tkgally

30 ms·

421.
by tkgally 2y ago
I currently use ChatGPT 4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro for various personal and work-related tasks, and I often compare them on the same task. It’s really hard to decide which performs best overall for me. Sometimes one will be o
422.
by tkgally 2y ago
I agree with that point, too. My guess is that current models are not yet able to provide consistent, effective motivation to learners over the long term. Like a lot of new educational technology, they might be fun and engaging in the short
423.
by tkgally 2y ago
That capability probably exists now if you are willing to accept cloud-based models and only moderately-sized contexts. With Claude 3.5 Pro, for example, one can put one’s reference data into a Project and query the model with that data in
424.
by tkgally 2y ago
Are separately trained models necessary for your case? As context windows get longer—Gemini 1.5 Pro now accepts up to two million tokens, and Google has talked of the goal of "infinite" context windows—couldn't a single base
425.
by tkgally 2y ago
Most commuter train lines in Japan are part of integrated conglomerates of complementary businesses. JR East, which runs the Yamanote Line, also operates shopping malls, hotels, gyms, and other retail facilities at and near their stations.
426.
by tkgally 2y ago
> There's no Irrawaddy for ... North Korea. Rimjin-Gang [1] publishes news from North Korea in Japanese, English, and Korean. Their articles are based on reports from people inside North Korea using smuggled cellphones and focus on
427.
by tkgally 2y ago
While more languages are great, compared to translation by the better LLMs Google Translate leaves much to be desired. Accurate and appropriate translation requires sensitivity to the context and purpose of the translation. While it is poss
428.
by tkgally 2y ago
No, I think you’re right. It’s just a matter of convenient tooling. I have been doing similar tasks with ChatGPT and Gemini, but, in the standard web interfaces, I end up uploading the same reference files multiple times and I tend to lose
429.
by tkgally 2y ago
This article was written before Anthropic added the Projects feature for Pro and Team users [1]. I spent some time yesterday experimenting with Projects, and, like Artifacts, it looks really useful. I like the idea of being able to have mul
430.
by tkgally 2y ago
One or two of my junior-high-school English teachers—public school, California, around 1970—showed us how to diagram sentences. I don’t remember diagramming (note the spelling) being pushed hard at us, but it did help me start to understand
431.
by tkgally 2y ago
I've also had good results with Gemini 1.5 Pro for some tasks. Just yesterday, it produced very good analysis and comments based on a 200-page document. ChatGPT 4o was much weaker, and the document was too large for Claude 3 Opus. (Thi
432.
by tkgally 2y ago
ElevenLabs just released something that is more controllable: https://news.ycombinator.com/item?id=40736536
433.
ElevenLabs Voiceover Studio [video](youtube.com)
9 points by tkgally 2y ago|2 comments
434.
by tkgally 2y ago
This came up on my YouTube feed today and I was impressed. I have no connection to ElevenLabs.
435.
by tkgally 2y ago
I just copied that sentence at random to get some text to feed into Glish. Here’s the original story: https://www.bbc.com/news/articles/c1rr92dwqnyo
436.
by tkgally 2y ago
An online demo and a link to an informative video are here: https://paralogical.dev/glish/ Input: South Korea has said it will resume propaganda broadcasts against North Korea for the first time in six years in respons
437.
by tkgally 2y ago
Both HN itself and prolific HN contributor simonw get shoutouts in the article: “The term [‘slop’] has sprung up in 4chan, Hacker News and YouTube comments, where anonymous posters sometimes project their proficiency in complex subject matt
438.
by tkgally 2y ago
Meta comment: This back-and-forth between Nition and bredren is one of the best exchanges I’ve read on HN recently. Thanks to both of you.
439.
by tkgally 2y ago
> The biggest mistake with AI is making it appear human. I’m guessing that that mistake is already being made on a large scale intentionally, not by OpenAI or Anthropic, perhaps, but by companies developing virtual boyfriends and girlfri
440.
by tkgally 2y ago
For an opposing view, see Leopold Aschenbrenner’s recently released “Situational Awareness: The Decade Ahead”: https://situational-awareness.ai He did a very long interview with Dwarkesh Patel, too: https://www.youtub
441.
by tkgally 2y ago
The poem discussed has been translated many times in many different ways. There’s even a book, by Hiroaki Sato, that collects more than a hundred versions in English [1]. Here are a few: R. H. Blyth The old pond. A frog jumps in— Plop! C
442.
by tkgally 2y ago
Interesting point. I looked at the article again and noticed more signs of LLM composition: well-organized paragraphs of about equal length; a lack of specific examples, data, and quotations; and no errors or infelicities that would suggest
443.
by tkgally 2y ago
In general I agree with you, but sometimes the “live”ness of demos is essential. I’ve given quite a few presentations about LLMs over the past year and a half, often to audiences unfamiliar with or doubtful about them. Being able to have on
444.
by tkgally 2y ago
It’s down for me here in Japan (17:50 local time), too—web interface, Mac app, and iOS app.
445.
What is interpretability? [video](youtube.com)
1 points by tkgally 2y ago|0 comments
446.
by tkgally 2y ago
Or Gemini 1.5 Pro. The latest multimodal models, while still far from perfect, do seem to be getting better at image recognition and OCR.
447.
by tkgally 2y ago
Good question. It's probably too early to tell. By signing DORA, the university leadership has stated support for a principle about how hiring and promotion decisions should be made. But it may take a few years before that principle is
448.
by tkgally 2y ago
Last year, my employer, the University of Tokyo, became the first Japanese university to sign the Declaration on Research Assessment (DORA) [1]. Like open access, that initiative is a positive move toward a better research environment, and
449.
by tkgally 2y ago
That an interesting question! I had never seen that usage of “low-key” either. Some similar examples found on the web: “Astro Bot is low key one of the best Sony franchises and I’ll fight anyone who says otherwise.” “Nitro navy is low key o
450.
by tkgally 2y ago
The Kanda-Jimbocho area of Tokyo has over a hundred bookstores within a few blocks [1]. Aside from a few years during the pandemic, I've been buying books there regularly for four decades. The neighborhood has evolved over time, and so

More ›