I’ve found this to be the case a lot, too. I also spoof my OS because a lot of government sites will refuse to work unless it says Windows. It’s stupid, but here we are.

YES! I study AI, and this is exactly how I feel!

Side note-One of my favorite things to do is ask people what their use case for using AI is, and watch them sputter out “uh…emails and productivity and things.”

I got a laptop back in 2018, and it shipped really fast. It’s not my daily driver, but it works well when I’m on the road, and the battery life is pretty good. Granted, I replaced the OS with a distro I prefer and customized the hell out of it, so that might contribute to my experience. Tbh, I was pretty impressed with it (still am), and I was going to buy a Librem 5 when they came out. I wanted to wait and not just throw money at them because I didn’t want to get burned. After all the horror stories and crap reviews, I passed on that and won’t touch the company with a 10 foot pole, and I thank past me for not throwing money at them.

I think that the company started with noble intentions and made a decent product at first, but they got in way over their heads and now they’re floundering.

The original paper itself, for those who are interested.

Overall, this is really interesting research and a really good “first step.” I will be interested to see if this can be replicated on other models. One thing that really stood out, though, was that certain details are obfuscated because of Sonnet being proprietary. Hopefully follow-on work is done on one of the open source models to confirm the method.

One of the notable limitations is quantifying activation’s correlation to text meaning, which will make any sort of controls difficult. Sure, you can just massively increase or decrease a weight, and for some things that will be fine, but for real manual fine tuning, that will prove to be a difficulty.

I suspect this method is likely generalizable (maybe with some tweaks?), and I’d really be interested to see how this type of analysis could be done on other neural networks.

This is awesome!

It’s not just convenient for them to do it; it’s how they are able to evade anti-trust action (not that the U.S. is great at it anyway but still). I also run my own mail server. It’s not impossible, and I wouldn’t even say it’s even hard. It’s just time consuming to set up (if it’s the first time), and there are a lot of hurdles to make it so impractical that it’s virtually impossible to the average person. Only the most patient or those who have a real desire to run their own mail server will even attempt it. Anyone can set up their own mail server, but most won’t because it’s not worth it compared to using something that just works from Google.

Congratulations on making the switch! I remember when I switched full time almost 10 years ago. It always feels like there’s something new to explore or to try with your computer. One of the most freeing things I learned was that most things are within my grasp if I put in the effort to learn about it. There’s nothing quite as fun as whittling the day away going down a configuration rabbit-hole to make something just right.

This is a much better article. OP’s article just shows the author’s surface understanding of how coding works and how well an LLM can actually code. There’s way more that goes into a programming task than just coding.

I see LLMs as having the potential of being almost like a super library. I can prompt GPT, Claude, etc. to write me a custom function that I copy, paste, test, scrutinize, and almost certainly change. It’s a tool that will make someone a more productive programmer. It won’t completely subsume a human’s ability to be creative and put the pieces together.

At the absolute worst over the next decade, I could see programming changing from writing and debugging code to prompting, stitching together, and debugging.

Yeah, I didn’t even get to say that I could change it (though I don’t recommend it) before she wanted to throw the whole thing out for not being “user friendly” enough.

Oh for sure! Sometimes it’s not even when something breaks but just a normal thing that’s different. I used to be a Linux evangelist, and when I convinced my to mom to simply try Linux, she was upset when she had to enter her password to do something (I think it was an update or something) rather than it just doing it. She was mad that it prompted for a password rather than “just updating.”

Explaining that giving permission is much safer than just running everything as Admin did nothing. She hasn’t used Linux since.

It’s funny you say that. I find the Linux way of getting software way more intuitive. Just hop in the terminal and use the package manager. When I used Windows, I always felt like I was doing something shady when I was getting a .exe. With drivers, I’ve only had an issue once; everything else was pre-compiled into the kernel. On Windows, I had driver issues a lot. For those reasons (and others), I switched full time to Linux almost a decade ago.

Totally anecdotal, of course, but I just thought it was funny how our experiences were complete opposites and sent us in complete opposite directions for the same reason.

This used to be me while I distrohopped. Now, I’ve settled on which distro I will use. I use Arch, btw.

Yeah, you’re right on a lot of chatbots just being paraphrased responses from the support database, but for a lot of people, that’s all they want or need. There are a great number of people who just don’t want to read the entire article to find their answer. For that, I don’t really mind chatbots because I get the use case. What I hate is when there isn’t an option to go to the next tier of support without going in circles forever with the stupid bot.