HNHacker News
TopNewBestAskShowJobs

qrios

227 karma · joined January 19, 2013

It is not a question if, but ‚when‘ we will see what we can call AGI. The ‚when‘ can be answered if we know the ‚how‘. The ‚how‘ we know if the ‚when‘ is answered by ‚today‘.
submissionscomments
qrios··on XAI lays off workers tasked with training Grok
https://archive.is/q56om
qrios··on Mistral raises 1.7B€, partners with ASML
Maybe it has to do with the "place and route" step in the synthesis of a chip design. Right now it is based mainly on random numbers and shuffling logic and wires. ML(not LLM) could help here in the same way as AlphaFold is using it for proteins.

[1] https://en.wikipedia.org/wiki/Place_and_route

qrios··on Mistral raises 1.7B€, partners with ASML
Thnx for the hint! I missed the news[1].

[1] https://euclyd.ai/#news

qrios··on America is in a serious jobs slump
> it's the uncertainty

Anecdotal reports from Europe (one insurance company, two banks, one systems integrator, one consulting firm; all multinational): All values in the risk assessment systems must be reassessed if an American partner is involved. For two companies, this must be done on a monthly basis.

A chancellor in Germany once proclaimed a ‘policy of steady hands’ when the opposition accused him of unsettling the economy with his many reforms.

qrios··on Show HN: Hacker News em dash user leaderboard pre-ChatGPT
Feature request 3: …
qrios··on Are OpenAI and Anthropic losing money on inference?
This is not my domain, but I assume the MMUs acting like a switch and something like multicast is not available here. I‘ve tried to implement such on a FPGA and it was extremely cost intensiv.
qrios··on Are OpenAI and Anthropic losing money on inference?
If multiple cores tries to get the same memory addresses, the MMU feeds only one core, the second one have to whait. Depends on the type of RAM, this will cost a lot of cycles.

GPU MMUs can handle multiple line in parallel. But not 10k cores at the same time. The HBM is not able to transfer 3.5TByte sequencial.

qrios··on Are OpenAI and Anthropic losing money on inference?
For sure an interesting calculation. Only one remark from someone with GPU metal experience:

> But compute becomes the bottleneck in certain scenarios. With long context sequences, attention computation scales quadratically with sequence length.

Even if the statement about quadratically scales is right, the bottleneck we are talking about is somewhere north by factor 1000. If 10k cores do only simple matrix operations each needs to have new data (up to 64k) available every 500 cycles (let's say). Getting these amount of data (without _any_ collision) means something like 100+GByte/s per core. Even 2+TByte/s on HBM means the bottleneck is the memory transfer rate, by something like 500 times. With collision, we talk about an additional factor like 5000 (last time I've done some tests with a 4090).

qrios··on Anthropic in Talks to Raise Up to $10B in New Funding
https://archive.is/2025.08.21-172524/https://www.bloomberg.c...
qrios··on Understanding Moravec's Paradox
Why the name „Moravec“ is two times correct in this article, but it is misspelled if it is a link text.
qrios··on Streaming services are driving viewers back to piracy
Right, I missed a whole decade. (For me 1989, the Mauerfall is the official end of GDR)
qrios··on Streaming services are driving viewers back to piracy
‘The Lives of Others’ is an outstanding film. However, it is a reappraisal of East German history and was made seven years after the collapse (the director grew up in West Germany and Western Europe).

US-America has looked at the subject of surveillance of its own population and its own (possible) collapse many times and often in a timely manner.

"The Conversation": https://en.wikipedia.org/wiki/The_Conversation

"Enemy of the State": https://en.wikipedia.org/wiki/Enemy_of_the_State_(film)

"The Siege": https://en.wikipedia.org/wiki/The_Siege

"In the Heat of the Night": https://en.wikipedia.org/wiki/In_the_Heat_of_the_Night_(film...

"Eagle Eye": https://en.wikipedia.org/wiki/Eagle_Eye

If you ask publicly, ‘What if I've seen XYZ?’ then it's actually already too late.

qrios··on Streaming services are driving viewers back to piracy
> I don't care if someone knows how many times I've watched Idiocracy.

I come from Germany, from East Germany. And some people there wanted to know if you had seen certain films and how often. And ‘Idiocracy’ would have been very high up on their list.

Not all films were banned right from the start (‘The Legend of Paul and Paula’ [1]), but right from the beginning the Stasi found it very interesting who had watched the film.

[1] https://en.wikipedia.org/wiki/The_Legend_of_Paul_and_Paula

qrios··on GPT-OSS-120B runs on just 8GB VRAM & 64GB+ system RAM
Some numbers are posted in the comments:

> … you can expect the speed to half when going from 4k to 16k long prompt …

> … it did slow down somewhat (from 25T/s to 18T/s) for very long context …

Depends on the hardware configuration (size of VRAM, speed of CPU and system RAM) and llama.cpp parameter settings, a bigger context prompt slows the T/s number significantly but not order of magnitudes.

Facit: gpt-oss 120B on a small GPU is not the proper setup for chat use cases.

qrios··on Tim Cook rallying Apple employees around AI efforts
Isn't "Private Compute Cloud" just a marketing term with some restrict sec architecture? The real personal assistant LLM would mean to have the realtime data available in hot memory (to make sure to give instant responses).

Audio, video, screen recordings, etc. from a single customer could be something between 1 and 10 GByte per day on average. After training you might get something like 3 MByte in additional model size per day. Even with 1 billion active users you would need to store additional data with 1 billion GByte (again on hot storage, like expensive GPU memory). The total amount of the memory of GPUs sold by NVIDIA is not even close to 400mio GByte (NVIDIA 3.8mio data center GPUs in 2023).

qrios··on Tim Cook rallying Apple employees around AI efforts
Current "good enough" models like Mistral Small require GPUs like the RTX 6000 to achieve user-friendly response times. The model quality is good enough, especially for narrow-scope tasks like summarization and classification. If Moore's Law holds for a few more years, a mobile device will be able to run it on-device in around 8 years (Apple's A11: 410 GFLOPS vs. RTX 6000: 16 TFLOPS [1]).

This is under the assumption that we don't see any significant optimization in the meantime. Looking back over the last eight years, the probability of no progress on the software side is near zero.

For a breakthrough in the consumer market, running LLM on-device with today's capabilities requires solving one key topic: "JIT learning" [2]. We can see some progress here [3, 4]. Perhaps the transformer architecture is not the best for this requirement, but it is hard to argue that it is impossible for Generative AI.

Due to today's technical limitations, we don't have real personal assistants. This could be the Mac for Apple in the AI era.

[1] https://gadgetversus.com/graphics-card/apple-a11-bionic-gpu-...

[2] Increasing context size is not a valid option for my scenario as it also increases the computation demand linear.

[2] https://arxiv.org/abs/2311.06668

[3] https://arxiv.org/abs/2305.18466

[Edit: decimal separator mess]

qrios··on US AI Action Plan
„Advance the Science of AI 9“?

Is this a reference to the AMD chip, or just a fragment of a removed numbered list?

Edit: It‘s a fragment of the PDF-to-HTML [1]

1: https://news.ycombinator.com/item?id=44661843

qrios··on Apple seriously considers buying Mistral:Mark Gurman
Source: https://archive.is/3iWxi (Bloomberg: "Tim Cook Isn’t Going Anywhere Soon, But an Apple Shake-Up Looms")
qrios··on LLM Inference Handbook
Thanks for putting this together! From now on I only need one link to point interested ones to learn.

Only one suggestion: On page "OpenAI-compatible API" it would be great to have also a simple example for the pure REST call instead of the need to import the OpenAI package.

qrios··on East German cities offer free stays to fight depopulation
„Deutsche Welle“ (dw) is a public funded station. It‘s like VOA of Germany, at least from the flow of money perspective.
qrios··on The German automotive industry wants to develop open-source software together
All three points are valid for every platform provider, and so for car manufactures.
qrios··on Lessons from That 1834 Landscape Gardening Guidebook
I remember "Halbgefrorenes"[1] as one of the main source of income for many restaurants and Cafés in the former DDR(GDR). Second recipe to generate nearly unlimited profit was Soljanka[2].

[1] https://de.wikipedia.org/wiki/Halbgefrorenes (german)

[2] https://en.wikipedia.org/wiki/Solyanka (english)

Edit: format

qrios··on Why it is (nearly) impossible that we live in a simulation
In my understanding the amount of energy in our universe is limited and this limitation is related to the very basic parameters of our universe. To escape this limitation you need to have an host universe with different fundamental constants. (chapter 4.5 in the paper)
qrios··on Why it is (nearly) impossible that we live in a simulation
There is a twist: author (Franco Vazza[1]) simulates universes:

> I study turbulence, shocks, magnetic fields and relativistic particles in the Universe using challenging large cosmological simulations.

[1] https://cosmosimfrazza.myfreesites.net

qrios··on Ask HN: Any good April Fools today?
Iconic Concorde back to the skies just in time for April[1]

[1] https://htxt.co.za/2025/04/iconic-concorde-back-to-the-skies...

qrios··on C. Elegans: The worm that no computer scientist can crack
Interesting article, but no reference to "DEVS"? Prediction of the worms behaviour is a key scene[1] in the opening episode.

[1] https://www.youtube.com/watch?v=xEfyT3vLD2c

qrios··on 'More than a hint' that dark energy isn't what astronomers thought
> Armed with the coolest business card in journalism, identifying me as the "cosmic affairs correspondent" of The New York Times …

The author Dennis Overbye about his retirement in December 2024 [1]

[1] https://archive.is/26oDb#selection-835.102-835.228

qrios··on Gödel's theorem debunks the most important AI myth – Roger Penrose [video]
The question is answered in the full title:

> "Gödel's theorem debunks the most important AI myth. AI will not be conscious"

Same statement from Penrose here with Lex Friedman: "Consciousness is Not a Computation" [1].

[1] https://www.youtube.com/watch?v=hXgqik6HXc0

qrios··on DualPipe: Bidirectional pipeline parallelism algorithm
In very simple words: it is one way to reduce the white squares in the picture from @danielhanchen[1].

In more complex words: imagine a processor which is able to process every instruction in 10 clock cycles. But also the processor is able to get new input for this instruction on every clock cycle and starts to process this new input in a pipeline. After the first input you have to wait ten clock cycles. But if you feed the input line every time you will get the output also permanently.

In the case of GPUs, it is now not only a topic of a single pipeline, but multiple in parallel. Depends on your data and algorithm it can be thousands in parallel.

[1] https://x.com/danielhanchen/status/1894937006352031832

qrios··on Show HN: Scripton – Python IDE with built-in realtime visualizations
Rerun is exactly what we have been looking for years. A game engine for the visualisation of multidimensional data without all the game 3D knowledge. I have so many questions!

(Can I access a mmap file, display it in real time with rerun and save a history at the same time?)

← PreviousPage 3 of 5Next →