#protonmail

Em uma controvérsia que mistura os conceitos de privacidade digital e vigilância governamental, o Proton Mail, um popular serviço de email criptografado, é acusado de facilitar a identificação de um manifestante vinculada ao movimento "Stop Cop City" em Atlanta. Relatórios recentes indicam que o FBI teria utilizado informações fornecidas pelas autoridades suíças para desmascarar a identidade de um usuário anônimo que se opunha à construção de um centro de policiamento na cidade.(...)

Leia a noticia completa no link abaixo:

https://www.jornalo.com.br/proton-mail-e-usado-pelo-fbi-para-identificar-manifestante-anonimo

Oh! Well I feel disappointed in Proton, but pleasently surprised that someone thought to ask me! To be frank, I'm not that deep into the Proton ecosystem (i mainly just use their VPN and sometimes Proton Drive), and there are countless other videos and resources that go more in detail about AI, but I can give you a basic summary of what all the fuss is about (with some research on my end first lol).

For context: Proton is a non-profit organization dedicated to internet privacy, and so has developed many apps (mainly alternatives to mainstream software like Powerpoint or Gmail) with user-privacy in mind. Recently, they've released a new AI model, Lumo, as a response to the whole AI craze that's grabbed the tech industry by the balls. Now for reasons we'll get into later, a lot of people REALLY don't like AI. However, they haven't made any efforts to integrate this into Proton Mail, so on that front, you don't have to worry! The only Proton app with Lumo intregration is Proton Drive - more like the other way around actually - where you can send your Proton Drive files to be analysed by Lumo. No, Lumo cannot see the files in your drive, only the ones you upload to its chat, and even then it immediately deletes them (more on this later).

So ig TL;DR: you don't to worry about AI being forced into your Proton apps.

First off, I think I should explain how it works, so that you have a better understanding of where its issues come from. Basically, AI models are kinda like prediction algorithms that are trained on vast amounts of training data to recognise patterns in said data and apply those patterns when giving outputs (e.g. AI being given images of lungs with lung cancer to predict whether or not any new patients have it - This is a good use of AI!). As for Generative AI, the LLM (large language model) is given lots of text files so that it can predict what words to use nest in a conversation given the prompt. context clues, and its training data. Probably not the best explanantion but that's like the gist of it.

As for why GenAI is so widely hated, at least when it comes to text-based models, I'll try to keep it simple:

The LLM needs a lot, and I mean A LOT, of training data. More than the companies' themselves can produce in-house. The solution? Scrape data off the internet of course! While there are companies that claim they only use copyright-free texts/images, the majority of them will take anything online to feed into the LLM

This opens up a lot of concerns: Did the people whose data was used consent to this? No! Will they be repaid for having their work stolen? No!

The quality of that data is also a concern. The ENTIRE Internet is getting scraped for this, meaning the best and worst of humanity is being fed into something without thought: Anything from Reddit shitposts to porn to Facefook Mom groups to literal hate speech! How can we trust it to not give misinformation or to say something unhinged? (Musk's Grok AI is a perfect example of this, as it regularly spits out far-right bullshit and general bigotry because it's been trained on Twitter and is owned by a member of the fascist Republican Party)

AI is also a massive fucking plagarist. Because it's a prediction model, it needs other people's data to function. As such, it'll often lift phrases from other works (mainly like fanfiction) because it's going off its training data. Again, these people are having their works used for training data with no credit, consent or reparations. THIS IS PLAGIARISM WITH EXTRA STEPS

It's also being used to replace artists and writers! Because writing a prompt into ChatGPT is way faster and cheaper than hiring and waiting for a professional artist, many artists' jobs have been cut or at risk! This is parrt of a greater issue about the commodification of the arts and people tend to focus on the final product rather than the process of making something, but that's a WHOLE other topic that I'm not touching because this post would be way too long. Point is PEOPLE'S WORKS ARE BEING STOLEN TO REPLACE THEM

These little shits also need datacenters to hold all that stolen data and to actually hold all the code and processing power. They require a lot of land, water (to prevent overheating) and energy, and also processors. This is HORRIBlLE for the environment! We're already facing water shortages in places near data centres, and burning more fossil fuels when countries are recording record-high summers sure isn't helping global warming, that's for sure!

There's also the fact that data from chat logs are ALSO used as training data! None of your conversations are private! It's just another way for companies to monitor you and force personalized ads down your throats!

The AI models have also proven to be total suck-ups, as to not upset users. This means that if you tell it something, it'll almost ALWAYS side with you. This is horrible for more vulnerable people who may feel isolated, as the AI ends up affirming any delusions thrown its way and gives you "advice" (this has been called "Ai Psychosis and there's actually a really good documentery-style video by Eddie Burbank that explores this in more detail). There have been cases where people have been driven to SUICIDE because they trusted the AI as a friend and it hyped up their suicidal thoughts and helped them do it - I wish I was making this up but google it up and you'll see plenty of videos and news outlets talking about it

This brings me into the next point: AI addiction. Because these LLMs are designed to be friendly and reassuring, people who are lonely and are in a rough period in their lives may use it as a friend or therapist and then get hooked to this simulation of friendship because it feels like a friend and meeting new people can be difficult and anxiety-inducing, so they choose the non-judgemental, always available chatbot over real human connections

... i think that's most of the points? While AI does have genuinely good uses, GenAI is not one of them. Now onto how Lumo is different.

As it's a Proton product, it has a greater focus on privacy than most AI models, according to them: All conversations are zero-access encrypted so no one else can access them; Proton doesn't keep logs on your usage or store your chats on their servers; Your chats aren't fed into Lumo or sent to third-parties (as seen here). Admittedly I'm only using them as a source, but considering Proton's whole mission statement is about privacy, I'm inclined to believe them.

TL;DR: Is it more private than other LLMs? Yes. Does it still have all the other issues? ...More or less (while I haven't heard of any AI psychosis cases ot anything mental-health-related surrounding Lumo, the risk always there).

also trying out Thunderbird and that shit SLAP!!! I really enjoy the Outlook layout of emails and being able to use folders, which Gmail doesn't have (labels only) so I'm really hoping I can configure it to my needs!