Here, the current 'AI' craze doesn't even in theory extrapolate to higher-order displays of intelligence.
I'd love to see the theory that claim is based on. I've never heard of any theory of "higher-order" intelligence (whatever that means) that tells us the current approach won't scale. If you're going to make claims about what is or isn't possible "in theory", those claims need to be based on an actual theory. Otherwise, it's empty rhetoric.
Many AI researchers believe the current approach can and probably will ultimately lead to human like intelligence. They reason like this. Humans have human like intelligence. We don't understand the details of how it works, but we're learning more all the time. Current approaches to AI are inspired by how the brain works. We're making really fast progress at it. And as we learn more about the brain, we can apply those new insights to AI too.
This gives us a pretty clear road that, if followed to the end, has a good chance of producing human like intelligence. We can't see to the end of the road yet. We can just see to the next bend. But that's ok, we don't need to. We'll just follow it one bend at a time and it'll get us there eventually.
A lot of mathematicians would say if you haven't studied ZF set theory, you don't understand 2+2 either. And if you have studied it, you know it describes addition as a process of algorithmic symbol manipulation. Exactly the sort of thing computers are great at.
Maybe the main difference is that humans delude themselves into thinking they "understand" things when really they're just following rules. Computers don't have that problem. So which is more intelligent?
I don't think they really get what makes YouTube popular. YouTube is all about making it easy to share videos. Facebook is all about locking you in and keeping you from ever navigating away from Facebook. When they start letting people without Facebook accounts view content, and when they let you embed videos into other web pages, then they might have a chance of making a YouTube competitor.
Oil is refined into gasoline to run cars. If only someone would invent an "electric car" that ran on electricity instead, we could power them with wind energy.
The analysis indicates that a network of land-based 2.5-megawatt (MW) turbines restricted to nonforested, ice-free, nonurban areas operating at as little as 20% of their rated capacity could supply >40 times current worldwide consumption of electricity, >5 times total global use of energy in all forms.
Sadly, it wouldn't. Making punishments more severe only has a weak effect on how well they work as deterrents. People always assume if you punish a crime really harshly, no one will commit it. But it doesn't work. People go on doing it anyway. If you're thinking of committing a crime, whether the punishment would be five years in prison or ten just isn't going to affect your thinking much.
The thing that actually does make a big difference is the certainty of punishment. If you think you can get away with it, you just don't consider the potential punishment much. But if you think you'll probably get caught, that becomes a big deterrent even if the punishment is a lot lighter.
We've been told that "fiat currencies" are bad because a government can print more money whenever they want, deflating the money people already hold. You can't trust it to keep its value. The solution is cryptocurrency! The supply is strictly controlled by an algorithm, so you can trust it to hold its value.
And what happens in practice? Cryptocurrencies are incredibly volatile. You can't rely on them at all. So instead someone creates a "stablecoin" that really does hold its value. And the way they do that is... by tying it to a fiat currency.
I have no clue what point you're trying to make. You made the assertion that "concerns over global warming should stop any day now". I explained all the reasons that's not true. We are currently emitting large amounts of CO2, we are currently on a path to keep doing so for decades (worldwide emissions actually increased last year), and even if we stopped today the planet would continue getting warmer for the next century. This has nothing to do with ideology. They're just facts. We can discuss the best way of solving the problem, but it shows no sign of going away on its own "any day now".
It's the total emissions that matter, not the per joule emissions. If we cut the emissions per unit energy in half but double the amount of energy we produce, the effect on the climate is the same.
It wouldn't be too much of a stretch for an electric utility to see this growing demand and decide that instead of buying new land and running new wires that they could instead put these funds from new electricity demand into upgrading existing power plants.
Sure, that could happen if they have a financial incentive to do it. A carbon tax for example. But it's not going to magically happen on its own. Most power plants are in places where land is cheap, and anyway solar takes more land than coal, not less. (Actually that's only true if you don't count the land used for mining the coal, but the power plant owner doesn't own that land, so it's not a factor for them.) Nuclear is super expensive to build. Hardly anyone is building nuclear anymore, because it just can't compete. What incentive do they have to throw our their existing investment, shut down a power plant with decades of useful life left on it, and build a cleaner replacement?
Eventually it will all get replaced, of course. But we're already well into the climate danger zone, and if we keep emitting CO2 at anything close to our current rate for another forty years we'll be in really bad shape.
Well, not solved exactly but the concerns over global warming should stop any day now
Only if you take a very long term view of "any day now". Most energy infrastructure (power plants and the like) is built to last around 40 years. Even if we completely stopped building non-renewable energy sources today (and we're not there yet, though the majority of new capacity is now renewable), existing power plants would keep producing greenhouse gases for decades. To prevent that, we need to accelerate the process and shut them down ahead of schedule. That's a lot easier (less expensive) to do if overall energy use is decreasing than if it's increasing.
And it's not just power plants. Transportation produces a big part of the world's CO2 emissions. Right now, only a tiny fraction of new cars being sold are electric. It will probably be about 10 years before the majority of new cars are electric, then more years after that before the majority of cars on the road are electric. At that point we'll need to add a ton of new generation capacity to power all those electric cars, which will make it even harder to retire old power plants.
Airplanes are a harder problem. Right now, electric planes just aren't practical. Airplane manufacturers are predicting within 5-10 years they'll become practical for short flights, but making them useful for long flights will take really huge advances in battery capacity.
And then there's the fact that the earth is just really big, and the climate has a lot of inertia. It takes a long time (decades to centuries) to respond to changes. If we totally stopped producing greenhouse gases today, the earth would keep warming for decades. Glaciers would keep melting and sea level would keep rising. That's because we've only seen part of the warming from the CO2 we've already emitted. To really "solve" global warming and stop that from happening, we would need to start taking CO2 back out of the atmosphere and get back to the level it was at several decades ago. But we're a long long way from doing that.
The summary is totally confused. It defines "outside" as "buildings other than their homes". So specifically not including being outdoors. Probably it just means more people shop online and telecommute, so they spend less time in stores and offices. Hopefully that leaves them more time to spend enjoying the outdoors.
They suffer when people can no longer afford to live in the area and move away. Property prices fall, the neighbourhood changes.
That's backward. If people are moving away because they can't afford to live there, that means property prices are *rising*, not falling. That's classic gentrification. When a neighborhood gets expensive, the poor people move away. That does change the neighborhood, but a lot of the rich people moving in probably think it's changing for the better. And income disparity within that neighborhood decreases, not increases, because the poor people are moving away.
What you really end up with is segregation. Rich neighborhoods and poor neighborhoods. And maybe crime goes up overall, but the rich people can deal with that by building gated communities and hiring private security services to police their neighborhoods. It's not a big deal for them, because the crime is happening somewhere else, not where they live.
Actually no. They mean the really ugly building with fluorescent tubes in the ceiling and lots of desks. Or maybe lots of shelves stocked with things to buy. They define it as "buildings other than their homes". The author seems a little confused about what "outside" normally means.
When they say the money was kept "overseas" and now they're "bringing it to the US", that doesn't mean what you think it does. In fact that "overseas" money was being kept in bank accounts in New York and managed by an Apple subsidiary in Reno. It's only "overseas" in a completely fictional sense invented by accountants. No new money is going to be entering the US. There won't be any investment boom. The money's already here and they've already been investing it. Now congress has given them a huge tax cut, so to say thank you they're pretending it will magically create jobs. They have to pretend it will somehow help regular people, not just their shareholders.
The questions in this test weren't like that. The reading passages were Wikipedia articles, and the questions asked about objective statements that were clearly given in the passage. Here's an example.
The test you're talking about looks at something totally different. It presents ambiguous sentences with no context. The reader is supposed to use their existing knowledge to resolve the ambiguity and infer what the sentence is talking about. These are both interesting and important problems. But they're totally different problems. This work was aimed at solving the first problem, not the second one.
I'm a firm believer in "one person one vote". People are always giving arguments for why some votes should count more than others, but I just don't buy them. When the country was founded, a lot of people argued that only rich people should be allowed to vote. They payed the taxes that supported the government, after all. But that would just have led to laws that (even more) favored the rich over the poor. A hundred years ago, women weren't allowed to vote. A lot of people firmly believed that was right too. They offered arguments that to themselves seemed really solid. Women were less well educated. They were too busy raising their children to pay proper attention to national affairs. They were inherently less capable of making important decisions. These were all just excuses. They kept the power in the hands of men who (surprise!) made sure the laws favored men over women.
For a present day example, people in small states get more power than people in large states (directly in presidential elections, indirectly by having more senators per capita). And they offer all sorts of reasons for why this is justified. To me, their arguments all sound like nonsense. But they somehow find them convincing, since of course they want to be convinced.
You're saying that uninformed people should have less say than informed people. But uninformed about what? I'll bet those "uninformed" people know a lot more than you do about the problems in their own lives and their own neighborhoods. And anyway, what makes being informed unique? Why is that the one and only thing we should weight people on? Should we also test their logical reasoning ability? That's often more important than raw information. Or should upstanding, moral people get more say than immoral ones? A lot of people probably feel that way.
In the end, the only position I find really defensible is one person, one vote. Anything else ends up as a way of taking power away from people you don't like.
That would be so easy to abuse depending on what questions you asked. People who think violent crime is a huge problem would probably get it right. People who think climate change is a much bigger deal and don't pay attention to crime statistics would likely get it wrong. It becomes really easy to favor certain groups by asking questions about things they're likely to know.
Can someone explain why I'm supposed to care about wireless charging? I spend less than 10 seconds a day plugging my phone in to charge and unplugging it again. How does this make my life better?
No, they really are trying to squeeze people for money, not protect themselves. They've been moving in this direction for a few years, and this is just the latest step. They already had cracked down on pretty much all the companies that install pre-built clusters. None of them have been able to sell clusters with GeForce cards for a couple of years, because NVIDIA wouldn't let them. If you wanted a cluster with GeForce, you had to buy individual servers, set up the cluster yourself, buy the GPUs at retail, and install them yourself. Now they're trying to close even that loophole.
In addition to this, even if you do need to scale Tensorflow beyond your local hardware, you still aren't bound to Google Cloud. Because after all, it's open source. That means any cloud provider is free to support it.
Contrast that with Amazon Lumberyard, which is a closed source on ramp. It's free (as in beer) to use, but if you want to use any cloud features, you're required to get them from Amazon. You're legally barred from using it with any other cloud provider. That's the sort of on ramp we should be worrying about, because unlike open source, it doesn't respect your freedom.
That means either they cannot be created with technology, or we are very very far from being able to do so.
Or maybe it just means they're bogus concepts.
People have argued for centuries about what "consciousness" and "intelligence" mean, and they still can't agree. So engineers roll their eyes, turn their backs on the argument, and get on with the job of creating useful things. And then people say, "It's not really intelligent! It's not really conscious!" Well, who's to say? If you can't define what the words mean, it's impossible to decide whether AIs meet the definitions. So give us some rigorous definitions of what you mean by the words "intelligent" and "conscious". Then (and only then) we can talk about how to build a machine that has those properties.
"A word whose meaning isn't defined" is not the same thing as "a concept we don't understand".
But, if one took the resources spent on gathering and curating corpora and instead invested in rule-based systems, you could get much further in less time.
Really? Why do you think that? Rule based is how all machine translation systems worked until just a few years ago. They worked, but not that great. And that's after decades of optimizing. Then the NMT systems came out and blew them out of the water.
And building a monolingual corpus is pretty easy. Have a shelf of books written in that language? Great, scan them in. Maybe there's a newspaper with an archive of back issues. There you go, you're set. Way easier than a bilingual corpus, where someone has to translate everything by hand, and match up sentences between them.
Have you tried it recently? Their old phrase based translations were terrible for Asian languages. Ask it to translate Japanese into English and you'd get garbage. Then they rolled out their new system based on neural networks, and it suddenly got a lot better. Not perfect, but now you can tell what it's saying. It's always easier translating between closely related languages, but the NNs are surprisingly good even for distant ones.
There's no standard definition of "free speech". Morally speaking, it means whatever you want it to. Everyone gets to claim they're exercising free speech, and everyone else gets to disagree with them, and no one is right. But legally speaking, laws and courts in different countries have defined it different ways, and those definitions almost never come down to, "the right to say anything you want, in any context, without restriction." It's not that yelling fire in a crowded theater is a form of free speech we restrict. It simply is not an exercise of free speech. Freedom of speech does not include the right to harm people by saying things you know are false. When the government punishes you for doing it, they aren't limiting your freedom of speech in any way. They're punishing you for doing something that (according to the law) is not free speech.
Here's how Webster defines freedom of speech: "the legal right to express one's opinions freely". dictionary.com calls it: "the right of people to express their opinions publicly without governmental interference, subject to the laws against libel, incitement to violence or rebellion, etc." According to Wikipedia it means: "the right to articulate one's opinions and ideas without fear of government retaliation or censorship, or societal sanction." Notice what all of those have in common. Freedom of speech is the right to express your opinions (and ideas and beliefs). If you honestly believe something, you have the right to say it (though we sometimes might restrict that right if it conflicts with other rights). But if you don't believe something and you intentionally lie to hurt someone or con them out of their money, that's not free speech.
Slashdot Mirror
Here, the current 'AI' craze doesn't even in theory extrapolate to higher-order displays of intelligence.
I'd love to see the theory that claim is based on. I've never heard of any theory of "higher-order" intelligence (whatever that means) that tells us the current approach won't scale. If you're going to make claims about what is or isn't possible "in theory", those claims need to be based on an actual theory. Otherwise, it's empty rhetoric.
Many AI researchers believe the current approach can and probably will ultimately lead to human like intelligence. They reason like this. Humans have human like intelligence. We don't understand the details of how it works, but we're learning more all the time. Current approaches to AI are inspired by how the brain works. We're making really fast progress at it. And as we learn more about the brain, we can apply those new insights to AI too.
This gives us a pretty clear road that, if followed to the end, has a good chance of producing human like intelligence. We can't see to the end of the road yet. We can just see to the next bend. But that's ok, we don't need to. We'll just follow it one bend at a time and it'll get us there eventually.
A lot of mathematicians would say if you haven't studied ZF set theory, you don't understand 2+2 either. And if you have studied it, you know it describes addition as a process of algorithmic symbol manipulation. Exactly the sort of thing computers are great at.
Maybe the main difference is that humans delude themselves into thinking they "understand" things when really they're just following rules. Computers don't have that problem. So which is more intelligent?
I don't think they really get what makes YouTube popular. YouTube is all about making it easy to share videos. Facebook is all about locking you in and keeping you from ever navigating away from Facebook. When they start letting people without Facebook accounts view content, and when they let you embed videos into other web pages, then they might have a chance of making a YouTube competitor.
Oil is refined into gasoline to run cars. If only someone would invent an "electric car" that ran on electricity instead, we could power them with wind energy.
Completely and totally false. http://www.pnas.org/content/10.... From the abstract:
The analysis indicates that a network of land-based 2.5-megawatt (MW) turbines restricted to nonforested, ice-free, nonurban areas operating at as little as 20% of their rated capacity could supply >40 times current worldwide consumption of electricity, >5 times total global use of energy in all forms.
That should put an end to it.
Sadly, it wouldn't. Making punishments more severe only has a weak effect on how well they work as deterrents. People always assume if you punish a crime really harshly, no one will commit it. But it doesn't work. People go on doing it anyway. If you're thinking of committing a crime, whether the punishment would be five years in prison or ten just isn't going to affect your thinking much.
The thing that actually does make a big difference is the certainty of punishment. If you think you can get away with it, you just don't consider the potential punishment much. But if you think you'll probably get caught, that becomes a big deterrent even if the punishment is a lot lighter.
We've been told that "fiat currencies" are bad because a government can print more money whenever they want, deflating the money people already hold. You can't trust it to keep its value. The solution is cryptocurrency! The supply is strictly controlled by an algorithm, so you can trust it to hold its value.
And what happens in practice? Cryptocurrencies are incredibly volatile. You can't rely on them at all. So instead someone creates a "stablecoin" that really does hold its value. And the way they do that is... by tying it to a fiat currency.
The irony is just incredible.
I have no clue what point you're trying to make. You made the assertion that "concerns over global warming should stop any day now". I explained all the reasons that's not true. We are currently emitting large amounts of CO2, we are currently on a path to keep doing so for decades (worldwide emissions actually increased last year), and even if we stopped today the planet would continue getting warmer for the next century. This has nothing to do with ideology. They're just facts. We can discuss the best way of solving the problem, but it shows no sign of going away on its own "any day now".
The per joule carbon output would be lowered.
It's the total emissions that matter, not the per joule emissions. If we cut the emissions per unit energy in half but double the amount of energy we produce, the effect on the climate is the same.
It wouldn't be too much of a stretch for an electric utility to see this growing demand and decide that instead of buying new land and running new wires that they could instead put these funds from new electricity demand into upgrading existing power plants.
Sure, that could happen if they have a financial incentive to do it. A carbon tax for example. But it's not going to magically happen on its own. Most power plants are in places where land is cheap, and anyway solar takes more land than coal, not less. (Actually that's only true if you don't count the land used for mining the coal, but the power plant owner doesn't own that land, so it's not a factor for them.) Nuclear is super expensive to build. Hardly anyone is building nuclear anymore, because it just can't compete. What incentive do they have to throw our their existing investment, shut down a power plant with decades of useful life left on it, and build a cleaner replacement?
Eventually it will all get replaced, of course. But we're already well into the climate danger zone, and if we keep emitting CO2 at anything close to our current rate for another forty years we'll be in really bad shape.
Well, not solved exactly but the concerns over global warming should stop any day now
Only if you take a very long term view of "any day now". Most energy infrastructure (power plants and the like) is built to last around 40 years. Even if we completely stopped building non-renewable energy sources today (and we're not there yet, though the majority of new capacity is now renewable), existing power plants would keep producing greenhouse gases for decades. To prevent that, we need to accelerate the process and shut them down ahead of schedule. That's a lot easier (less expensive) to do if overall energy use is decreasing than if it's increasing.
And it's not just power plants. Transportation produces a big part of the world's CO2 emissions. Right now, only a tiny fraction of new cars being sold are electric. It will probably be about 10 years before the majority of new cars are electric, then more years after that before the majority of cars on the road are electric. At that point we'll need to add a ton of new generation capacity to power all those electric cars, which will make it even harder to retire old power plants.
Airplanes are a harder problem. Right now, electric planes just aren't practical. Airplane manufacturers are predicting within 5-10 years they'll become practical for short flights, but making them useful for long flights will take really huge advances in battery capacity.
And then there's the fact that the earth is just really big, and the climate has a lot of inertia. It takes a long time (decades to centuries) to respond to changes. If we totally stopped producing greenhouse gases today, the earth would keep warming for decades. Glaciers would keep melting and sea level would keep rising. That's because we've only seen part of the warming from the CO2 we've already emitted. To really "solve" global warming and stop that from happening, we would need to start taking CO2 back out of the atmosphere and get back to the level it was at several decades ago. But we're a long long way from doing that.
The summary is totally confused. It defines "outside" as "buildings other than their homes". So specifically not including being outdoors. Probably it just means more people shop online and telecommute, so they spend less time in stores and offices. Hopefully that leaves them more time to spend enjoying the outdoors.
They suffer when people can no longer afford to live in the area and move away. Property prices fall, the neighbourhood changes.
That's backward. If people are moving away because they can't afford to live there, that means property prices are *rising*, not falling. That's classic gentrification. When a neighborhood gets expensive, the poor people move away. That does change the neighborhood, but a lot of the rich people moving in probably think it's changing for the better. And income disparity within that neighborhood decreases, not increases, because the poor people are moving away.
What you really end up with is segregation. Rich neighborhoods and poor neighborhoods. And maybe crime goes up overall, but the rich people can deal with that by building gated communities and hiring private security services to police their neighborhoods. It's not a big deal for them, because the crime is happening somewhere else, not where they live.
Actually no. They mean the really ugly building with fluorescent tubes in the ceiling and lots of desks. Or maybe lots of shelves stocked with things to buy. They define it as "buildings other than their homes". The author seems a little confused about what "outside" normally means.
When they say the money was kept "overseas" and now they're "bringing it to the US", that doesn't mean what you think it does. In fact that "overseas" money was being kept in bank accounts in New York and managed by an Apple subsidiary in Reno. It's only "overseas" in a completely fictional sense invented by accountants. No new money is going to be entering the US. There won't be any investment boom. The money's already here and they've already been investing it. Now congress has given them a huge tax cut, so to say thank you they're pretending it will magically create jobs. They have to pretend it will somehow help regular people, not just their shareholders.
The questions in this test weren't like that. The reading passages were Wikipedia articles, and the questions asked about objective statements that were clearly given in the passage. Here's an example.
The test you're talking about looks at something totally different. It presents ambiguous sentences with no context. The reader is supposed to use their existing knowledge to resolve the ambiguity and infer what the sentence is talking about. These are both interesting and important problems. But they're totally different problems. This work was aimed at solving the first problem, not the second one.
I'm a firm believer in "one person one vote". People are always giving arguments for why some votes should count more than others, but I just don't buy them. When the country was founded, a lot of people argued that only rich people should be allowed to vote. They payed the taxes that supported the government, after all. But that would just have led to laws that (even more) favored the rich over the poor. A hundred years ago, women weren't allowed to vote. A lot of people firmly believed that was right too. They offered arguments that to themselves seemed really solid. Women were less well educated. They were too busy raising their children to pay proper attention to national affairs. They were inherently less capable of making important decisions. These were all just excuses. They kept the power in the hands of men who (surprise!) made sure the laws favored men over women.
For a present day example, people in small states get more power than people in large states (directly in presidential elections, indirectly by having more senators per capita). And they offer all sorts of reasons for why this is justified. To me, their arguments all sound like nonsense. But they somehow find them convincing, since of course they want to be convinced.
You're saying that uninformed people should have less say than informed people. But uninformed about what? I'll bet those "uninformed" people know a lot more than you do about the problems in their own lives and their own neighborhoods. And anyway, what makes being informed unique? Why is that the one and only thing we should weight people on? Should we also test their logical reasoning ability? That's often more important than raw information. Or should upstanding, moral people get more say than immoral ones? A lot of people probably feel that way.
In the end, the only position I find really defensible is one person, one vote. Anything else ends up as a way of taking power away from people you don't like.
That would be so easy to abuse depending on what questions you asked. People who think violent crime is a huge problem would probably get it right. People who think climate change is a much bigger deal and don't pay attention to crime statistics would likely get it wrong. It becomes really easy to favor certain groups by asking questions about things they're likely to know.
There's already software that does this. It uses census blocks, so it tends to keep neighborhoods intact.
Can someone explain why I'm supposed to care about wireless charging? I spend less than 10 seconds a day plugging my phone in to charge and unplugging it again. How does this make my life better?
No, they really are trying to squeeze people for money, not protect themselves. They've been moving in this direction for a few years, and this is just the latest step. They already had cracked down on pretty much all the companies that install pre-built clusters. None of them have been able to sell clusters with GeForce cards for a couple of years, because NVIDIA wouldn't let them. If you wanted a cluster with GeForce, you had to buy individual servers, set up the cluster yourself, buy the GPUs at retail, and install them yourself. Now they're trying to close even that loophole.
In addition to this, even if you do need to scale Tensorflow beyond your local hardware, you still aren't bound to Google Cloud. Because after all, it's open source. That means any cloud provider is free to support it.
Contrast that with Amazon Lumberyard, which is a closed source on ramp. It's free (as in beer) to use, but if you want to use any cloud features, you're required to get them from Amazon. You're legally barred from using it with any other cloud provider. That's the sort of on ramp we should be worrying about, because unlike open source, it doesn't respect your freedom.
That means either they cannot be created with technology, or we are very very far from being able to do so.
Or maybe it just means they're bogus concepts.
People have argued for centuries about what "consciousness" and "intelligence" mean, and they still can't agree. So engineers roll their eyes, turn their backs on the argument, and get on with the job of creating useful things. And then people say, "It's not really intelligent! It's not really conscious!" Well, who's to say? If you can't define what the words mean, it's impossible to decide whether AIs meet the definitions. So give us some rigorous definitions of what you mean by the words "intelligent" and "conscious". Then (and only then) we can talk about how to build a machine that has those properties.
"A word whose meaning isn't defined" is not the same thing as "a concept we don't understand".
But, if one took the resources spent on gathering and curating corpora and instead invested in rule-based systems, you could get much further in less time.
Really? Why do you think that? Rule based is how all machine translation systems worked until just a few years ago. They worked, but not that great. And that's after decades of optimizing. Then the NMT systems came out and blew them out of the water.
And building a monolingual corpus is pretty easy. Have a shelf of books written in that language? Great, scan them in. Maybe there's a newspaper with an archive of back issues. There you go, you're set. Way easier than a bilingual corpus, where someone has to translate everything by hand, and match up sentences between them.
Have you tried it recently? Their old phrase based translations were terrible for Asian languages. Ask it to translate Japanese into English and you'd get garbage. Then they rolled out their new system based on neural networks, and it suddenly got a lot better. Not perfect, but now you can tell what it's saying. It's always easier translating between closely related languages, but the NNs are surprisingly good even for distant ones.
There's no standard definition of "free speech". Morally speaking, it means whatever you want it to. Everyone gets to claim they're exercising free speech, and everyone else gets to disagree with them, and no one is right. But legally speaking, laws and courts in different countries have defined it different ways, and those definitions almost never come down to, "the right to say anything you want, in any context, without restriction." It's not that yelling fire in a crowded theater is a form of free speech we restrict. It simply is not an exercise of free speech. Freedom of speech does not include the right to harm people by saying things you know are false. When the government punishes you for doing it, they aren't limiting your freedom of speech in any way. They're punishing you for doing something that (according to the law) is not free speech.
Here's how Webster defines freedom of speech: "the legal right to express one's opinions freely". dictionary.com calls it: "the right of people to express their opinions publicly without governmental interference, subject to the laws against libel, incitement to violence or rebellion, etc." According to Wikipedia it means: "the right to articulate one's opinions and ideas without fear of government retaliation or censorship, or societal sanction." Notice what all of those have in common. Freedom of speech is the right to express your opinions (and ideas and beliefs). If you honestly believe something, you have the right to say it (though we sometimes might restrict that right if it conflicts with other rights). But if you don't believe something and you intentionally lie to hurt someone or con them out of their money, that's not free speech.