up6w6

crazy how they only show benchmark results against their own models

I am very suspicious of the results. A few months ago they published a LLM benchmark, calling it "perfect" while it actually contained like only 50 inputs (academic benchmark datasets usually contain tens of thousands of inputs).

Very audacious to call it "almost perfect" when it has only what appears to be 50 questions. For comparison, MMLU contains 57 tasks and more than 100k questions.

I remember that you can detect the "curl | bash" server side and serve a different script than what the user would get by downloading using other methods[1]. But yeah, the binary itself already has enough attack surface.

[1] https://news.ycombinator.com/item?id=34145799

Not only that, but Servo is only receiving less 3000 USD/month in donation, less than their goal of 10k/month, and much less than what they deserve if consider Ladybird is receiving millions.

https://servo.org/blog/2024/06/28/input-text-emoji-devtools/

Does anyone knows what is the current status of Apple silicon hardware emulation in qemu?

Related news: Servo Web Engine Continues Advancing But Seeing Just $1.6k In Monthly Donations

https://www.phoronix.com/news/Servo-Engine-May-2024

> Lack of speed and ghosting felt like it made traditional Eink impossible to do most computing tasks. So we focused on making the most Paperlike epaper display that has no ghosting and high refresh rate - 60 to 120fps. We started working on this in 2018.

The website mentions 60hz, will it also support 120hz?

The Opus model that seems to perform better than GPT4 is unfortunately much more expensive than the OpenAI model.

Pricing (input/output per million tokens):

GPT4-turbo: $10/$30

Claude 3 Opus: $15/$75

Chatgpt-3.5 price reduction seems to be a direct response to Mixtral, which was cheaper (~0.0019 vs 0.0020 for 1K tokens) and better (https://arena.lmsys.org/) until now.

Sadly we will still have to wait for application. Firefox doesn't have HDR support even on Windows right now.

I think the medium is trying to compete with Anthropic's Claude than Openai's products

https://www-files.anthropic.com/production/images/model_pric...

For shell commands, there is a tool from Github that does it quite well (and it's included on the copilot plan)[1].

It's perfect to run fast bash commands that I forgot but are simple enough that the first Google search result would solve it anyways.

[1] https://www.npmjs.com/package/@githubnext/github-copilot-cli

Love how they want to keep the focus to the personal usage instead of selling themselves into building features for business.

Does anyone knows what is the current state of the voice commands? I always wanted a self hosted version of Google Assistant. I bet it's possible to use a fine tuned version of llama and open source models like whisper to have a local assistant smarter than any of these smart speakers available in the market.

Check out https://github.com/seemoo-lab/openhaystack

True. The results from codex are actually from code-cushman-001 (Chen et al. 2021), which is an older model that Copilot was based on.

Even the 7B model of code llama seems to be competitive with Codex, the model behind copilot

https://ai.meta.com/blog/code-llama-large-language-model-cod...

> usually can have their TDP modified in BIOS, via Ryzen Master, or third party tools like RyzenAdj (Phoenix support: https://github.com/FlyGoat/RyzenAdj/pull/256) to perform pretty closely.

I didn't know about that. Do you have citations or benchmarks?

Usernames is a pending feature[1], something discussed in their "secret" non-indexed forum.

[1] https://community.signalusers.org/t/usernames-in-signal/9157