barrell

32 ms·

361.
by barrell 2y ago
I've worked with clojure for years, and it becomes pretty normal, but I don't think it inherently get's better than the alternative, just on par... although it's realllly nice not to ever have to think about operator pre
362.
by barrell 2y ago
Yes I watched the demo. True those things were not possible, so if that’s what’s blowing you away then fair enough I guess. For me that doesn’t impact at all anything have ever used voice for or probably will ever use voice for. I’ve voice
363.
by barrell 2y ago
Honestly, as someone who has been using this functionality almost daily for months now, the times that break immersion the most by far is when it does human-like things, such as clearing its throat, pandering, or attaching emotions to its
364.
by barrell 2y ago
Flat out impossible? If you mean “without clicking anything”, sure, but you could interrupt with your thumb, exit chat to send images and go back (maybe video too, I’ve never had any need), and honestly the 2-3 second response time never on
365.
by barrell 2y ago
Did you use any of the GPT voice features before? I’m curious whether this reaction is to the modality or the model. Don’t get me wrong, excited about this update, but I’m struggling to see what is so magical about it. Then again, I’ve been
366.
by barrell 2y ago
Well, I for one am excited about this update, and skeptical about the AI scaling, and agree with everything said in the top comment. I saw the update, was a little like “meh,” and was relieved to see that some people had the same reaction a
367.
by barrell 2y ago
I’m working on phrasing.app partially to solve this problem. My solutions are: - study flashcards from l3 <-> l4 to cut down on time - practice switching between the languages and distinguishing between them (I use interleaved flashca
368.
by barrell 2y ago
There’s a hint function on phrasing.app :) we’re launching the public beta this week sometime
369.
by barrell 2y ago
This feels disingenuous. You may have coded everything from scratch, but I couldn’t tell if this was a linear plugin or separate app from the first few seconds. To say linear was just a source of inspiration is just not true. Call a spade a
370.
by barrell 2y ago
Meta Llama 3 8B vs Claude Haiku according to their press releases if anyone else was curious | Benchmark | Llama3 8B | Claude Haiku | | ------------- | ----------- | ------------ | | MMLU ____ | 68.4 ____ | 75.2 _______ | | GPQA ____ | 34.2
371.
by barrell 3y ago
Even in this example, I don’t think that the middle of lions, tigers, and bears would be further away from lions than tigers. But it might also help to understand that most often your embedding smaller chunks of data, specifically for the r
372.
by barrell 3y ago
Agreed. I use LLMs quite extensively and the amount of production code I ship from an LLM is next to zero. I even wrote a majority of my codebase in Python despite not knowing Python precisely because I would get the best recommendations fr
373.
by barrell 3y ago
While I do agree there is some amount of secret sauce, keep in mind the training takes several months. So from someone to see the success of GPT4, decide they want to invest that amount of money to train the same, raise the money to train t
374.
by barrell 3y ago
The comments in this thread alone would indicate the the crowd is pretty split even after this blog post
375.
by barrell 3y ago
I don’t think anything about this is cut and dry. You may have a strong opinion on the matter, but at the most charitable it’s people doing their best in a murky situation
376.
by barrell 3y ago
In my experience it’s not such a simple question. If you want to be able to speak in nuanced non-English and have it pick up on the intricacies, or have it respond in rich correct non-English, then it’s not the best model (Cohere recently r
377.
by barrell 3y ago
If you tap on the settings/cog icon on a video, there’s an item for “additional settled”, which then has another item for playback controls or something. Once you want to do anything else you have to exit everything and open it back up
378.
by barrell 3y ago
Yes! I mean, we won't supply the content for all 7,000 languages, but we can make the core learning process and core pipeline language agnostic, so you could learn any language. With the exception of sign languages of course, as those
379.
by barrell 3y ago
We support Tamil :) I don't think we have any voices added for Tamil, but it's possible that a Tamil speaker voice clone in Hindi could be sufficient. Would love to try!
380.
by barrell 3y ago
We're definitely going to get there! For what it's worth, nobody has been signed on yet without receiving a tour and demonstration of the product, and nor would I expect anyone to pay money without having seen the product. Sometim
381.
by barrell 3y ago
For what it's worth, we're cooking up a lot of fun stuff with vector embeddings along these lines. For now we've decided on getting the product stable and launched, although everything is set up to go in this direction.
382.
by barrell 3y ago
Heh yeah that was just brought to my attention. It's kind of always been the mission of the project, and what we've been building internally. We've been adding 20 languages a month since we started, and we're not plannin
383.
by barrell 3y ago
I mean, I'm sorry your experience wasn't great. But this is fantastic feedback, so thank you for that. We're working on an FAQ section, we'll make sure these are all answered there. If you were genuinely asking here, I&#
384.
by barrell 3y ago
My experience from learning from childrens books has not been very good. They tend to be quite whimsical, which is hard to follow, use quite old language, which is hard to follow, and use really child-appropriate language, which is hard to
385.
by barrell 3y ago
While we offer people the ability to do machine translations, there's absolutely no reliance on them. We actively advise against using machine translations when onboarding new users It's a good point about the "every language
386.
by barrell 3y ago
A recent example that comes to mind is I wanted to refresh my Italian by watching my favorite TV show in an Italian dub. Even though I can converse in Italian, and quite the show by heart, I couldn’t really follow the program. Loading it up
387.
by barrell 3y ago
To clarify, there’s nothing stopping you from learning Japanese with it. Almost all of our beta testers are polyglots, studying 5+ languages, most of them studying at least Japanese or Chinese. It’s just that the alignments don’t work well
388.
by barrell 3y ago
I know Anki uses a modified version of SM2 so I can't say for certain, but if I recall correctly they are still using a linear algorithm. Most forgetting curves are, well, curved. More logarithmic. Once you go too far away from the &#x
389.
by barrell 3y ago
When we built this demo we figured the official translations would be well translated, apparently they are not. Some people elsewhere in the comments are saying it's just a weird English sentence so it has a weird French sentence. My F
390.
by barrell 3y ago
Try https://phrasing.app/ The /playground link I posted to HN was meant to give HN something to click around on. The root index page is meant to be the primary marketing materials.

More ›