HackerTrans
TopNewTrendsCommentsPastAskShowJobs

wuschel

2,135 karmajoined 13 ปีที่แล้ว
Former scientist/entrepreneur, looking what to do next

hn (at) rtoip (dot) com

Submissions

IPO: Lime (SEC S-1 Form, Neutron Holdings)

sec.gov
2 points·by wuschel·2 เดือนที่ผ่านมา·0 comments

Microsoft Paper: LLMs Corrupt Your Documents When You Delegate (Arxiv.org)

arxiv.org
7 points·by wuschel·3 เดือนที่ผ่านมา·2 comments

Nvidia and Thinking Machines Lab draw multi-year chip deal

siliconrepublic.com
3 points·by wuschel·4 เดือนที่ผ่านมา·1 comments

The New Way to Build a Startup (YC Video, YouTube)

youtube.com
1 points·by wuschel·5 เดือนที่ผ่านมา·0 comments

comments

wuschel
·7 ชั่วโมงที่ผ่านมา·discuss
Side question to you, considering your occupation:

Could you please list a set of core papers (or other resources) that give a beginner an overview or even understand of the fundamental concepts and techniques with LLMs?

Thank you!
wuschel
·เมื่อวานซืน·discuss
Depends. Oral examninations will be more important as a test of skill now.

During my university time, all main exams were oral examinations: One professor, one protocol keeper, and the student. Duration: 1-2 hours.
wuschel
·4 วันที่ผ่านมา·discuss
I looked into hackable printers 10+ years ago when building a demonstrator prototype of a bioproduced ink dye. The key challenge challenges is that they are an integrated system between mechanical, electronical, and chemical (ink formulation, surface chemistry) domains.

I would believe most relevant patents are now expired and that we should be able to build a printer that at least gives sufficient quality prints. Given sufficient motivation and time investment, that is.
wuschel
·8 วันที่ผ่านมา·discuss
Hmm. I am not so sure about that statement. There are many models at play in the market, not one model. What mechanistic patterns are driving those models that are falsified by this paper and lead to the outcome of increasingly extractive/stagnating economies?
wuschel
·15 วันที่ผ่านมา·discuss
My understanding is that in addition to your comment and the development of a method to separate the training data for distributed learning, the latency/bandwidth of systems connected on the internet is a challenge, too. Information has to be sent around before and after any hypothetical number crunching.
wuschel
·16 วันที่ผ่านมา·discuss
Thank you for your comment. I enjoyed it a lot. Good food for thought.

Your analogon is a bit leaky abstraction in the sense that it misses out on the broad stastical nature of LLMs. However, I find it is a good way to illustrate the potential industrial transformation.

It is hard to say what the future will bring. The original AsK HN post is definitly an omen for things to come.
wuschel
·16 วันที่ผ่านมา·discuss
Seconded. I applaud BI for doing such a clean open source release. So many things to learn from!

I wish other game studios would release their software so many years after release.
wuschel
·19 วันที่ผ่านมา·discuss
That is incredible, but I must ask: Beside really getting to know Rust's trait and type system, and besides experimental / educational tinkering, could this have any practical use case?
wuschel
·22 วันที่ผ่านมา·discuss
For those not in SF, could you please clarify what you mean by "labs"?
wuschel
·26 วันที่ผ่านมา·discuss
I would like to know whether someone was able to use lower tier models for activities other than coding e.g. a limited version of a personal note manager - and what the hardware requirements in RAM for these models were.
wuschel
·เดือนที่แล้ว·discuss
> "He was 38. Shortly before his death, he had mailed a manuscript describing a new three-electrode semiconductor device to Physical Review. The paper was lost in the wartime Atlantic. Five years later, Shockley, ..."

I wish the article had a reference for that claim.

I remember from my childhood that my father told me that in the old soviet system, publications from were invented and dated back in order to demonstrate the superiority of their science. Both sides might have done it.

Now, a story from my father is not strong data point. But falsification of scientific theories, statistics and publications was a thing in the Soviet Union [1,2].

Then again, the guy might have really done it.

[1] https://communistcrimes.org/en/falsification-memory-history-... [2] https://history.stackexchange.com/questions/33071/how-often-...
wuschel
·เดือนที่แล้ว·discuss
Thermals are one among many really big challenges that require costly solutions.
wuschel
·เดือนที่แล้ว·discuss
Could you perhaps cite the core papers for LLMs beyond „Attention is all you need“?
wuschel
·เดือนที่แล้ว·discuss
Would you know anything about stable forks for rsync that are not affected by performance quality degradations?
wuschel
·2 เดือนที่ผ่านมา·discuss
Thank you for that. Refreshing! :)
wuschel
·2 เดือนที่ผ่านมา·discuss
Yes, it is injectable, which is a unique property a material that exhibits some properties of steel.

The downside is that it is brittle.
wuschel
·2 เดือนที่ผ่านมา·discuss
Yes, of course people would be shocked.

The problem is just the data, the model, and the biomechanics solutions (…)

For now, I will put my trust in my human dentist.
wuschel
·2 เดือนที่ผ่านมา·discuss
I guess other languages that follow a stricter/larger compile time regimen e.g. Haskell, OCaml, Idris, ATS, F* would be suitable as well for this kind of approach. Like Go, OCaml seems to offer low compilation times.
wuschel
·2 เดือนที่ผ่านมา·discuss
Unfortunately you can't record meetings in many jurisdictions, including court sessions. Hence we have to rely - for worse, or perhaps even for better - on human driven note taking.
wuschel
·2 เดือนที่ผ่านมา·discuss
It is a small model, so what utility can I / Google expect from it? What is the on-board model used for?