So it sounds like the spirit of what this person is saying is, if somebody bad were to use the phrase "don't do genocide", then we should go consult people being genocided and ask them if it's okay to say "don't do genocide" because it's not really clear if it's the right thing to say, since somebody bad used it. 🫠
“On a snowy October night, I was eating a microwaved burrito, when the editor called and ordered me to go meet some actually interesting people”
"As soon as I stepped outside, it started to rain. The droplets gathered on the ground and I had a profound realization. Gravity pulls things down. I wondered: Have the Chinese ever figured this out? No, surely not. The Russians? Not in a million years. But I, a white man living in New York writing for an imperialist publication, am the first to have ever had this realization. I quickly jotted my thought down on a used napkin, ignoring the properties of water that made it wet in the rain, and hurried to the patent office to get my scientific discovery recorded in the annals of capitalism."
Lol, that is a funny coincidence. Hope you enjoy it!
Oh yeah, on some apps, there definitely seems to be profiles that are liking everybody. I've heard it speculated (not sure if confirmed) that they tend to give you a visibility boost when you are a new profile, which is why you can end up with a few likes in the beginning then nothing. Though with Hinge, I found I wouldn't necessarily get any even when new, possibly because Likes are more limited per day than some apps and it allows you to send a message with a like, so it's a bit more conscious.
Np, hope it helps (and feel free to let me know how it goes, I'm curious to know if it works for anyone else). And yeah, I remember trying CharacterAI briefly in the past for language learning myself; main difference I find with this type of setup, is 1) The corrections are "on the side", not organically part of the conversation itself, so you have the main conversation which the AI is focused on and then you have suggestions/corrections you get if you say something its evaluation thinks has incorrectness for that language. 2) Both apps have forms of conversation where they can be more unguided in terms of what you talk about and have "roleplay" type of scenarios that are a little more structured for practicing specific kinds of things. For example, I was just doing a little talking with a "roleplay" on Tutor Lily called "explaining symptoms to a doctor."
So if I compare it to trying to learn through just any LLM, point 1 seems to be the most significant difference. Since technically you could already get an LLM to roleplay most things, albeit with more effort than with these apps' scenarios. But getting corrections on the side while talking to an LLM seems to be a more specific engineering/design thing that goes beyond LLMs alone. I might try to ask the creators and see if I can get an answer, but I'm unclear still on whether these apps are powering the corrections purely through an LLM itself or some other kind of AI evaluation with it.
There's one I've been on for... I'm not sure, could be more than a year, though I was inactive for a while during periods of that. I have a grand total of 4 likes and 3 of them were recently when I was inactive, which seemed like the app was giving me a bit more visibility to draw me back in. (And none of them are people I want to match with.)
Based on the things I've heard, the game seems to be that these apps tend to "rank" you early on and then from there, you're mostly stuck where you are unless you pay to get more visibility. And because rejections are not something you "see" unless you match and then the other person unmatches, you have no way of knowing for sure if your interactions (likes, or on Hinge, messages without being matched) are being seen by anyone or if they are buried in the stack.
I know on Hinge, from the end of receiving attention, there's a limited number of likes you can see at a time without paying. So presumably that means that if, for example, somebody gets flooded with likes/comments and gets 100 of them, they'd have to go through and match or reject with each one to see all of them if they are a free user. And because some women get flooded with more attention than they have the time to engage with, that effectively means you might never get seen at all.
I know that's not exactly an encouraging way of looking at it, but considering the mechanisms of it helps remind me that it's likely not something to do with me and is far more likely I got in a bad spot early with the "ranking" and can't get out of it without paying. So sometimes I go through it for the hell of it to remove people in my stack, or send out the occasional like/message, but I try not to spend too much time on it when it's designed to work against me.
So, I don't know about Spanish specifically (I've been focused on Mandarin myself), but I do have an idea you can try, that I've been experimenting with. There's an app called Tutor Lily and one called Univerbal, and both make use of a combination of AI text generation / languages models (for one side of a conversation) and some kind of "corrections" to point out mistakes in grammar or that kind of thing. Both have a form of trial or free use before you have to pay for anything (though with Univerbal, you might have to sign up for the full trial and then cancel in your store subscriptions if you don't want to get charged after it ends).
I haven't tried these for long yet, but I'm liking Tutor Lily more so far for just basic chatting and writing a message in the language every now and then during a day. And the main benefit I find is in getting me to use the language more flexibly vs. the set examples that course-based apps tend to have. In this way, one thing I do sometimes is I'll go find a grammar rule I learned in the past, then try to make use of it in the chatting app, within the context of the conversation. I reach for Google Translate a fair bit during this, as my understanding of Mandarin is still limited, but I know enough that it's not too painful.
The one main caveat here is, LLMs (large language models) can "hallucinate", as in confidently BS on things. Presumably, these apps are designed to try to confine the output to some degree and are (hopefully?) specially trained on grammatical corrections and the like to make them more accurate. But I still take it and its corrections with a grain of salt when conversing with it. So it's not something I'd recommend for someone new to a language who is very dependent on correct info, but at the level you're at, it might be a helpful supplementary thing.
Edit: Just noticed you said you have ADHD. I recommend trying this approach all the more in that case. I'm pretty confident I have ADHD too and the interactivity of this approach helps me stay engaged with it.
It's interesting to hear your experience on it. I've long felt that the visceral hatred some places have for self-promotion is odd, but I have no evidence to back up how it goes in practice. Personally, what bothers me most about self-promotion or large scale corporate stuff is when it's sneaky and manipulative as such. I would much prefer to encounter someone who is open about it than come across a thing that has "hello fellow kids" energy and it's some marketing employee for a corporation. The problem with the sneaky way being that it blurs the lines between casual social interaction and formal business transaction; which I don't think is necessarily bad intrinsically? But when it's done expressly for the purpose of manipulating and has no sincere relationship building behind it, it's creepy as hell. The nature of business under capitalism means you kinda need to draw some lines between "casual social interaction and formal business transaction" or you're going to get taken advantage of.
Anyway, all of this is to say, I think you make a good point and I don't think self-promotion is intrinsically bad either. And in fact, being hard on it probably pushes more of it "underground" so to speak, to the creepy subtle stuff.
Yes and no. It's not a solved problem, but a worked around problem. Diffusion models struggle with parts that are especially small and would normally have to be done with precision to look right. Some tech does better on this, by increasing the resolution (so that otherwise smaller parts come out bigger) and/or by tuning the model such that it's stiffer in what it can do but some of the worst renders are less likely.
In other words, fine detail is still a problem in diffusion models. Hands are related to it some of the time, but are not the entirety of it. Hands were kind of like a symptom of the fine detail problem, but now that they've made hands better, they haven't fixed that problem (at least not in entirety and fixing it in entirety might not be possible within the diffusion architecture). So it's sorta like they've treated the symptoms more so.
Surreal. This while Columbia University students protesting genocide are facing threat of the national guard being brought in, while students at University of Austin Texas are dealing with violent cops just for peacefully demonstrating.
Then there's the part where they are so deep in paranoia and racism they think you are a foreign spy if you say anything sympathetic about the country. (I actually had this happen to me once online.)
I feel that. It's especially awful to me how much of the US, there is no alternative and you're essentially being forced to do a dangerous high-risk activity on a regular basis just to get to and from places for basic survival (that is, if you can afford a car and are capable of driving and so on - and if you can't, you may be dependent on someone else doing that dangerous high-risk activity). The amount of cumulative stress from that, not even getting into the number of injuries and deaths, has gotta add up to a lot.
And the whole concept of driving as the main means of getting around is so backwards and ineffectual that the richest people bypass it and use private planes instead. All the traffic, the accommodating different directions people are coming from, lights, stop signs, turnoffs, you end up with so much starting and stopping, it's wasteful for gas use, wastes time, and isn't even freeing like it gets portrayed as because you're highly limited by the roads and their traffic patterns and design. And don't get me started on how so many road designs have no consistency at all and vary widely from moment to moment, because forbid any of it makes any sense and was planned ahead on.