Indeed. Animals without linguistic ability (like fruit flies) need sleep, but after ChatGPT's release in 2022, now tech bros think LLMs specifically might model the animal brain in general because of anthropocentrism and anthropomorphism.
It's also a fundamental misunderstanding of how LLMs work, mixing up inference with training.
Come on, don't be uncharitable, language isn't inherently necessary for models like LLMs, you can train something similar on visual inputs. Fruit flies have neurons that pass around ~probabilities/signal strengths to each other to represent their environments and basic concepts, it's not way off as an analogy.
The statement was "AI frenzy almost convinced me that sleep was the training of our neural network with all the prompts of the day."
Prompts are specific to LLMs. Most neural networks don't have prompts.
Additionally, prompts happen during LLM inference, not LLM training. There are many non-technical people who claim they have experience "training" LLMs, when they are just an end user who added a lot of tokens to the context window during inference.
> There are many non-technical people who claim they have experience "training" LLMs, when they are just an end user who added a lot of tokens to the context window during inference.
Since in-context learning is a thing, “adding tokens to the context window”, at least with the intent and effect of having a particular impact on capabilities when inference is run on the context to which they were added, is, arguably, a kind of training.
Sure. Foundation models aren't fine-tuned, and companies fine-tune foundation models to optimize user experience. So they are modeling the animal brain on an even more specific type of LLM that happens to be related to being a consumer of AI products.
You're being pretty pedantic about the specific term used. Everything they said makes sense if you change "prompts" to "training examples" and you wouldn't expect someone who hasn't implemented an AI model to know the difference.
It's like someone said while driving the car "let's give it some gas" and you said "but the tank is almost full" when they obviously meant "let's press the accelerator pedal"
Funnily I am interested in this semantic argument.
Do LLM trainers actually feed their « beast » with prompts from the past? Especially ones that are human corrections upon false assumptions hallucinated by the LLM?
As a non-specialist I would definitely see a lot of value in doing so, but I let you, experts, clarify that point.
Immaterial how independent they are because it's completely impossible to get honest opinions of repressive regimes. The people within the regime have no real way to know whether a poll response will make it back to the government or not, so they must assume that it will. When the repercussions for having the wrong opinion are that you disappear or find yourself "volunteered" for the front line, it's best to either lie or say you think the leader is a top bloke.
You can watch Youtube videos of citizens refusing to answer contentious questions quite easily. I believe William Spaniel has produced videos (relating to the Russian General Election) where he points this out, too.
When asked in a way where the opinion can't be identified, the support numbers do drop significantly, but the approval is still estimated to be about 50-70%. In western countries governments with clear minority support start to be almost the norm.
UK government approval has surpassed 50% in a handful of polls in over 10 years, and approval peaks are typically immediately after elections before the government starts to implement its policies. The approval is currently 14%.
> When asked in a way where the opinion can't be identified, the support numbers do drop significantly
Bear in mind that the person answering the poll still lives in an oppressive regime where wrongthink can get you killed. You spend your whole life training yourself to never say anything bad about the government in public. Would you be able to turn it off?
Also, there’s literally no free press in these countries. The government will get primarily positive coverage whatever they do! The current Labour government could only dream of such coverage!
How can you tell it is a repressive regime? They have elections, a press and they are pretty satisfied about their form of governance, actually much more than their western counterparts.
So let me sum this up. We cannot ask the people. We cannot base ourselves on how their institutions function and how well they perform.
This discussion highlights how westerners suffer from some serious superiority complex where only THEY can experience genuine freedom and democracy(probably due to their superior phenotype or some inane bs), and everything outside of their little group of friends is a masquerade. The issue with that is that westerners disconnect themselves from reality. They are losing ground and it shows.
In Switzerland we don't elect our Federal Council, which is our executive branch. A bit like in the UK too. Would you say its what matters in a democracy?
Germans and Americans refuse to answer contentious questions about the genocide of Palestinian... They also probably live in a repressive regime, right?
(Also I agree with you, Russia is a capitalist dictatorship)
"Elections in the People's Republic of China occur under a one-party authoritarian political system controlled by the Chinese Communist Party (CCP). Direct elections, except in the special administrative regions of Hong Kong and Macau, occur only at the local level people's congresses and village committees, with all candidate nominations preapproved by the CCP. By law, all elections at all levels must adhere to the leadership of the CCP."
> This discussion highlights how westerners suffer from some serious superiority complex where only THEY can experience genuine freedom and democracy(probably due to their superior phenotype or some inane bs)
There is democracy in Japan, South Korea, and Taiwan.
Just say, "I'm a tankie and I support Russia's invasion of Ukraine."
> Elections in the People's Republic of China occur under a one-party authoritarian political system controlled by the Chinese Communist Party (CCP). Direct elections, except in the special administrative regions of Hong Kong and Macau, occur only at the local level people's congresses and village committees, with all candidate nominations preapproved by the CCP. By law, all elections at all levels must adhere to the leadership of the CCP.
I personally see nothing wrong with this. The word "authoritarian" is virtually meaningless. And those local elections are paramount; Locally elected representatives end up electing MPs on the provincial level, then they chose MPs of the National People's Congress. The rest is common sense: just because we are used to "elect" pedophiles, racists and parasites doesn't mean all other countries should do the same.
There were/are people who didn't personally see anything wrong with Stalin et al.
> doesn't mean all other countries should do the same
That's the great part right? If you were in place like that you wouldn't know that you "elected representatives" are any of those things since there would be no free press and exposing them would probably be illegal anyway. Ignorance is bliss, I guess that's one way to experience the world..
> This discussion highlights how westerners suffer from some serious superiority complex where only THEY can experience genuine freedom and democracy(probably due to their superior phenotype or some inane bs)
You are quite literally commenting on a topic where Brits are complaining about our democracy. You will find reams of articles about the problems with western democracies.
However, you're also commenting about countries that quite literally changed our governments in the last year. USA voted in Trump, the UK voted in Labour. Germany just voted in a new party.
China and Russia, the main comparison points, have not changed government since the 90s. This is nothing to do with phenotypes, it's 100% just looking at the facts.
Russia is very similar to the rest of western democracies, so I won't comment further on that.
Regarding China, their leading party hasn't switched in 80 years, but their policies have and have plenty actually. Changing parties matters only a little bit in the grand scheme of things. I'd argue, for example, that Japan, that has been ruled by a single party for all of his modern existence, is still considered by many in the west as a functioning democracy.
In any case, I think all the replies have made my point for me that your dismissal of our rhetoric as based on "western arrogance" are simply nonsense. It's in fact you who's displayed a lack of understanding of those you argue against.
It's not defamation if it's true. Why do you think women warning other women about rapey and stalker men are mostly lies? Even if it's only 5% of men, wouldn't the discussion focus on that dangerous 5% over persecuting the innocent 95%, as a matter of self-preservation?
An irony in this conversation is how normalized it is for women to be concerned about men as a demographic when it's only a small minority that inflict harm. While it's controversial for men to be concerned about women as a demographic when it's only a small minority that inflict harm.
I still maintain my pet theory that this is a downstream effect of the normalization of paranoia around pedophiles that began hitting the mainstream in the '80s. The modern world is exceptionally safe, yet to the average person, it feels exceptionally dangerous.
...While I've got the hood up, I'll continue soapboxing.
I've started seeing rare instances such as a young woman walking around a corner and there is a man rounding the same corner, surprising her by mistake, and the woman starts crying or breathing in a panicked way, unable to regulate herself for several minutes. It's not always walking around the corner at the same time, but there's a common pattern of being surprised by a man just going about his day and experiencing a severe fear response to that interaction.
When I look at a lot of cultural related issues today, beyond just gender, I see many signs of pervasive psychological issues. I don't know what the solution is, but I'm very confident that the root cause is more complicated than something you can describe in a single sentence.
Maybe it's different now, I have no clue, but I'm in my 40's now and don't make a habit of hanging out with 20 year olds.
But I was friends with my wife's friends before we got married, and in a sample size of ~20 women my age, every single one of them has experienced inappropriate and unwanted touching in social settings. And a large number of them were victims of outright rape.
In comparison, I have many male friends and of them, I only know one who has been wrongly accused of sexual assault (the lady openly talked about doing it to help with a promotion...)
So even if both sides may have a few bad apples, one side is a much more prevalent problem when it comes to the number of victims.
> An irony in this conversation is how normalized it is for women to be concerned about men as a demographic when it's only a small minority that inflict harm.
The same hypothetical 5% can inflict harm to multiple women, that's why multiple women and girls complained about Epstein and Trump.
Do you think a women's dating safety app is mainly about women lying and intending to hurt men, because it's rare for men to stalk or sexually assault women?
A few days ago a video leaked of a woman riding in a Mexican taxi, who was demanding the driver went faster. He refused because it'd be dangerous, and she immediately started threatening to report him as a harasser to the police. She even said he had to speed up or else the police would be waiting for him when they got there. She didn't realize her whole conversation was recorded on camera.
A lot of men have had experiences like this one. Either directly or they know someone it happened to. Yeah #NotAllWomen but way too many will exploit the feminist #BelieveAllWomen culture to gain even trivial benefits. An app devoted to letting women anonymous gossip and engage in reputation warfare without fear of consequence, or even fear that the man might reply in self defense, is going to get flooded with women like the taxi passenger.
"A lot of men" is doing a lot of heavy lifting here.
Go read some statistics on the number of women harassed, abused, raped, and killed every day—every single day—because they are women.
Go ask your mother, your sister, your wife, your female best friend, when they had their last abusive encounter.
Go ask your friends of both genders what the worst things are that could happen to them when walking home at night, and compare the responses.
Go read some historic accounts of how women were treated for… pretty much all of history.
Go look up news articles of what can happen to women when riding a taxi. Spoiler: it’s not just a threat.
Yes, there are some abusive women out there. Yes, it’s fucked up when that happens to you. But trying to insinuate the levels of violence against men would be even remotely comparable is just plain awful.
By the time a man has hit his 40’s, it is exceptionally uncommon he hasn’t seen someone hit with a false rape claim - or had one himself - by a vindictive ex. Or has been threatened with (or directly attacked) with physical violence.
By people going on the same sort of rants like you just did.
Some People are terrible, especially when they think they can act without consequences.
Does that excuse men doing bad things too? No.
But it sure does (or should!) make anyone with a brain question hyperbolic claims of abuse or violence without actual evidence.
The problem is that you're equating the wide range of violence against women with a specific kind of violence against men by calling both "bad things", insinuating those are even remotely comparable. They are not. 90% of rape victims are female. In the US alone, every 68 seconds, a woman is sexually assaulted.
After the big war, some Germans were quick to point out that their people had suffered when they were displaced from the land they occupied in Poland, for example, and that "both sides had suffered". I assume you're also incapable of understanding why the victims of the Nazi regime were completely aghast by that?
> But it sure does (or should!) make anyone with a brain question hyperbolic claims of abuse or violence without actual evidence.
What do you suggest to do instead? Sexual violence is often a crime with only the perpetrator(s) and the victim as witnesses. In most cases, rape doesn't leave persistent traces. Rape victims tend to be in shock, however, and often need time to process what happened. Your suggestion seems to be that we should question these claims?
Judging these cases correctly is incredibly complicated, and claims of wide swaths of men falling prey to abusive women don't really help anyone affected.
There is a reason ‘he said, she said’ is widely known as the shittiest type of situation, eh?
Yes, we should question those claims, and any others. Or everyone who wants to be shitty will do it via that route. It’s basic shitty human behavior.
That it screws actual victims is why people gaming the system should be punished.
But not challenging these claims just makes more victims too. And eventually people will just tune out accusations, because the shittiness has gotten too pervasive. And then the predators/shitty humans will get be doing more actual rape eh? Which is terrible.
This is why it’s also prudent to be very careful who anyone is alone with, favor video recording of public spaces, etc. as well. Because the best way to avoid a situation is to make it as difficult as possible for the situation to occur, and minimize the chances of any ambiguity. Which is also shitty for everyone.
Personally, I also don’t trust the stats because I’ve seen many (5+) women retcon clearly consensual behavior (that they were even bragging about before!) into ‘he raped me’ when someone tried to shame them for it later, or there was some leverage they could get out of it. I had one who literally admitted to me when I investigated that she was doing it to punish the guy for refusing to date her later. Another was fine until she went home and her mom gave her crap about her dating behavior, and then all the sudden it was rape. Until we started to interview her for her story, and then she admitted it was consensual.
I very much believe actual rapes and SA’s occur. I personally have literally never seen an accusation for rape or SA that stood up to even the lightest scrutiny, within the environments I’ve been responsible for. And not because I was trying to avoid them!
The joys of being a manager of mixed sex groups eh?
If we could figure out the actual truth of these situations, then we could punish actual offenders and not constantly be in this BS situation.
I do. Not as an indictment of women but an indictment of social apps. Apps like this are way too hard to moderate, manage and verify. They quickly get swarmed by bad actors and misused. Again, not because women don't have genuine safety concerns in the dating world but because apps are not a viable way to manage those concerns.
Some social problems just don't have technological solutions.
You know the answer to that is zero. There is no viable system a company, let alone a small unfunded startup, could use to verify the identity of the reporters let alone guarantee the trustworthiness of the account.
Those ten reports could be made by one person. That one person might not even know the person they're accusing. That one person might be a man. That one person might be a bot.
You'd have to ignore the last three decades of online identity, trolling and social media pitfalls to not recognize that.
And please don't compare reviewing a can opener on Amazon to accusing someone anonymously of a heinous crime on an app built by one person.
But I'm not sure I'm going to convince you with words so I'll suggest this:
Go and build this app.
Build it, see what happens. Nobody else has been able to crack this but maybe you can.
I'm sorry and I'll be voted down for this, but I do think that it will attract plenty of fibbing and deliberate or not-so-deliberate stretching of the truth. Anyone who is rejected tends to be a bit angry about it. In this case, women who are ghosted can say whatever they want.
This isn't all of the people, but in my experience in life it's more than enough to make this app impossible to filter.
> Do you think a women's dating safety app is mainly about women lying
That's not what it is intended for, but many people after relationships end can be extremely emotional and sometimes very spiteful. It's not uncommon for people to embellish or lie about the truth to make themselves look better and the other person look shitty. Especially if you're the one being dumped, you may be even more likely to engage in petty behaviour.
I personally have experienced an ex making up a sexual assault story. This kind of app didn't exist then, but she even went as far as reporting me to the police. Luckily the police investigated and could easily discern it was a lie. Going to the police is obviously a much higher burden than using an app, and yet many females still go make false SA claims there. Do you really think it wouldn't be a common problem for people to do the same in an app at a much higher rate?
People often believe things like SA claims without any evidence and will often even attack people trying to defend the person or insist on some kind of proof. It means that someone making up bull crap on these apps is going to be treated like it is true, yet the rates of lies would likely be pretty high.
People can just be so crazy when it comes to relationships/love. Especially when it comes to people in their teens or early 20's, the brain isn't fully developed and dealing with these emotions is even more challenging and leads to even more rash decision making.
We grant a tremendous amount of leeway and power to accusations made by women against men in society today. There are always honest people using things for their intended purpose. Though they are also dishonest people using things for their own ulterior motives.
A well-designed system will maximize utility for the former, and minimize utility for the latter. An app where women can leave what are practically anonymous reviews for men is not such a system.
What was leaked was women's personal data, like driver's licenses. What they shared with each other was their experiences with men who sexually assaulted them or stalked them and their names, not the men's personal data.
Men's driver licenses were not distributed online. Only women's driver licenses were distributed online.
I'm not familiar with this app, but surely those accusations of sexual assault are only useful to other users of the men are sufficiently well identified?
The article says that what gets shared with the app is a picture of the man, and it's not just "those who sexually assaulted them or stalked them" but anyone they want feedback about.
I assume the app then runs facial recognition.
This may be legal in the US, but not under GDPR. Pictures of faces are biometric data (explicitly listed as such), which falls under additional restrictions beyond personally identifiable information.
A drivers license with the picture blacked out would be less sensitive than the picture itself!
There are soo many examples from the US showing why GDPR is a good thing: Clearview AI (biometric mass surveillance, essentially "search the internet by face"), car manufacturers collecting and selling location data, phone companies collecting and selling location data, ISPs collecting and selling browsing behavior, companies running mass surveillance on license plates and selling the data to law enforcement and really anyone who pays, some DNA sequencing related abuses that I don't remember the details of, all the data collected by the ad "ecosystem" (note that this still happens in GDPR-land because enforcement is lacking), this, ...
You need to do this for background checks for employment, even though the employees for the background check service might be outsourced to a different country, and your government data had no protections in their jurisdiction.
I absolutely realize that this is a function of privilege but I've declined to participate in these sorts of background checks and have still been hired on a few occasions.
It's also a fundamental misunderstanding of how LLMs work, mixing up inference with training.
reply