The demo video notes that it uses Phoenix as the most-commenting, and Edgeworth as the second (and the rest of the cast following that); in addition, it uses a NLP to determine the tone of the voice, and mark an Objection! if it's negative.
I've wondered if the AI superintelligence everyone is watching out for is actually slowly being built from the bottom up, and encompasses evening in the entire world. Deep Thought from Hitchhiker's Guide is seeming more and more plausible to me. Perhaps this bot that makes Reddit videos doesn't do anything "useful" per se, but it is doing what a human might otherwise do, and adds itself to Reddit as an entity that could be indistinguishable from a person making silly videos.
What percentage of Reddit comments are bots? I'd be curious to know. SaaS and Lambda functions and and APIs are all like very complicated neurons that link together to form this world wide web of interactions.
I've wondered if an approach to the AI alignment problem is really to see that the entirety of all the computers in the world are a giant brain that is continually self-improving. Phoenix Wright bot is one neuron, like any other. And so to align the world AI, you have to align the culture that makes the AI. And so, basically, anything that one does to improve the culture in a sense improves AI.
> a giant brain that is continually self-improving
...nope, does not qualify. If it cannot manage to replicate itself completely (as in: create another separate 'internet' capable of further replication) it is as good as a single human living on Mars. It's a short ride, one small mistake and goodbye Mr. Superintelligence. (I'm not claiming humans will be able to survive that though.)
You are spot on. We are so self-centered that we think an AI superintelligence will be something like a superintelligent-human-mind/person inside a computer, when most likely, it will be nothing like a human and probably completely unrecognizable to us.
In fact, I'd argue it's already here and already taken over.
Most of us are now enslaved by screens and the Internet, we are addicted to them, even if we don't want to admit it. Some might say these things are run by humans, but are humans really in charge? Can any single person turn off the internet or prevent everyone else from using their phones/screens for hours every day? Is even a group of people capable of doing it? And even if they could, would they?
Maybe if humans went extinct, the Internet and screens would disappear, but that just means there's a symbiotic relationship, not that we are by any means in control.
Are you implying Google did it on purpose to prevent this super-intelligence from taking over?
To me it seems like the opposite, the outage was untintentional and then humans worked really hard to fix it. It's almost as if we were the cells of the body of this super-intelligence.
> Are you implying Google did it on purpose to prevent this super-intelligence from taking over?
Why are you reading that from what I said? I wasn't responding to the superintelligence thing (as I hadn't quoted it), and that seems like the weirdest reading from what I responded to.
The claim was: "Can any single person turn off the internet or prevent everyone else from using their phones/screens for hours every day?"
And I responded with proof that that is marginally possible. Probably not a single person, but a well-trained well-connected group could do it very easily.
I feel like everyone stating that "the superintelligence is around, and has already won" in this thread is acting based on what they wish to be true, rather than what we can see based on fact.
You have turned a potentiality for which there is as yet no evidence, into a certainty. A rational mindset in general tends to frown on such things.
I feel I should clarify, for the sake of anyone that bothers to browse my profile / lurkers, I don't mean Rationality inasmuch as Jordan Peterson does[0], or any of the alt-right. I mean Rationality in the Humanist sense, i.e. the progressive Rational Wiki's left-wing approach.
I understand. However, what I'm saying is that even if that capability exists, in practice it doesn't matter, as it is something that we will never act on becauae we've already been taken over.
That outage example shows that people will immediately act to restore, preserve and expand the system, never destroy it.
This is a rather spurious assumption, that you have there.
'The system' that you're referring to in most likelyhood is just an aspect of science fiction, as such there is no provable threat from the internet, and it doesn't make any kind of capitalistic sense to turn off the infrastructure upon which rests, for example, the stock market, the banking infrastructure, a small portion of the communications infrastructure.
Now, if there were a provable, tangible threat to these systems remaining, and it was seen to likely be a GH-0 'Dead Greenhouse' scenario[0], then I think it would push people to take action.
But I don't think you can claim that "people will never turn off the internet, ever", simply because as-such they have never done it. With that line of logic the USSR would never have got humanity to space[1].
> I've wondered if the AI superintelligence everyone is watching out for is actually slowly being built from the bottom up, and encompasses evening in the entire world.
every time someone tries this idea of training an AI on "the internet" it comes out badly racist / bigoted.
I've looked at various political debate subreddits over the past few years. But over time I considered how these spaces could be used nefariously. If I were a propagandist, I'd throw all of this content into some deep learning doohickey and let it generate sample responses for my army of trolls, and let them quickly edit the best samples to create more complicated arguments more quickly, until the bots are consistent enough to run amok as truly convincing persons.
Kinda like when after Kasparov lost to Deep Blue, there was a thought for a while that chess tournaments would include humans working with a computer model to help them make decisions. For some time a human working together with a computer to make decisions was superior against a human or computer alone -- this hasn't been the case for a while now as chess models advanced.
Apparently it's not running anymore, but https://www.reddit.com/r/SubredditSimulator/ is sort of apropos. It's an entirely bot-generated subreddit. You can still read through the old posts.
maybe so, but I don't think this particular program does anything like AI. glancing at the code, it looks like it just pastes the full text of a reddit comment into a video template, using a different template for each unique commenter in the thread. it no more "does what a human might otherwise do" than how a sorting algorithm might emulate how I sort my socks.
Did you know that it's actually a rather pointed parody/criticism of the Japanese justice system? I had no idea myself! Perhaps I was the only one who didn't know.
The conviction rate in Japan is over 99% -- going to "trial" is essentially the same as being found guilty.
The prosecuting attorneys have a lot of discretion over what is prosecuted, and are effectively de facto judges. They may decide to simply not prosecute a case if they feel "character, age, environment, gravity of the offense, circumstances or situation after the offense" makes it unnecessary.
So it's easy to imagine there might be some with quite an ego like Edgeworth or Von Karma -- and a protagonist like our heroic defense attorney would be quite the unusual attorney, indeed! I loved the games even without knowing any of that, but that really makes them all the more poignant and funny for me.
I wonder if this is partly an extension of what you might call "confirmation culture", which could be described as a way of avoiding disgrace/humiliation.
For instance, when applying for a promotion, I've heard that in Japanese culture it is common to get pre-approval from various parties beforehand so that it is ensured before a formal/public application is made; this way, nobody is disgraced by being turned down. Not sure how true that is. But it could apply to the justice system in a similar way; having a high conviction rate means the justice system is effective, whereas trying someone and not convicting them might make the justice system look bad.
Your article says that a study of every legal case since WWII found that it was due to prosecutors not trying uncertain cases, which would seem to be a good thing for the innocent.
There are criticisms about allegedly forced confessions and such, but that seems to be unrelated to the conviction rate itself.
Fair enough, but that criticism of the conviction rate always stuck out to me--and this is by no means the first time I've heard it repeated--simply because it's one of those cases where there is no right answer.
I recently started playing it on the Switch, and my first thought was that I hope the Japanese court system isn't really like this. It makes me appreciate the idea of "innocent until proven guilty" in the US. Ace Attorney feels much more "I don't like you, so I'm going to assume you're guilty unless you have absolute proof that you're not".
Justice systems are complicated. From what I've read, it's true that the japanese court system indeed gives prosecutors more leeway and the trial is more of a formality. The conviction rate of 99% (vs 70-95% in the us, depending how you count) speaks to this.
However, that's not all there is to it. You're still innocent until proven guilty in Japan.
There's also probably more in common with the japanese justice system and the US system than different. In practice, the majority of cases in both don't go to trial. In japan, the prosecution will just not prosecute any case they don't think they'll win (and have broad leeway to do so). This results in a 99% conviction rate because, well, there's no reason they can't just drop any case without rock solid evidence. Even in cases where they're confident the suspect is guilty, but they don't think they have enough evidence, they may opt to just drop the case.
In the US, similarly, most cases don't go to trial. The vast majority are also ended under prosecutor discretion, via plea bargains. Both in the US and Japan, most cases skip the proper court system due to the prosecution having power to do so. It's just in Japan, the prosecution also has power to drop the case, while in the US the prosecution can only skip the courts if they get a plea deal, but not if the suspect keeps pleading not-guilty.
In america it's routine for people to be convicted on nothing other than a suspect agreeing to a plea deal even though they didn't do it. "You can get 1 year in jail and get back to your family if you take the plea, or you can risk 30 years in jail if you keep saying not-guilty" isn't at all unusual in the US system, especially since a cop's word is often enough for a guilty sentence.
I'd be happy to be wrong on any of he above details if anyone has separate experience, understanding, or sources.
All that being said, I'd much much rather be arrested in Japan than in the US.
The first Ace Attorney game must have been incredible value for Capcom's investment. The spartan, comic-like sprites and straightforward visual novel gameplay likely fit into a smaller ROM size. Given the charming narrative, I bet they made their ROI pretty quickly in the first print run.
This is just like the CBT mindfulness technique for defusing from unwelcome thoughts. Imagine the painful discussion going on in your head is being carried out by the Ducktails kids or something ridiculous. To draw attention to the fact that they are just words.
It would be cool to see some bot like this to feed in unwanted thoughts and get back a cartoon.. there are so many great things we could be doing with computing to redefine our selves and our interfaces with world and others, but all we got so far is thins reinforcing the same old tired power relations of yore
You do often win the game by being pedantic and pointing out small issues that didn't have anything to do with the original main testimony, so maybe parent is on to something after all :)
The HN hivemind would not allow such discussions to take place in the first place. Which kinda sucks, if you have some other view that is not accepted by the general mindset here, have some downvotes.
Pocket does Text to Speech, IIRC. I think I tried it on reddit at some point. It turns out the ability to just skip comments and not read through all the threads in their entirety is pretty important.
The readme doesn't mention it, but I like that it seems that anim.py is pretty standalone so you could potentially use it not just for Reddit comment threads, but pretty much any conversation at all.
Also like that it seems fairly easy to modify to use another prosecutor as the second main character. I know who I'd use!
My sides are splitting. This is the lighthearted use of technology that I live for. Thanks for making my weekend, OP! And now to find good subreddits to target...
Absolutely genius.