139 karma · joined July 11, 2018
Obviously we’ve got to tame the version of LLMs we’ve got now, and this kind of thinking is a step in the right direction. What I take issue with is the way this thinking is couched as a revolutionary silver bullet.
I've been looking at these recently but not specifically to solve "LLM issues".
What oddly coincidental timing ... I finished setup of Tailscale just yesterday and ran into this exact issue when testing it. I didn't think too much of it and blamed the USB connection I'm using to connect my external drive.
I'm surprised more effort wasn't put in to making the habitat homely. It's very austere and scientific. I'd have thought that this experiment would also provide means to test ways to provide comfort in trying situations.
I love that ListLang includes listening exercises too. Vocab + listening is the way to go!
Thank you for this. It's an incredibly polished app so far, very excited to see how it progresses. Ideally you'll avoid Duolingo's heinous use of forced ads. Anki on IOS costs $40 or so. I'd pay that for ListLang provided it proves itself overtime.
What would be helpful: 1. An option to use a classic flashcard format, with one word on it. Fill-in-the-blank can feel a bit hand-holdy. 2. Ability to do listening practice only on sentences learnt so far. I want to practice listening as I learn new words. 3. Sentences in listening practice that use learnt words, but new sentences different to the one used in 'Learn'.
What I found most powerful was the ability to interrogate its responses. I can ask broad/general questions, and then follow-up with 'why is X used in this way? How else can I do that?' or 'you mentioned Y, how does it work and why isn't X used instead?'.
I had ChatGPT explain to me how Rust's implementation of `Vec` works, and importantly, WHY things have been done in a certain way, e.g. why a separate struct called `RawVec` is used.
This may be simple information, but to be able to get it by asking questions in ways that make sense to ME, is so valuable.
I experienced this in practice. I was pairing with an inexperienced engineer who was using Copilot. He was blindly accepting every Copilot suggestion that came up.
When I expressed doubt in the generated code (incorrect logic + unnecessarily complex syntax), he didn't believe me and instead trusted that the AI was right.
Thanks for the insight.