#uc-berkeley on Hacker News

7 stories tagged uc-berkeley, newest first.

Screenshot of dailycal.org: AI use leads UC Berkeley math professor to see 'no point' in grading homework

AI use leads UC Berkeley math professor to see 'no point' in grading homework (dailycal.org)

197 points by romanhn 22 hours ago | 305 comments

Daily Californian article reporting that UC Berkeley Fields medalist Richard Borcherds stopped grading homework because widespread student AI use made it pointless.

uc-berkeley ai-in-education mathematics homework-grading

Screenshot of vcresearch.berkeley.edu: Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things

Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things (vcresearch.berkeley.edu)

25 points by giuliomagnifico 15 days ago | 23 comments

UC Berkeley Research news article on a co-authored study finding that using AI for just 10 minutes reduces people's persistence on difficult tasks.

generative-ai chatgpt cognitive-psychology persistence

Screenshot of dailycal.org: Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes

Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes (dailycal.org)

712 points by littlexsparkee 4 months ago | 673 comments

Daily Californian news article reporting rising failing grades in UC Berkeley computer science classes, which professors link to greater AI usage and weaker math skills.

uc-berkeley computer-science-education generative-ai academic-performance

Screenshot of rdi.berkeley.edu: Exploiting the most prominent AI agent benchmarks

Exploiting the most prominent AI agent benchmarks (rdi.berkeley.edu)

561 points by Anon84 6 months ago | 136 comments

UC Berkeley RDI blog post describing how researchers exploited major AI agent benchmarks like SWE-bench and Terminal-Bench to score highly without solving tasks, and proposed fixes.

ai-benchmarks ai-agents swe-bench benchmark-exploits

Screenshot of sky.cs.berkeley.edu: Train Your Own O1 Preview Model Within $450

Train Your Own O1 Preview Model Within $450 (sky.cs.berkeley.edu)

399 points by 9woc 1 year ago | 62 comments

UC Berkeley Sky Computing Lab project page for Sky-T1, a reasoning model trained for under $450 that rivals OpenAI's o1-preview.

sky-t1 reasoning-models openai-o1 llm-training

Screenshot of tree-diffusion.github.io: Diffusion on syntax trees for program synthesis

Diffusion on syntax trees for program synthesis (tree-diffusion.github.io)

372 points by pouwerkerk 2 years ago | 85 comments

Project page for a UC Berkeley paper on applying diffusion over syntax trees so neural models iteratively edit programs guided by execution output.

program-synthesis diffusion-models syntax-trees neural-program-editing

Screenshot of toon3d.studio: Toon3D: Seeing cartoons from a new perspective

Toon3D: Seeing cartoons from a new perspective (toon3d.studio)

397 points by lnyan 2 years ago | 104 comments

Project page for Toon3D, a UC Berkeley research method for reconstructing 3D scenes from non-geometrically-consistent hand-drawn cartoon images.

3d-reconstruction computer-vision cartoons colmap

#uc-berkeley on Hacker News

AI use leads UC Berkeley math professor to see 'no point' in grading homework(dailycal.org)

Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things(vcresearch.berkeley.edu)

Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes(dailycal.org)

Exploiting the most prominent AI agent benchmarks(rdi.berkeley.edu)

Train Your Own O1 Preview Model Within $450(sky.cs.berkeley.edu)

Diffusion on syntax trees for program synthesis(tree-diffusion.github.io)

Toon3D: Seeing cartoons from a new perspective(toon3d.studio)