yiyingzhang

5 ms·

1.
by yiyingzhang 2mo ago
This approach only works for small context requests. For large context and relatively smaller output (say understanding a huge code base), the cost will mainly be on prefill, and sending the large context to multiple models will only increa
2.
by yiyingzhang 2mo ago
With AI, the speed of finishing a project and publishing a paper is very much accelerated. But you are right that the industry role in the program diverges from the original goal of independent academic research. But my guess is that there
3.
by yiyingzhang 2mo ago
What you described sounds like this: https://github.com/GenseeAI/gensee-crate . It offers whole workspace fork and merge. But the merge is still somewhat cumbersome and require manual overseeing.
4.
by yiyingzhang 2mo ago
Many AI startups open source part of their work and write blog posts without publishing papers, including the startup I work for. Sometimes, startups just have for more pressing things to do and don't have time to write up scientific p
5.
NSF partners with universities and industry on pilot 4-year PhD programs(nsf.gov)
3 points by yiyingzhang 2mo ago|3 comments
6.
by yiyingzhang 2mo ago
The U.S. National Science Foundation announced a $47 million investment over ﬁve years, along with additional ﬁnancial commitments from nearly three dozen universities and private industry, to launch a pilot of a new four-year Ph.D. model t
7.
by yiyingzhang 2mo ago
Cool idea! How do you guarantee privacy?
8.
by yiyingzhang 2mo ago
What stunned me is actually OpenAI's response to the incident. They team up with Hugging Face to put in more guardrails. Is this really the answer? Self-evolving AI can always find a way out, given more time, in a sense.
9.
by yiyingzhang 3mo ago
> My list is like 200 items now Do human developers check for 200 items when they do code review? How long would that take? It's quite clear that AI code review could be better even with some error. What's easy for human to rev
10.
by yiyingzhang 3mo ago
For academia people who care about publication, arxiv is more like a place to claim a spot before others do and before a paper gets accepted somewhere. It's also easier to get open-access publications from arxiv, as some journals are s
11.
Meta building cloud business to sell excess AI capacity(reuters.com)
2 points by yiyingzhang 3mo ago|0 comments
12.
by yiyingzhang 3mo ago
Very good point! Accountability is very hard if the ecosystem is monopolized by a few giants.
13.
by yiyingzhang 3mo ago
The uncomfortable part is not that Anthropic wants to detect resellers or distillation pipelines. That is normal adversarial business. The uncomfortable part is that a “safety” company put a covert classification channel into the system pro
14.
by yiyingzhang 3mo ago
Unfortunately, most students today just want to find the easiest way to get a good grade. The percentage of students truly want to learn is very low. For the most, they'd prefer instructors who feed them with exam problems. This is ver
15.
by yiyingzhang 3mo ago
We've been discussing this in our department for some time. No one gets fired yet :)
16.
by yiyingzhang 3mo ago
As a university professor, I honestly don't understand the point of grading. Who will look at and care about grades? Likely company HR. But then why should we (professors) do the screening for companies for free? Also, grades have long
17.
by yiyingzhang 3mo ago
Curious what's the typical switching frequency in your experiments and experience. How do you control the tradeoff of cache matching and model efficiency?
18.
by yiyingzhang 3mo ago
It all depends on the context window size. A small context size with fast performance won't be very useful today, as most workloads (like requests behind codex) usually have very long context.
19.
by yiyingzhang 3mo ago
How's this different from Firecracker?
20.
by yiyingzhang 3mo ago
This is cool! I may introduce it to the undergrad OS course I teach at UCSD. Does it have memory hierarchy?
21.
by yiyingzhang 3mo ago
This a a step away from university-driven research towards giving money to industry. FWIW, NSF's prior industry investment has rarely yield any true impact. Most successful startups spun from universities directly go for VC fund, esp.
22.
by yiyingzhang 3mo ago
Isn't this true since the beginning of software development? AI hasn't changed that yet
23.
by yiyingzhang 3mo ago
"Safety evals are an exception I believe eval startups can work when they're targeting safety benchmarks specifically. Researchers who want to work on safety evals tend to be ideologically opposed to working on capabilities, which
24.
Decoupling Compute and Memory for Async GPUs
8 points by yiyingzhang 3mo ago|2 comments
25.
by yiyingzhang 3mo ago
This is another Cerebras? fwiw, it took Cerebras many years to finally get a handle on the yield and the cooling problem. Wondering if they just hired a bunch of people from Cerebras.
26.
Show HN: GPT-5 available for free on Gensee(gensee.ai)
6 points by yiyingzhang 1y ago|0 comments
27.
Show HN: Free access and one-click swap to GPT-OSS and Claude-Opus-4.1 on Gensee(gensee.ai)
3 points by yiyingzhang 1y ago|0 comments
28.
Show HN: Optimize and launch a travel-planning AI application in minutes(gensee.ai)
4 points by yiyingzhang 1y ago|0 comments
29.
Show HN: Gensee – Free AI Agent Optimization and Deployment(platform.gensee.ai)
13 points by yiyingzhang 1y ago|0 comments