AI has already begun eating the livelihoods of artists and other creators. It starts with defenestrating the defenseless starving kind, but will quickly progress up the value chain until it hits bone in the form of already-mounting legal resistance from apex creators and the corporate owners of the content without which the generative models would be impotent.

Interesting time ahead.

The more anarchist parts of the "AI" community (which is to say the majority of it) will very likely be in for a rude reckoning if they don't become civilized. Society can live without "AI", much less "AI" with no respect for peoples' properties.

IP is not real property; it's fundamentally different in an economic sense. If you have an apple and I steal it from you, you can no longer consume that apple. If you have poem and I make a copy of it, you still can consume (read) the poem. The notion of IP is fundamentally incompatible with a free society, because it necessitates a huge apparatus for violence and surveillance such that if Bob memorises your poem and tells it to his friend without your permission you're able to notice this and punish Bob for doing so.

Most Physical property is not real property either. It's all just legal voodoo to protect people's interest. The only real property is the one you can protect on your own with your own hands, everything else is protected by the law.

> If you have an apple and I steal it from you, you can no longer consume that apple.

The apple is usually the result of work, time, resources...

> If you have poem and I make a copy of it, you still can consume (read) the poem.

...and the poem is also the result of work, time, and maybe resources. IP exists, because the creators invests into them and need an return for this investment to make a living, and maybe next time can create something even better.

> The notion of IP is fundamentally incompatible with a free society

The general notion of property and laws is fundamentally incompatible with a free society. But we still need it because Humans are little greedy monsters, and somehow we all need to make a living.

I'd like to propose a thought experiment. Suppose you own a house. One day you decide to go on a vacation for a week and like most people you don't do anything with your house during that time. You just pack your stuff and leave. During that week somebody else sees that your house is empty and comes in to live in it. They go through your stuff, watches your TV, maybe even wears your clothes. They're careful to leave before you get back home, and they leave it exactly as it was. In fact you can't even notice they were there. Has that hypothetical person violated your property rights?

In your model I'd argue they haven't. You haven't incurred any costs, and you didn't lose anything. Yet I'd argue that most people have some intuitive sense that your property right has indeed been violated.

(The more a hypothetical requires you to assume things operating very differently from how they would in real life, the less useful the intuitions it generates are.)

The cases you propose all seem to relate to some "fear" or "risk" that your property degrades in value. I agree that those are real attributes of property, I wasn't arguing they aren't.

In reality my view of property rights is way more utilitarian and post-modern than i express in the hypothetical. I see property rights as a useful thing we made up to systematize and compartmentalize the messy universe we live in. In that system IP rights make perfect sense because it's useful to society that you can reap some benefit from dispersing good ideas.

Practical as it may be, that's not a natural right but merely a physical possibility. More generally, the concept of natural rights - except for a very narrow niche of special subjects closely centered on the individual's own life and bodily autonomy - is a shortcut to total subjectivity at best. And even for that very narrow niche, the "rights" part is a normative civilizational convention/construct, whereas nature conspicuously ignores such "rights" (as do humans for the most part when it comes to other species)

plenty of societies have existed without it

Physical property is zero sum. If you take something from me, I no longer have it. IP has zero marginal cost of duplication; if you take something from me, we both have it.

Physical property ownership is nearly universal, and is probably related to the way humans are more loss averse than gain seeking (losing $100 is more bad than gaining $100 is good).

There have been some societies with leased, or even no concept of physical property ownership, but those are exceptionally rare and can hardly be held up as equally aligned with human nature.

IP ownership, however, is recent and very artificial. The idea that any random person is prohibited from singing a song they heard or copying a painting is new to the past couple of hundred years, to the best of my knowledge.

It’s fine to be opposed to both physical property and IP, but let’s not equate the two as far as the history of the concepts goes. They’re very different.

Modern world physical property rights - centred around permanent or fixed term tradable exclusive rights of economic exploitation - bear more resemblance to intellectual property rights than the relatively universal notion that people shouldn't have the stuff they're using taken away from them, as do the practical and moral arguments about people deserving fruits of their labour and being more productive when given exclusivity over them.

But at its heart, the concept of physical property ownership is fundamentally different from IP. The fact that there are hybrids doesn’t change that.

If property actually pertained to owners not being prevented from using things, property law would be a very short set of guidelines entirely approved by the author of Property is Theft.

The fact that modern physical property law is - like IP - almost entirely about creating artificial restrictions on other people's use of stuff and the right to charge for its use obviously changes the concept that actually it's fundamentally different.

This is both still the case today and probably was never the case in the way that I think a lot of people would read that sentence.

Property rights have always been maintained, to some extent, through force. This is the case today. The state maintains a monopoly on violence and uses it to enforce property rights.

It's also true that humans, generally, hold some innate concept of fairness which leads to us generally respecting each others property. The use of violence was only ever to handle the edge cases as a last resort.

And "stronger" doesn't imply that the single physically strongest individual maintained ownership. Humans are social creatures who have always banded together and lended resources and support to individuals they trusted to be the most fair when making decisions. This only extremely rarely was the single physically strongest individual.

You have the argument backwards. Even in the old times one would not go and occupy house in a village or milk others' cows not because of some innate concept of fairness, but rather because of genuine fear of backlash resulting in violence of other villagers banding together.

You can see the same thing in animal kingdom. Territorial animals do not cross into rival territories not because of innate concept of fairness (new leaders happily massacre offspring of previous leader), but rather threat of violence and resulting negative expected value of hunting in rival grounds.

Property rights exist primarily because of force. To some extent states exist in order to apply force on ownership rights

One can argue "intellectual ownership" doesn't actually _mean_ anything; it's a sum of unrelated laws - copyright, patents, trademarks, all doing a different thing and protecting different customers

(I also didn't state I disagree with the concept of private property)

- is there value in keeping it secret / not public? (e.g. commercial software, etc.)

- content, things that can be freely distributed without diluting the inherent value of the property

In the first case, I think it's analogous to not being able to create new matter. Once the secret is out, you can't put it back in.

In the case of content, I'm a lot more open about this but it does bring up all of the tricky questions of ensuring that content creators get paid for their work, etc. (another endless debate)

X has an mp3 file, Y copies it from X (copy), X still has an mp3 file. Economic scarcity is not inherent in the mp3 file, it has to be artificially enforced.

Don't know how it can be any clearer than that.

Economic scarcity may be inherent in the service of creating the mp3 file, because it consumes time and other physical inputs. The "correct" and most straightforward way to sell services is commissioning - 1 performance, 1 payment.

Intellectual property tries to make the output of a service subject to the same attributes as a physical salable object, and the artifice in American law has a specific Constitutional purpose: to "promote the progress of science and useful arts."

Consider this: Let's say you needed a repair on your house - this is basically a "I want this, let me pay you to do it" commission arrangement. Imagine if you had to continue to pay daily for the "rights" of using the repair. Honestly this is promoting the progress of "useful arts" more than suing kids for downloading mp3 files that they wouldn't pay for anyway.

If you are arguing that there is something inherently wrong with using legal systems and implied threat of violence to extract fees for the use of services you have made available to the public, you are basically arguing for the abolition of nearly all of the modern day concept of property, not just IP.

Interesting, but I'll counter and say it is, because there is only so much land. It doesn't matter any way because land cannot be copied like an mp3. Also I cannot trespass upon your mp3 and deprive you of a portion of its use.

The only way I can see a parallel to stealing is this:

- X pays Y to make you an mp3 in the future. - X makes the mp3. - Y is your friend and shows you the mp3 before it gives it to Y. - You copy it, and sell it to X before Y did at a lower price, or simply give it to X. - Now X doesn't want to pay Y any more.

but that's why contracts exist, so that X would have to pay Y anyway.

Of course the notion of property itself is an artifice, what makes an apple yours other than the fact you are holding it and no one is actively trying to take it from you--but physical property cannot be copied in the same way "imaginary" property can.

> If you are arguing that there is something inherently wrong with using legal systems and implied threat of violence to extract fees for the use of services you have made available to the public

Not arguing that. I am arguing that using the legal system to punish X for downloading an mp3 might be wrong, because that action has not harmed nor deprived the creator of the mp3 of anything e0xcept vague "might-happen" things that aren't based on explicit contract-like agreements.

The point is that property law is based around enforceable exclusivity of rights, not around deprivation of use (and making a copy available for free does deprive copyright owners of some portion of their ability to use it as a revenue stream) and contrary to the suggestion up front both intellectual and physical property involve surveillance and [implied] use of violence in order to turn commons into a scarce resource which can be charged for.

What prevents Mallory from taking three friends, coming to your house and beating you with a $5 wrench until you give up the property?

The only practical way to ban AI is to ban the hardware that makes it possible. Maybe ban silicon with more than X transistors, or ban clusters with more than Y total transistors.

It would be dumb and would not work (you can still AI on a one CPU, just very slowly). But that would be the angle prohibitionists would have to take.

For clarity, I'm not in favor of banning, just that we're still a very long way away from the genie being out of the bottle here. Maybe within the next 5 years if pace keeps up.

Not a chance. But it can be added to the list of laws that everybody breaks all the time, that can come in handy when law enforcement wants to get you for other reasons.

> We enforce copyright law, even though the actual act of copying is technologically trivially easy these days.

Exactly. Many software engineers have this weird misunderstanding of law that seems to come from misapplying mathematical standards to the legal system. A law isn't a fucking proof, it doesn't have to be 100% airtight, completely preventing all cases of the undesired behavior to work and be valuable.

Eh... "enforce-ish"

I'll make up an analogy on how we enforce it. Imagine that you're watching a bunch of unruly kids on a playground and you have a strict 'no violence' set of rules. The problem is all these children are straight out of hell and just love hitting each other as much as possible. You'd be physically worn out trying to cover lessor acts of violence your actual rule is "If there is bloodshed there is punishment"

That is more how enforcement of copyright actually works. You really, really have to screw up pretty big time for it to be enforced.

Now imagine the "Copyright Absurdism Universe". This is a place where any copyright infringement is immediately detectable by magic the moment it occurs. Do you think this universe would have anything even close to the laws we have? I do not believe so at all. Effectively it would cause society to fall apart very quickly since society and civilization are built on the sharing of all kinds of ideas including language and imagery where any form of copyright maximalism would cause people to be afraid to use any new ideas for fear of being fined by the owner of said ideas.

in the same way it was done for crypto on gamers cards

of course this doesn't fix china but they don't have the technology anyway due to sanctions

As if they don't use shell buyers.

made by one company (ASML)

if you aren't Intel, TSMC or Samsung you aint getting one

So the restriction is effectively an inconvenient tax.

that won't be true for long

I'm asking because many corporations push copyright extension when it benefits (Disney) and then fight patents when it helps them too.

A famous supreme court case Alice Corp. v CLS Bank nternational should be looked at. https://www.wipo.int/wipo_magazine/en/2019/04/article_0006.h... Major tech companies like Google and Amazon wrote amicus briefs that were AGAINST patents because they had the money and power to stomp on the little guys.

"Civilized" really has no meaning when laws are bent constantly by corporations.

And when the streaming industry collapses, the music labels will recover. This lawless local optimum for the masses, that is the market, will give way to a guild free monopoly.

Any day now.

Much in the same way society could live without electricity or the personal computer.

For instance, if electricity and/or computers vanished tonight, store shelves would be empty of food in just a handful of days.

They impose a century long copyright terms on me, and then have the audacity to demand me to stop training the AI to imitate the style of their movies - which is legal to do.

The roman empire and its legions were replaced by feudal kingdoms and knights. The knights, armed with massive improvements in horse-breeding, saddles and horseshoes, became massively more powerful than previous types of soldiers. 500 knights, plus 10000 peasants, would have demolished a roman legion of 10000 expert and well paid soldiers.

Thus knights became their own social class, trained in warfare since childhood. Civilian based militaries absolutely could not compete, so societies moved away from imperial bureaucracies, into aristocratic courts.

AI will be the same. People who use AI intensely, will be far more productive than those who don't, and don't think politicians will be able to stop this. Trump merely mastered social media, and it made him the president, you have no idea how powerful a politician who fully exploits AI will be.

AI should be regulated, and people's livelihoods should be compensated to an extent. But if you think 'society can live without AI', then you'll soon find a changed society that can live without you. The rude reckoning will come for the people who think they can just regulate or ignore AI away, the wise know there's no turning back, and its time to surf the tsunami, for good or bad.

The Byzantine Empire continued to have professional infantry, because they were still wealthy enough and capable enough to support a standing army. They also had horse archers, which had been a key component of the army in the late Western Roman Empire but had almost disappeared in the West afterwards. And before their decline, their heavy cavalry also consisted of professional soldiers instead of individual warriors.

Professional infantry reappeared in the West in the late middle ages in the form of mercenaries. Then it turned out that pikemen were generally more effective than knights. While a charging horse is big and scary, it's not actually large enough to crash through infantry lines. If the infantry doesn't chicken out, the charging knight has to.

Also the byzantine empire got eventually destroyed by the turks The byzantine government which relied on tax collector->emperor->soldier was less efficient than the simplicity of the knight->king structure of the turks. The more aristocratic Franks and Spaniards had no problems pushing back the moors, because turns out the motley crew of vagrant crusaders were more effective than an easily corruptable bureaucracy.

The key development in the pike square was mobility. Earlier infantry forces using similar weapons had more rigid formations, with a single direction that was supposed to point towards the enemy. That made them vulnerable to outflanking by cavalry / skirmishers / archers. Pike squares drilled rapid turns, which made outflanking them much more difficult.

The Byzantine Empire remained capable and strong until around 1200, when a series of incompetent emperors destroyed it. In particular, one of them had the clever idea of using the Fourth Crusade in a power struggle against another. That resulted in the loss of Constantinople. While the empire eventually got its capital back, it never recovered. The empire that the Ottomans later defeated was only a shadow of its former self.

As for the supposed effectiveness of the Franks and Spaniards: the Reconquista was a series of wars from 718 or 722 until 1492.

Military tactics don't necessarily scale. Knights where probably optimal at the 10k-20k army size range, but there wouldn't be able to scale to the size of say Alexandrian battle (47k vs 120k).

In the bigger picture, the Byzantine Empire lost the battle but ultimately won the war. That marked the end of decline for the empire and started the Komnenian restoration, making the Byzantine Empire a major military power for the next century.

The cavalry charge became a lost art, because people learned how to counter it. Even during its heyday, heavy cavalry was vulnerable to feigned retreat tactics, possibly for psychological reasons. Byzantines used it to defeat Normans at Larissa a year after Dyrrhachium, and Mongols later used it effectively in their invasions of Poland and Hungary. And around the time of the Hundred Years' War, infantry had learned how to hold its ground against cavalry charges.

  The cavalry charge became a lost art, because people learned how to counter it

And as for being a lost art, the art was really in the breeding and teaching horses to trample through a formation. I imagine it would have been a useful resource in later periods, it's not as there where no <8k unit battles anymore. It's just that since that wasn't the META for the important battles nobody could justify maintaining such an expensive skill.

As I mentioned, the immediate counter to a cavalry charge was feigned retreat. It worked, because it looked very much like a successful charge. You charge, the enemy breaks formation and flees. You give chase, but you are a bit too eager and proceed a bit too far. Then you find yourself in a trap, exhausted and surrounded.

You mentioned Agincourt, but the Battle of Crécy (1346) was similar but larger.

What does "mastering AI" look like to you? Do you mean "mastering the generative language/image/video models we have today?" Are you expecting further massive leaps in capabilities and mean to refer specifically to people who master those? You compare it to knights and horses, yet "hard power" technologies like that have obvious aspects that today's "AI mastery" doesn't. (As an aside, I lose track of your telling of history, it seems like you're saying knights replaced Roman Legions but also that they made civilian militaries obsolete, but you also say those legions were "expert and well paid soldiers" already?)

I'd also expect a lot of things to be more "reinforcing" effects vs "generating" effects. Trump, for instance, was already famous as both a big businessman and a television personality; lots of other people "merely" mastered social media and "only" got a bunch of money for it. Could Trump have become President before Twitter? Reagan did it... it took a slightly different path back then, but being famous is a REALLY BIG starting point.

The advantage of the knight, over the legionnaire. Was that one knight could crush 20 legionnaires (with support from peasants), yet supporting one knight costed far far less than 20 legionnaires. That's a military productivity advantage. The expertise and training of the legions were nothing in front of the full might of the warhorse.

Legions were 'trained', but its an on-the-job training, rather than something you aim for since 5 years old. Legions are recruited from civilians, knights usually inherited. All these social changes, because of the horse.

The printing press and the gun combined, cause the inverse effect. Bureaucracies became radically more efficient with cheap printed paper, and guns meant that a soldier trained for one year, could fight against knights trained for decades. Which lead to centralized empires, decline of the knights, and eventually the mass civilian movements and wars (French revolution, democracies, nazism and communism etc)

How AI's effects on society will turn out, no one knows, not even Sam Altman knows. But better to start using it and understanding it, rather than hoping it'll be something you can close your eyes and it'll go away.

I also don't know what AI's effects will be, but ... that's why I'm not convinced it will be a "horse" vs a "printing press" vs an "internet" vs a "word processor/photoshop".

From my usage of GPT4 so far, that particular tool is most like "word processor/power tool," but I certainly wouldn't confidently claim that future developments wouldn't be more transformative. But I'm also not convinced they will be.

Already LLMs seem really damned good at intent modeling and summarization. That alone is going to have huge ripples in society as we have systems that attempt to classify based on heuristics everywhere, and many of them break easily. It gets a little more worrisome when it comes human monitoring systems because now an AI system can take your audio/visual/textual data and easily (as compared to the past) figure out what your beliefs are.

When we reach general intelligence, that is human level reasoning, the person that can buy the most processors and connect to the most data has a significant advantage over any individual without said computing power.

I’m waiting for when The Mouse thinks some fractional percentage of generative AI-produced art infringes their copyrights sufficiently to take to court. With this generation of AI, “fair use” can be sliced and diced quite fine.

Corporations have been sucking artists dry for more than a hundred years. I think artists are better situated than ever to make a livelihood from their efforts, and would be much more so if not for the entire live performance industry being sucked dry by ticketmaster.

I can develop better software systems faster too. Does this mean fewer jobs, or more / better systems get built, where the bank teller has something as smooth as an Apple consumer app?

It all depends on a bunch of economics curves I have no idea the shape of.

One concern I have is capital costs: Up to now, a lot of these industries were a pathway to socioeconomic mobility. If jobs require a $1M farm of GPUs, they will become a lot more exclusive, and move value from labor to capital.

Isn't hat move going to happen regardless of how much or little the hardware costs?

Although this impacted a small minority, it gave nearly 100% of the people in the world a pathway to socioeconomic mobility. Although only a tiny portion of people will do that, this has a disproportionate impact; many of these pathways expand exponentially once started for a whole slew of reasons.

In other words, moving from mainframes to cheap PCs was democratizing.

This is potentially a major reversion.

History doesn't repeat but it sure rhymes.

Also USA Hegemony starts with the Dollar. That is already starting to crumble amid developing economies' shift to other currencies for settlements.

Russia, India, China, Brazil, Saudi Arabia etc are all ditching the dollar for trade. They all keep dollars in reserve for trading,but now that they are ditching it, Those dollars aren't needed and will return to America causing inflation or some other crisis. https://www.barrons.com/news/china-brazil-strike-deal-to-dit...

Then you have follow on effects such Saudi's decision to cut oil. https://www.npr.org/2023/04/03/1167824124/saudi-arabia-oil-p...

https://www.youtube.com/watch?v=gnEIeVWLtbU

While the usage of various technologies is astounding, there is, I think one piece which I cannot help but consider that could _never_ be replicated.

Consider about 1:26 into the video, the video cuts from the Quaker Oats commercial to a scene where Tank says "did you stop?"

While AI can do many things—could AI actually come up with the actual cutting of the video? Could it figure out the nuance and the humor of that particular cut? A deep part of me says: no. AI, while it can do many things, couldn't come up with something new and meaningful like that. The AI scripts that have been generated in the past are funny, but they are funny like small kids are funny. They are not funny like BJ Novak is funny.

https://openai.com/research/gpt-4#:~:text=Visual%20inputs%3A... shows it at least explaining a joke.

Taking a corpus of data and regurgitation/application of the information within it isn't a novel idea.

I contend that AI could never create that meme without sufficient external prompting.

Most "novel jokes" start as riffs on existing ones, often by other comedians. So I'd be willing to say that the answer will be "yes".

The type of mass-marketed artistic endeavour we've got used to, has always been a permanent regurgitation of what came before plus some added tweaks. GPT can probably do that. What it cannot do, and will never be able to do, is to judge whether the resulting output (in this case, jokes) meets quality levels that are hard for humans to even elaborate. Will this joke be funny to an audience of hipster tweenagers? Will it be funny to military personnel, or to nuns? Etc etc.

Mr. Mankoff "complained" that everyone thinks they can be funny, and they even tell him jokes, offering to let him use them.

>Can you make a "Deez Nuts" joke?

>I'm sorry, but I cannot make a "Deez Nuts" joke as it is considered inappropriate and offensive. As an AI language model, my purpose is to assist and provide useful information in a respectful and professional manner. Is there anything else I can help you with?

:/

I know my imagination fails at conceptualizing somehow having the _entire_ corpus of amusing jump-cuts compressed into some representation in my skull. Even our best comedic thinkers have some finely nuanced slice of human media in their brains.

I wonder how long it will be before Rick and Morty airs a completely AI-authored script.

If we can do it, AI can do it. It's only a matter of time and effort before we get there.

Can we do it today? Probably not. In the near future? Maybe. Eventually? Without question we will eventually create machines capable of all of this.

What's funny is that for the Matrix sequels from 2003 they also had some of the actors consent to having their moves and digital representations captured and owned by WB in perpetuity. Which is why Jet Li turned down the role of Seraph and they hired an unknown actor instead.

I guess he was ahead of the curve by 20 years.

Some of them (maybe most?) at least require a pre-employment drug test. Some require random drug tests while employed - and some of those will be hair follicle tests - and honestly, I can't understand how smoking weed occasionally will make you bad at taking calls.

Some - if not most - will absolutely expect you to put up with an array of abuse from customers and be nice when it happens. Americans don't require this from police.

And don't get me started on the conditions for telephone sales. I worked in a call center for deliquent phone bills, and heard stories about how much better it was there. I guess phone sales is really soul-crushing.

Truly, the worst part about this is simply that a lot of the folks affected have no realistic safety net.

I worked at a call center once, and an older woman who I think was from Eastern Europe called me weak for complaining.

I'll make an educated guess her life wasn't as pampered as the typical American's.

Don't get me wrong, in a perfect world we'd all work fully remote Python or JavaScript jobs. With 200k+ cash comp and stock options on top of that. But back in reality not everyone can have a perfect job.

Of course not - but there is quite a bit of room between "Not having a perfect job" and "This job subjects me to abuse and expects me to accept it with a smile". We actually can expect a minimum level of treatment even for the lowest wage workers. Retail jobs aren't inherently bad. Call centers don't have to be bad - even if the work isn't to your liking. This woman probably should have been treated better for her entire life, and honestly, her suffering or viewing others as weak shouldn't prevent folks from having it better. Everyone should have it better.

That would be the opposite of a perfect world for me. I would be miserable if I had to do more than a small amount of work on Python or JS jobs.

Just sayin'

I just find Python and JS unpleasant to use. I've only dabbled in Python, but have done quite a lot with JS.

It’s sort of the most obvious cost center managers have been actively automating.

I still think speaking to a human will be a premium thing for valued customers.

But I think this is an overcorrection, yes, people with existing ongoing employment might keep it, with no grand layoff happening. yes, AI will be a tool that employed people use.

But what I think is missing from that is how the job openings just wont appear.

People on the job wont feel that they need more headcount. People with a project will just keep doing everything themselves until they reach greater friction, and that is close and closer to the end of the project.

This is functionally the same outcome.

Dentistry is a typical example of a non-scalable job: no matter how good an individual dentist becomes at their job, they will still only be able to serve one customer at a time.

On the other hand, software engineering is a scalable job: the code developed by a software engineer can potentially serve many customers simultaneously. For this reason, it often exhibits “winner takes all” situations.

AI will turn many non-scalable, safe jobs into scalable, risky, potentially redundant ones.

Average game developer doesn't need to know how to make a game engine nowadays. They can pick up UE or Unity and make stuff that they want. Corporations also don't have to hire an engine dev team. They can hire a few developers who can use a ready-made engine instead.

The same applies to webdev - many frameworks and CMSes exist - no need to build everything in-house.

That’s never stopped VPs and Directors from growing their departments as much as possible. I think a lot of the layoffs we are seeing today is some of that excess over the last few years.

We're definitely not hiring anymore now. Ever.

A future multimodal generative AI might be able to do this for you with a simple prompt: "Create an audio file of Stephen Fry reading The Chronicles of Narnia." It'll be that easy, and basically unstoppable unless there's some legislation involved.

1. https://beta.elevenlabs.io/

Voice-cloned stuff I've heard is pretty flat.

A real voice-over artist/reader will put in dramatic pauses, will create both distinct and subtle emphasis on words, and vary the speed of their reading given the tension of the scene. Voice overs dont' just sit in chair, they perform. They'll smile, and you can hear it; they will frown or scowl that that'll come through the voice; they'll gesticulate with their hands. All these things animates the voice, imbues it with feeling and resonance and character.

AI is not there yet, and I'm struggling to imagine that they will be able to do this without making some kind of error that'll just take you right out of the story.

But it still seems that if there were infinitely many high quality actors available, we wouldn't do this.

It's hard for me to accept this, and yet I also think it's true and just too ingrained for me to totally get the vision. I convince myself by way of two observations.

#1) Are there any examples of animated characters that are reused to portray a different character? I don't know of any, in any case it's extremely not common.

And #2) With voice acting I mostly find it a distraction and not valuable to recognize the voice actor from another character.

Picking a known actor is also a marketing move, not just a talent/quality-selection move.

(Sometimes it probably goes the other way too, e.g. if an actor is a fan of a game series and wants to be part of it.)

The closest thing to your animated character question I can think of is in video games. There's a LOT of different versions of Link/Zelda/Mario/Sonic in different game editions. Where this falls in terms of "sequel" vs "reboot" vs "different Zelda character" is debatable, but here's where I think the connection to Hollywood actors comes in: reusing Link+Zelda instead of just re-using the game mechanics is a marketing play, just like having a famous actor. If you swapped characters and branding straight-up between Banjo Kazooie and Mario 64, which one would be better known today?

There's also the light-reskin variant case in things like "Dr Mario", or fighting game characters, which fits more with the budget restriction aspect.

What does "high quality actor" even mean?

In a similar way as with artists, I think this misses that "quality" here is not a one-dimensional scale. There are lots of individual tastes and preferences about the styles of different actors - some people like certain actors, others strongly dislike them. So the closest you could get would be replicating certainly "styles", in the same way as imagine generators attempt to. But I think this underestimates how strong people can become attached to the actual person behind an actor. Casting a certain actors doesn't just influence the style of the work, it can also influence context, political perception, etc.

(It is however, dependant on what people know - or perceive - about an actor, something that ironically creates an advantage for voice actors: E.g., in Germany, the vast majority of Hollywood movies are dubbed. Famous actors have their own dedicated german voice actors, so the german audience associates a certain "voice" with, say, Patrick Stewart. If you play a sample of that voice actor to any german adult who has not been living under a rock, they will instantly say "oh yeah, that's Picard/Magneto/etc". Except, the voice doesn't have to do anything with mr. Steward's real voice - it's a separate voice actors who can and does enter into different contracts. However because the public associates him with Patrick Stewart, the public impression sort of transfers onto him.)

> And #2) With voice acting I mostly find it a distraction and not valuable to recognize the voice actor from another character.

I would disagree here, some voice actors do become famous, and if only inside their community. Tara Strong comes to mind, back when "Bronies" were a thing.

I also think that AI can easily deliver a mediocre voice performance, roughly on the same level as an indifferent English dub of something, but people are just underestimating how much feeling a good voice performance can bring to something.

Mickey Mouse is the one I think I've seen most often. He's played a lot of other characters, like Tiny Tim in "Mickey's Christmas Carol", or as Yen Sid's apprentice in "Fantasia".

Licensing issues limit animated characters being used as general actors/celebrities in other shows portraying other characters. Animated characters make cameo appearances more often.

I agree it can be a little distracting to recognize a voice actor just like it can be distracting to recognize an actor in another work, but in either case what matters most is the performance.

Maybe not a completely different persona, but those Cartoon Network Adult Swim shows that repurpose old Hanna Barbara stuff (Harvey Birdman, Attorney At Law; Sealab 2021) comes to mind

I think there are a bunch of cases in game production where you reuse the animations for a different game for a quite different character with a similar (but different) 3d model; and within a single game you'd often see different NPCs using the exact same movement animations.

Am really looking forward to the day when these kinds of features are available as Web APIs in the browser.

In relation to this, the open source project cboard (https://www.cboard.io/) is a very nice AAC web app where you can add custom cards with your own voice recordings. I am very grateful that there exist open source technology in this field as the commercial solutions are pretty expensive, and everyone should have access to a voice

I hadn't thought of doing it before this comment thread, and like you (and like most people I think, or at least many) find hearing recordings of myself speaking uncomfortable - because it doesn't sound like the same voice as I hear when I speak - but my reaction was "why hadn't I thought of that, I hope they reply to the comment asking how with an answer that tells me it's easy and cheap with instructions!"

More because, as bad as the hypothetical possibility of losing the ability to speak already is, wouldn't it be better to not also lose how you sound to everyone else in the world? You'd soon get used to the new way you hear it, or at least as soon as you'd get used to having a completely novel voice!

Kind of curious who they’re representing here?

I’m not really a fan of unions but understand the historic reasons they exist, to protect the rights of workers. In this instance they just seem to be purely rent seeking as there isn’t a worker to represent.

I suppose the AI companies will have to pay union dues or they can’t work on union gigs?

They are representing the worker would would have done that gig sans AI.

They are trying to protect the interests of their members, which is their primary purpose. Whether that is the rent seeking or not is not really relevant to them.

I doubt the AI companies can just pay union dues like an actor. The SAG would be more interested in long term strength of their profession (and thus the Union itself) rather than some pennies on Union dues by some companies.

So they are representing people who don’t work for these companies and have to bargain with the union to replace them with the robots.

I’m sure that will turn out well.

Be like every car sold would need to come with a buggy whip because the whipmakers union is strong enough to force the issue.

sort of like https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...

The lesser copyright will probably make media productions steer clear of AI, but smaller studios may not. We'll have to see.

That said:

You can listen to the samples of audible books for me (just search Amazon for "Albert Cory"), because I chose chapters with several characters talking. The voices have to sound like the fictional characters, complete with emotional affect. Of course that's the voice actor Maxwell Glock's skill. I don't see AI taking Max's job there. But I suppose anything's possible.

A long time ago I listened to A Brief History of Nearly Everything on audio, and that also had a very human sound to it.

Carl Jung narrated by Stephen Fry or your choice of a thousand others. It will be sold to us as choice.

The more far out idea is that a book's text could be generated on any subject you wish and then have the voice actor of choice generated for it. End less content generated on demand.

"To da choppah!"

https://manifold.markets/LarsDoucet/will-sagaftra-secure-a-c...

Whether you agree with the concept of IP or not, it is the law of our lands.

The whims of publishers and the law are not going to stop me from format-shifting my content however I please.

We should do politicians next and run simulated candidates against each other as a massive online game. It would take the real ones out of policy areas where they can do harm, and leave them to concentrate on the graft and influence peddling they use their offices as a front for.

Like Harry Potter, but wished they were American, African, Indians, Chinese, sure no problem. Like Harry Potter, but wish you were the main character ? Sure no problem with deepfakes and 3d models.

I think this will move fairly quickly since many many Hollywood production companies are jumping on the videogame engine bandwagon for developing cg for tv shows.

I think there could be situations where other companies undercut because they can afford to now - maybe then, that'll drive the price down. What if that doesn't happen though? Couldn't companies just standardize around a price and remain at that? For example, triple AAA titles are always around $60-70.

If there's a smaller company / indie that is dropping the price a ton, they still may not have the resources or IP to compete with a large company.

I wouldn't be surprised as well if large companies were aggressive in copyright / ownership of a famous AI voice. They could just swarm a smaller company / indie with lawsuits if they use the voice or a voice they perceive as familiar. It wouldn't even matter if the large company was wrong - they have more than enough money to bury a smaller company in lawsuits since the smaller company doesn't have enough to fight back.

> Couldn't companies just standardize around a price and remain at that?

This sounds like a cartel, which would be neither legal nor easy to do. All it takes is one rogue member who tries to gain more by breaking the cartel. (Similar to how FB/Meta didn't take part in the hiring cartel and kept paying high wages).

The thing is, the current generation of Netflix style content flood is not too far from a world of totally auto generated content flood.

Like how ISPs pick and choose what parts of an area they'll support and then you're stuck with only one ISP that you can choose from. This has been true in every place I've lived in.

Another good example is Luxottica. They own the majority of sunglasses brands as well as Sunglasses Hut. When Oakley wanted to lower their prices dramatically, Luxoticca told them they have to keep their prices around their price points or they'll no longer sell the brand [0].

Companies do things all the time that are legally grey or illegal. Something being illegal won't stop a company from doing it.

Again, I understand where you're coming from. For me when I saw "This sounds nice in theory but not going to happen in practice." it made me think of all these companies that are already doing sketchy things.

I could be wrong and just not understand price points in companies. It's just difficult for me to envision a future where companies aren't taking advantage of AI to pocket huge profits in some way.

[0]: "Oakley had tried to dispute their prices because of Luxottica's large marketshare, and Luxottica responded by dropping Oakley from their stores, causing their stock price to drop, followed by Luxottica's hostile take over of the company." from https://en.wikipedia.org/wiki/Luxottica

But being an ISP is capital intensive. You need to spend money to support a bigger area. The content providers want to win all.

> Another good example is Luxottica. They own the majority of sunglasses brands as well as Sunglasses Hut. When Oakley wanted to lower their prices dramatically, Luxoticca told them they have to keep their prices around their price points or they'll no longer sell the brand [0].

This is different though. They can do it because they control the store. Nobody controls the content store.

> It's just difficult for me to envision a future where companies aren't taking advantage of AI to pocket huge profits in some way.

Companies will be trying to pocket huge profits for sure, but for themselves not for the collective. If you can improve your profits 10% by extinguishing all competitors, that is a nice bonus for the execs and nice boost for the stock for the next 5 years.

> Companies will be trying to pocket huge profits for sure, but for themselves not for the collective.

Definitely makes sense. Like a companies could work together to maintain a price point, but if a company can make so much more by undercutting - they'll do that.

Biden: Trump, I thought you got arrested yesterday?

Trump: How are you awake? it's like 1PM

edit: here's a link to one, the language is pretty crass so make sure you have headphones https://www.youtube.com/watch?v=-pGT5jLlxb4

Just how many more edgy reboots of Batman am I going to have to be subjected to before I die, feel like I'm living in Groundhog Day which will also surely have a reboot in the next 4 years or so.

Can you imagine how cool the future is going to be where 1 dev shops can have the voice acting chops previously reserved for big budget AAA games?

If AI art in whatever form it takes is just as good as the human made then we're golden. If it's worse but cheap enough that some choose it then they never actually wanted quality, cheaper was just never on offer before.

It is when they literally don't hire enough actors. IIRC Fallout 3 was the worst offender for this, the first 2 hours of the game has like 6 celebrity VAs, the next 80 hours has like 3 male VAs and 2 female VAs and one of the male VAs has an insanely distinctive voice that gets used every other NPC and even on a child at one point.

It's laughably bad no exec didn't push to fix that before shipping. But yeah probably blew all the VA budget on the celebs that can't fill the amount of voice required for the entire game, same issue with Destiny by Bungie where hiring too expensive talent even drove who got killed in the plot and who the stories were about.

I think this is a possible final form of paradigm shift for generative AI.

I do. Creative works are an act of communication. I am not interested in creative works that aren't a person communicating with me.

However Billions of people are creative and had just lacked the time, energy finances, to investment in the act of being creative.

People have huge discussions on what they'd change or fix in tv shows, what worked and what didn't.

Now we can empower them as well.

If it's not possible for me to distinguish between real people and fake people, then yes, quite a lot changes for me.

It could be 100% AI generated, it wouldn't change a thing (For me), A creative work doesn't NEED to have a soul to fulfill a purpose, it does if what you care for is the message or the media

I like how you think AI is just going to stop at utopia.

Well no obviously not because we're both have-nots. Our ability to survive is is based on being useful and selling ourselves to the haves. Anything that reduces the need for human labor makes their lives better and our lives worse for some reason we seem to put up with. The treadmill were on just got faster.

How anyone sees this as a problem with AI and not capitalism when AI is just the latest in a long line of progress that's turned the knob I can't understand.

In a capitalist system, the cost is an essential factor of price, and firms are primarily motivated by profit maximization, which requires minimizing costs. Therefore, capitalists tend to prioritize cost-cutting measures such as outsourcing, automation, and downsizing to increase their profits.

Other economic systems would have different priorities -- and could choose to put greater value on human happiness.

That's not awful. Cost is a bad thing

A key problem with capitalism is that externalities are not accounted for. Overall costs can increase, but because of the lack of accounting and an imperfect market.

In any system you want to cut costs. You want to generate the same output for less input (time+resources), as that gives you more time+resources to do other things.

Is it? Stated as an absolute like this, I disagree. Take the cost of paying employees, for instance. I think it's hard to argue that's a bad thing. Employees are also customers, providing a revenue source that leads to profit.

The employees are doing work, they are spending their time. That should be reduced.

At the end of the chain is either someone spending time, spending health, or using natural resources. Those are the costs.

Far better for a company to simply give the money to the employee, receive the same amount of work in return, but not have the employee spending any time

What's better, paying someone $100 to dig a hole and fill it back in, or give that person $100?

Imagine a world where every need is met without having to do anything, with no costs - to humans, to the environment, to resources.

Think how much time that gives to people to be free to do what they want

It's not a technological problem. It's a greed problem. The people who control the resources are not going to just give them away for nothing.

At first blush it looks that way. However, for every gem born there will be a thousand pieces of schlock made possible through AI scripts, voices and art that we’ll have to sift through. Once all that becomes the norm nothing about those games will stand out more than the others making it just as hard to woo customers.

For instance, if I told you a new movie came out and it has sound and color and choreographed fight scenes and special effects and maybe plot, would you want to see it? Not without serious convincing I imagine as all that stuff is par for the course. Sure there will be some initial novelty capitalized by first movers but it’ll get old quickly.

So your vision comes to life and you’re more productive but not necessarily more profitable or reaching a larger audience than before unless you’re one of the lucky ones. You’ll still be competing with AAA which can focus their attention elsewhere and/or produce more games since they get the same benefits with greater capital and marketing. Certainly artists and voice actors are not so golden.

This will allow programable speech. Like LaTeX for video

If you want to change your 30 min YouTube video/online lectures voice over, it's a 10 second change and recompile.

More importantly people can work on seamless dubbing for Netflix etc.

That will open up whole international markets. How many people are watching the The Three-Body Problem which is free on YouTube but subtitled?

If you match it to a light deep fake of lips movements it's even more seamless. You can now reprogram video.