a human/team behind it with grit and care and user empathy. You can have some* confidence in the continued stability and maintenance of the software. (*relatively speaking, and with huge error bars).

Without knowing the details of Fluxer, if someone tells me about a vibecoded Discord clone, my immediate thoughts are like: - how much thought and care went into this? - how much work from Discord has been cannibalized? - will the authors be around next week to keep maintaining and making improvements and thinking deeply about their users?

I have also seen a lot of fantastic demos and stuff with AI coding. I use AI coding all day. I just think there's a weird gap still, where code and images and text are all "cheap", but none of this has translated into anything I personally use or care about.

I think its a useful exercise to try and name a cultural artifact (picture, video, game, music, book, etc) that was improved, or made substantially possible, by this technology. Ideally something that has moved you or provoked a lasting response.

Conversely, try to think of examples where Generative AI has betrayed a lack of creativity or care from the creator, and thus left a sour or ambivalent or impression.

I find the former impossibly hard, and the latter depressingly easy. Do others feel the same?

Yeah but also despite being the richest we've ever been on avg., somehow we're working more hours than ever. Especially in well paid disciplines.

But more to the point, it seems that 2 or maybe 3, > $1 trillion dollar companies are poised to capture all the economic value from all knowledge work ever, despite doing it off the backs of artists, programmers, writers etc.

(Assuming they don't collapse the financial system , internet or power grids first).

I mean, clearly "tooling" is not a perfect analogy for AI. As others have pointed out, if you pay money and get back a full website, app, game etc. it feels a lot closer to a service.

How many of those jobs do you expect them to retain as they build increasingly capable models?

How much of the profit will be redistributed to all the artists, programmers, scientists, etc. whose works are the foundation of their product?

The poorest 50% of global citizens hold the same combined wealth as the top 10 individuals.

How are massive and increasingly concentrated corporate profits going to help them? (Or anyone else for that matter).

My puzzlement is that people object to attributing agency to things that are clearly "agent-like". That doesn't mean the developers and companies building it get a free pass.

Mightn't be so bad if all the value wasn't being captured by trillion dollar corps, despite them benefiting off the labor and creativity of countless non consenting individuals

I mean.. There are degrees of control and agency right? I really don't understand the impetus to pretend AI is just doing exactly what its told.

Presumably this is only a bright point if you dont think its false hope?

This seems highly optimistic to me.

What's to say current AI won't be highly power-concentrating by default?

The best models are owned by a few companies, and displacement of knowledge-workers mainly seems to benefit the capital class.

On-device / edge computing makes sense in a few very limited scenarios. And economically, price or watt/token (or watt/task completed) might always be better in large data centers.

No reason to think we are on a trajectory towards broad empowerment right now

Define LLMs. Current models look pretty different to the vanilla transformer from 2017. And its not even known what architecture and training innovations the frontier is using, right?

Its turtles all the way down

People want human art. Probably they will want it more than ever in post AGI world.

What does this even mean. There is strong evidence of LLMs doing in context learning.

Some of the linear RNN layers in recent models are provably doing SGD in hidden space during inference

What unique differentiator could musk possibly offer though?

Is there really a high demand for racist chatbots with no safety controls?

And even if there is, will governments even allow it to keep operating?

I find the presentation of this as "factoring training" very confusing.

Factoring is something we do to distributions, which we then parameterize with our neural net.

An autoregressive model pertains to a certain factorisation of a joint distribution. And if we choose this factorisation it may have consequences for our training and sampling etc.

But to say "most models on factor generation".. I dont quite get this.

This would be better work if it didnt claim it was a new pretraining axis also. Its another way of scaling compute right?

If it were a new axis, IWAE would have an equal claim to it (as others have pointed out).

I immediately thought of IWAE as well.