2,745 karma · joined November 27, 2019
Randomly pick one.
> Trying to generate a model that's "free of correlative relationships" is impossible because the model would never have the infinitely pedantic input text to describe the exact output image.
Sure, and you can never make a medical procedure 100% safe. Doesn't mean that you don't try to make them safer. You can trim the obvious low hanging fruit though.
If you want the model to understand what a "nurse" actually is, then it shouldn't be associated with female.
If you want the model to understand how the word "nurse" is usually used, without regard for what a "nurse" actually is, then associating it with female is fine.
The issue with a correlative model is that it can easily be self-reinforcing.
If they don't, then it's completely unconvincing because it doesn't actually provide any new information.
So if the goal is to gripe to other people who "get it", then sure, vague criticisms are great.
If the goal is to convince people that something is an issue, then all vague criticisms do is establish what your opinion is; they fail to actually provide information that could (justifiably) convince anyone.
At best, they can convince someone to be interested enough to look into it on their own.
Also, if you want to make a "send money" feature, why would it being interoperable preclude you from doing it? Bank transfers are already interoperable, why can't a send money feature be?
> e2ee with group chats
See: https://interoperability.news/2022/03/end-to-end-encrypted-g...
Can you give an example of a "new feature" that you think would be stymied by this?
> how would it work with group chats where you can add and remove people arbitrarily?
I don't quite understand what issue you're pointing to. I'm not sure if you're unsure of how e2e encryption works or something else. Can you ask your question again with more specificity?
You can say that's not a getter/setter, but then your definition is just different than the people you're responding to.
I'm talking about "artistic skill", which is the ability to effectively put ideas or emotions into a work of art.
An analogy which might explain my thinking: A software engineer who writes a 1,000 line program in a month is producing work beneath their "skill" because they could've written a 100,000 line program that does the same thing, in that same month. Typing is a technical skill for programmers, but I wouldn't say that typing skill makes you a better software engineer.
> They allow a group of relatively untrained and loosely organized defenders who know the terrain well to deploy extremely effective resistance against an attacker, as long as it's at short range. A drone swarm can quite literally destroy all hostile forces within an area without risking a single person, but it can't do this beyond say 100 miles out.
We already have this "drone swarm", we just call it a guided missile.
The hard part in fighting a modern army isn't killing them, it's finding them. The defender is inherently at a disadvantage in this regard because they have things to defend, which necessitate that they're position in the vicinity. Russia is struggling at the moment not because defenders are inherently advantaged but because they're relying on conscripts and relatively untrained soldiers.
> The musket allowed relatively untrained militias to enjoy superior firepower over the knights and longbowmen that had trained professionally their whole lives. As a result, smaller city-states and colonies could defend themselves against the large standing armies that kings and emperors could wield, and so the feudal system collapsed.
The exact opposite of what you're describing happened with the wide utilization of gunpowder. Pre-gunpowder, city-states and small kingdoms enjoyed relative independence due to the sheer expense of penetrating walls. Post-gunpowder, artillery (not rifles) required a whole professional organization to be utilized effectively, and formed the backbone of the army, so small states could no longer field or effectively defend against larger states, leading to increased centralization of authority, well before the creation of nation-states. "Makers of Modern Strategy from Machiavelli to the Nuclear Age" covers this transition pretty extensively.
Have you tried having DALL-E draw a blue cube, red cube, and yellow sphere in the same image?
https://mymodernmet.com/wp/wp-content/uploads/2016/12/1-pica...
Than at 35:
https://mymodernmet.com/wp/wp-content/uploads/2016/12/6-pica...
And that he was an even worse artist at 90:
https://mymodernmet.com/wp/wp-content/uploads/2016/12/13-pic...
People without functional legs can't simply decide to walk up some steps.
People use Firefox can simply decide to use Chrome.
To be more direct; what's your definition of "broken"? Is it that it doesn't work for you in the manner that you'd like?
If one new oil well is discovered every year, then you can say that who own that growth is zero-sum. You can also say that the amount anyone can increase their wealth via oil wells by is zero-sum.
Sum over the growth of capital over the year, and you find that the allocation of the ownership of growth is also zero-sum. And so are the proceeds of that yearly growth.
If your claim that the growth of the world's productivity on a yearly basis is unbounded, then sure, zero-sum isn't an applicable term.
If your claim is that you can generate growth without hampering the growth of others; that's true to a point. Someone needs to sweep the floors and take out the trash. You can make robots do it, but those robots need to be made and maintained by someone. Someone needs to make food as you do it. Someone is on "robot design duty" and someone is on "farming duty". The person on "farming duty" can't stop without someone coming in to replace them.
But I haven't been able to see any evidence of that claim being true.
The game itself doesn't have intrinsic chance, but as with everything in life, the players do.
I do agree with your last point though, winning doesn't mean you played well, it just means you played better than your opponent.
I think you've inverted the expectation here. What I was saying is a machine passes if you can't find anyone in the world who can distinguish between it and an actual human. Meaning that if the "world's smartest person" can distinguish between it and an actual human, it fails, even if it can fool everyone else.
A machine can pass the test by deliberately feigning to be a human with limited communication capabilities (ex: I think we can simulate 2 month old baby talking via text). But then all you've shown is that your machine is as capable of thought as a 2 month old baby, which probably isn't the bar that you're trying to reach.
Then there's the conceptual argument of the Turing test, which we can turn into a test of intelligence. It relies on the idea that we can abstract the mind into a "thinking" black box which takes inputs and outputs. And then posits that any black box which can't be distinguished from an actual "thinking" black box may as well be "thinking".
Passing the literal Turing test is a sign that some humans can be tricked for some small domain over some small period of time. Passing the conceptual argument that the Turing test relies on shows that there are non-human entities which cannot be separated from humans on the basis of how they interact with a human (through some limited medium).
The repetition, duration and prizes are just practicality; prizes incentivize people to try, repetition ensures that the results are robust, and duration ensures that the domain the AI can be tested over is equivalent to a humans.
Also, it shouldn't matter if the human is someone who worked on the AI, or has read the code, or has seen every previous Turing test that the AI underwent. There shouldn't be any information a person could know that would allow them to tell that it's an AI.
1. This judge is aware that they have to discern whether the 'bot' is real or a machine.
2. The judge cannot discern whether the 'bot' is real or a machine better than random chance.
This failed 1. And even given that advantage, might have failed 2 as well?
Often I see headlines along the lines of "X fools people and beats the Turing test!". But the point of the Turing test isn't to trick a person, it's to make it functionally impossible for a person to distinguish between the real and simulated thing, no matter how hard they try. For something to pass a Turing test, it would need to be able to pass the following:
"Anyone can play as judge any number of times. You can take as long as you want, and if you're successful in breaking it under controlled conditions (IE, you don't cheat and use an out-of-band communication protocol with the 'bot'/human), we'll give you a 10,000,000$."
The "One Million Dollar Paranormal Challenge" (https://en.wikipedia.org/wiki/One_Million_Dollar_Paranormal_...) is a solid example of a Turing test for magic.
So it's not even that the bank can't pay out, it's just that it costs them a little extra.
To add to that point, decks can be good or bad based on the ecosystem (meta) they're playing in. A deck that aims to rush you down might be great in a meta where players expect slow decks, but terrible in a meta where many decks have means of healing themselves.
It also doesn't remove the human factors from redistricting; humans would have to choose and approve this method. It's just laundering the gerrymandering through an algorithm to give it credibility.
But if you're the party that stands to benefit from changing redistricting laws, then the system loses credibility because you picked this method.
And if you're the party that stands to lose from changing these laws, then you're unlikely to self-sabotage yourself by doing so.
It would be the equivalent of going to a restaurant and asking for something non-standard and off menu (say change a meal to vegan or gluten free), being denied that, noting that they seem only seem to be denying black patrons off menu orders, and then claiming the restaurant violated your rights.
Why do you ask?
Is there a source for this? I'd guess that there's a percentile where American health care has better outcomes than other countries, but 50th seems pretty low.
I see the niche. The risks are:
- the mid market is constantly churning; either players become too big and you can't meet their requirements or they go bankrupt. Customer acquisition becomes a pretty big expense.
- selling CPU cycles is a cutthroat business which competes pretty directly with AWS, Azure, and Google Cloud. Their edge will likely be ease of use, but at some scale, the larger providers will be able to undercut them hard.
- selling a solution for managing datasets and training models using cloud CPUs is a crowded market.
- not sure how trustworthy the company is with private datasets. Easier to trust an established vendor.
But it wouldn't be a startup if there weren't risks.
1. Investors expect Huggingface to extract more than $100M from the market. Otherwise they'd be called 'donors'.
2. If they openly publish models, then their APIs will be undercut by other providers who can take the published model and host it for cheaper. It would be cheaper for other companies because: they don't need to pay the cost of training the model, and they can specialize in simply hosting models.
3. Because of 2), Huggingface would need to avoid allowing other companies to host models, including internal APIs (because then providers would simply spin up to making hosting those internal APIs easy).
4) Because of 3), their policy of publishing trained models openly has to change.
So the question that the original poster was asking is: what Huggingface policies will change, given the need to make returns on this investment?
The original poster is likely thinking of OpenAI, which went down a similar route (starting training open models, took in a bunch of money, realized that openly publishing them wasn't sustainable, kept the models secret and created locked down APIs for accessing them).
> So if you want someone to answer precisely how they'll extract hundreds of millions of dollars from an emerging market, I have to imagine this isn't the correct forum to expect such answers.
This market isn't new; Google, AWS, OpenAI, etc. all have APIs they charge for. They also have services to host trained models for you. How will Huggingface make money without resorting to hiding its models?
That doesn't answer the original question. The question was:
"How will you extract $100M+ from the software market?"
Your answer was:
"I think that Huggingface will produce $100M worth of value."
Which may or may not be true, but just because something produces X amount of value doesn't mean the project will be able to extract that value. See any open source project.