Yesterday's wonders

AI is all the rage right now. Something changed; opening HackerNews 4 months ago, you would have been bombarded by posts about Rust. Right now, everyone speaks about AI, Large Language Models and such. Even Wozniak and Musk are signing a petition urging everyone to stop for a minute, and think. It feels like there are two camps in this trend (or rather three) :

• the enthusiasts, which are already feeding ChatGPT to itself and creating their home version of Jarvis
• the alarmists, which are worried about the influence these tools will have on jobs, on security, and more;
• the silent majority, of which I am a part of, that does not really know what to do with this information.

I don’t know what to think about AI. On one hand, I’m all for progress and the development of better tools; on the other, I can’t help but feel like AI is one these things that people create just because they can. One such field in my opinion is voice reproduction. Nowadays, you can take a sample of someone’s voice, feed to an AI such as Microsoft’s VALL-E, which will then reproduce it saying whatever you want. A 3 second base sample is enough to have convincing results. Now, that is an amazing technology for sure; text-to-speech has been around for a long time and is certainly extremely useful to thousands of people around the world. But voice reproduction ? What is the use for such a tool, apart from impersonation ? This technique has already been used to scam people; the good old “call some corporate accountant pretending you’re the CEO” technique just got suddenly a lot easier. I could not find any usecases for it coming from Microsoft…

This post is not really about AI. As I said, everyone talks about it, and I don’t have anything smart to say about it. My goal was just to share a feeling : I’m growing tired of these technical prowesses. Take email : 30 years ago, it was a wonder. You could write to anybody, anywhere, in a matter of minutes. At its core, SMTP is a simple protocol. This was pretty elegant. Where are we at today ?

Part of my job as a sysadmin is managing the email systems of my workplace. We have our own email servers, with a few hundreds of users; not exactly tiny, but we are not big players either. And guess what ? Emails are the main vector of attack on our systems. Social engineering is so powerful; email makes it so easy. We have to constantly get in touch with users who clicked on fraudulent links, got their passwords stolen, etc. Nowadays, you have to use 2FA to secure everything, which makes the process of logging in tedious, etc. This situation sucks ! Collectively, we have this system that I still find amazing : thousands of servers around the world, all of them talking more or less the same language (SMTP, since I am talking about emails). And nowadays, all that power is wasted on spam, spam, spam. I hate emails. It could have been so nice, but it sucks. Everyone is using Gmail to avoid the hassle, but that’s not a solution either, since Google is reading all that you send and receive. There are solutions, sure (I use Proton, which is pretty good), but the underlying system won’t change, and spammers won’t stop spamming.

So yeah, about AI : I hope Wozniak and Musk are not the only ones willing to stop and think about the implications. It’s not that I’m scared of ChatGPT reaching singularity, or anything of that sort : just, it would be nice to have something well implemented, that will be elegant and frugal in the future, that everyone will be able to use without having to resort to selling their soul and privacy to some corporation. And also, things that are truly useful and safe.

I hope I make sense here. I’m not sure, to be honest. If I’m talking stupid, or you have something to say, write to blog@squi.fr - I will be happy to read your comments; not only I because like learning, but also because it will prove to me that emails can still carry some human interaction.