HNHacker News
TopNewBestAskShowJobs

smy20011

415 karma · joined May 26, 2016

submissionscomments
smy20011··on Nvidia Announces Financial Results for Second Quarter Fiscal 2025
Numbers are impressive, not sure why it's down in AH.
smy20011··on Nvidia Announces Financial Results for Fourth Quarter and Fiscal 2024
Cisco in 2000: Gross margin in fiscal 2000 was 64.4%, compared with 65.0% in fiscal 1999. The following table shows the standard
smy20011··on [dead]
[flagged]
smy20011··on OpenAI Deal Lets Employees Sell Shares at $86B Valuation
It must be a big offer
smy20011··on An influx of Chinese cars is terrifying the West
Some cheapest BYD cars sells in China cost 14k USD. It's cheaper than a lot of EVs in US even considering tariffs.
smy20011··on Show HN: Inbox Zero – open-source email assistant
Thanks but I would like to have a localLM version. I don't want my email sent to any third party for analysis.
smy20011··on AMD MI300X 30% higher performance than Nvidia H100, even with optimized stack
I don't quite understand. Most of LLM inference is MatMul and Softmax. It's not rocket science to implement matrix multiplication in any GPU. We don't have very complex custom kernel for the LLM inference. If AMD can provide fast GPU with cheaper price, a lot of people don't mind to add support for them. GGML shows you can create a working LLM inference solution with one man's work.
smy20011··on LLM Visualization
Really cool!
smy20011··on Are Open-Source Large Language Models Catching Up?
I tried to use Sakura LLM to translate some JP novels. It's really good and half the price of GPT3.5 turbo.

https://github.com/SakuraLLM/Sakura-13B-Galgame/tree/dev_ser...

smy20011··on OpenAI: Increased errors across API and ChatGPT
You should at least have some basic ideas about your stack.
smy20011··on AIConfig – source control format for gen AI prompts, models and settings
Thank you. I think most things we store in the config are non-business logic. Like the connection string of database or some feature flags.

However, the prompt is your business logic in most cases and put your business logic into a separate file make it harder to read and harder to maintain.

smy20011··on AIConfig – source control format for gen AI prompts, models and settings
Why not store them in code? Easier to read and source controlled.
smy20011··on New models and developer products
Other than coding, do we have a good application of LLMs?
smy20011··on Mozilla.social mastodon waitlist
The UI looks like ELK (https://github.com/elk-zone/elk/)
smy20011··on We are wasting up to 20% of our time on computer problems, says study
Only 20%? I think the numbers are close to 80%.
smy20011··on Bluesky facing degraded performance due to record high traffic
Cannot see anything. Do I need to login?
smy20011··on Tell HN: Twitter switched temporarily to rate limited mode
People getting rate limited message cannot see that due to rate limit.
smy20011··on Tell HN: Twitter is now rate limiting accounts
It's fine that website down. How can it down 3 hrs? Where are the engineers?
smy20011··on [dead]
racism against white That's really rare. More like "I don't like lossing my privilege"
smy20011··on Generative AI is also a revolution for computer interfaces
I don't think so. Chat based interface is horrible on phone. You have to type what you want instead of tapping things.

Chatbot should be part of your ui instead of replacing the ui.

smy20011··on AMD Expands AI Product Lineup with GPU-Only Instinct Mi300X with 192GB Memory
One thing AMD can do is working with ggml to make llama.cpp running on AMD GPUs. Compile modern ML framework is a quite complex due to number of OPs. However, running LLMs does not need a lot of OPs. Just einsum & relu & softmax. Having LLM works with llama.cpp could be done by a team within a week or so.
smy20011··on Apple, Goldman Sachs Debut Savings Account with 4.15% Annual Yield
Why not just buying bonds?
smy20011··on Show HN: GPT-4-powered web searches for developers
My test query is "why LOLIN S3's LED is not working". All LLM failed on that query, this included.
smy20011··on OpenAI's ChatGPT plugins may revolutionise the traditional customer journey
We build web browser for a reason. Talk to a chatbot like talk to customer service. It's far slower and sometimes frustration.
smy20011··on ChatGPT Plugins
It seems that OAI have their preference of choosing the first movers of their ecosystem.
smy20011··on Bing Image Creator
But you can already do it for a long time.
smy20011··on Show HN: Alpaca.cpp – Run an Instruction-Tuned Chat-Style LLM on a MacBook
Alpaca model leaked again?
smy20011··on Large language models are having their Stable Diffusion moment
Maybe out-dated by 1 day LOL.
smy20011··on GPT-4 is coming next week – and it will be multimodal, says Microsoft Germany
2 AI winter so far!
smy20011··on Show HN: ChatGPT+ for $5/month on iPhone/iPad/Mac(1M words)
Dude, 1M words cost $2. You charge $3 for what?
← PreviousPage 3 of 4Next →