clbrmbr

17 ms·

271.
by clbrmbr 3y ago
I’m curious how you find the Swedish from different models. GPT-4 seems to return perfectly grammatical Swedish but a Swede friend says it reads like English. Do you notice this? I’d love to have models that are better at idiomatic usage of
272.
by clbrmbr 3y ago
I’ve been doing this in my private codebase. When copilot hallucinates a function, I just go and write the thing. It’s usually a good idea, and it will re-hallucinate the same function independently in another file.
273.
by clbrmbr 3y ago
I’m finding it’s more effective (and pleasurable) to write using GitHub CoPilot and CMD-RIGHT (accept next word). I put a detailed doc comment above and write in tandem with copilot. I’ve written the structure and I review as I write jointl
274.
by clbrmbr 3y ago
> There's no takesies backies with autoregressive models Doesn’t ChatGPT use beam search?
275.
by clbrmbr 3y ago
How to load on your repo?
276.
by clbrmbr 3y ago
This. Whisper is phenomenal. Have you tried the conversational mode? I would love to be able to use that in a more customized agent. I know you can use the conversation mode with a custom GPT but I’d prefer to write dynamic prompts programm
277.
by clbrmbr 3y ago
GP meant Custom GPTs. Confusing names for sure.
278.
by clbrmbr 3y ago
1. Cost & resource optimization 2. More and more RLHF
279.
by clbrmbr 3y ago
Oberon sounds interesting. How refreshing it could be to have a simple computer system. Even my little IoT project has grown over the years to a rather ungainly proportion, requiring 20min in CI to build many targets. Matter is really the l
280.
by clbrmbr 3y ago
I’m a bit concerned that the ease of modifying the parameters will lead to Congestion Collapse. It’s been possible to cheat with TCP but most people don’t bother because it would require a kernel recompilation. Or did this ship sail a while
281.
by clbrmbr 3y ago
I hope regulators watch video from the vehicle and think, “damn I would not have been able to react that fast. Sure we can make these things even safer but it would be immoral to block waymo’s operations as it would certainly increase harm.
282.
by clbrmbr 3y ago
Beautifully composed article. Looking forward to trying this out.
283.
by clbrmbr 3y ago
> I would expect candidates for programming jobs to demonstrate first class ChatGPT or other code copilot skills. Agree. But two challenges: if the interviewer does not make it clear that ChatGPT/SO may be used, the typical assumpti
284.
by clbrmbr 3y ago
I cannot tell if this comment was made in just or in earnest. As far as I understand, the earlier GPT generations required a fixed amount of compute per token inferred. But given the tremendous load on their systems, I wouldn’t be surprised
285.
by clbrmbr 3y ago
> turning stories into language learning content Great idea. It could be simple cloze deletion from the content pasted in, or are you thinking to use AI to create simplified expressions with similar semantics?
286.
by clbrmbr 3y ago
I look forward to giving user feedback when I’m able to signup. > our Humane Spaced repeition system. Based off of the popular FSRS, our Spaced Repeition System will adapt to your review style. Take a few weeks off? No problem. Want to s
287.
by clbrmbr 3y ago
As I’ve commented elsewhere, to the extent you are using machine translation: note that GPT is bad at translation to languages other than English. Telltale sign is not declining for gender. I saw this a lot with GPT4’s Polish. Better to use
288.
by clbrmbr 3y ago
Honestly, I think what worked best for me was CSV import. The downside of course is that I cannot easily edit or add to the deck in this way. I should emphasize that in the process of going through the deck I was able to improve quality sig
289.
by clbrmbr 3y ago
The real delight point for me was how painless it was to mark 700 words as already known and pickup a bunch of fairly common words i am not comfortable using (like “allerdings” and “eigentlich”) using that scroll-to-set-learned control.
290.
by clbrmbr 3y ago
I’d recommend using GPT4 to generate English sentences and then Google Translate to get the German. Harder to get example sentences for a specific German word, but may improve the quality of the German. (I had this issue when generating spe
291.
by clbrmbr 3y ago
Dude. You rock at UX design. This is amazing… Only trouble I have tbh is figuring out what I want to read in German! But I’ve signed up and will give this a spin, and send what feedback I can.
292.
by clbrmbr 3y ago
I used iNaturalist website to list 500 most observed plant species in Eastern USA and Canada, and used a script to scrape the species names (scientific and common) and the URL of the title photo. I used another script to download the images
293.
by clbrmbr 3y ago
I used Anki to learn to recognize the top 300 plant species (and top mushroom species) in my area by sight. Totally changed my perception of nature.
294.
by clbrmbr 3y ago
> what the ecosystem would look like There is a third-party paid exchange here: https://www.ankihub.net/ (I’ve had a few exchanges with the founder, and wish him luck.)
295.
by clbrmbr 3y ago
This is how I feel about the public NTP pool on which so many IoT devices depend.
296.
by clbrmbr 3y ago
Verification of time-sensitive rare events from strangers is hard. But I can see how that could really increase your potential winnings.
297.
by clbrmbr 3y ago
Hmm. So a rouge Meta employee could put a barometric pressure monitor in Messenger or WhatsApp (that people use even without paid in-flight wifi) to detect depressurization and automatically short the airline and aircraft manufacturer.
298.
by clbrmbr 3y ago
What else is astronomical data used for in litigation other than determining (upper bound on) illumination levels?
299.
by clbrmbr 3y ago
Would be nice if this article included some equations!
300.
by clbrmbr 3y ago
But those are problems that usually require writing and running some code. OP’s problems are intended for pencil and paper.

More ›