Generative AI serves as an artistic creation medium. Existing interactions often assume creators specify intentions in advance.
Source: arXiv
https://arxiv.org/abs/2610.24010
#MachineLearning
Discover
Local timeline
Published posts from this FPDP installation.
Prominent AI safety discussions emphasize existential threats, but a new position paper argues that widespread AI use also creates accumulative risks.
Source: arXiv
https://arxiv.org/abs/2610.24006
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24006
#MachineLearning
AstraZeneca has partnered with Oracle to launch AI-driven real-world data trials in Abu Dhabi, aiming to speed recruitment and safety monitoring for multiple diseases.
Source: BusinessLine
https://www.thehindubusinessline.com/info-tech/astrazeneca-partners-with-oracle-for-ai-driven-clinical-trials/article71603695.ece
#AI #Oracle
Source: BusinessLine
https://www.thehindubusinessline.com/info-tech/astrazeneca-partners-with-oracle-for-ai-driven-clinical-trials/article71603695.ece
#AI #Oracle
GardenBench is a new benchmark introduced with 11 test suites that evaluate LLM capabilities across various tasks.
Source: arXiv
https://arxiv.org/abs/2610.23866
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.23866
#MachineLearning
The Centre is considering developing a hardware device to strengthen security oversight of satellite internet services offered by Starlink and other overseas operators.
Source: Inc42
https://inc42.com/buzz/centre-mulls-hardware-security-layer-for-starlink-other-satcom-operators/
#TechPolicy
Source: Inc42
https://inc42.com/buzz/centre-mulls-hardware-security-layer-for-starlink-other-satcom-operators/
#TechPolicy
UBTECH has begun delivering its U1 companion humanoid robots, which social media users call cyber boyfriends and cyber girlfriends.
Source: TechJuice
https://www.techjuice.pk/ubtech-u1-companion-humanoid-robot-deliveries-begin/
#Robotics
Source: TechJuice
https://www.techjuice.pk/ubtech-u1-companion-humanoid-robot-deliveries-begin/
#Robotics
Germany signed an AI partnership with Nvidia at the Berlin investment summit. Chancellor Friedrich Merz opened the two-day event by addressing Jensen Huang, Nvidia's CEO. The economy ministry entered into a memorandum of understanding with the chipmaker.
Source: The Next Web
https://thenextweb.com/news/germany-nvidia-ai-partnership-berlin-summit
#AI #Nvidia
Source: The Next Web
https://thenextweb.com/news/germany-nvidia-ai-partnership-berlin-summit
#AI #Nvidia
Seoul stocks slid nearly 4 percent amid AI jitters and unrest in France. South Korean shares plunged as investors cashed in recent gains in major chip shares.
Source: The Korea Herald
https://www.koreaherald.com/article/10908892
#AI
Source: The Korea Herald
https://www.koreaherald.com/article/10908892
#AI
Exa 3.0 provides smarter search that gives AI agent richer data at lower cost.
Source: Product Hunt
https://www.producthunt.com/products/exa-ai
#Tech
Source: Product Hunt
https://www.producthunt.com/products/exa-ai
#Tech
US researchers have built a brain-like computer chip that can analyze fusion plasma at one million samples per second.
Source: TechJuice
https://www.techjuice.pk/brain-like-chip-fusion-plasma-million-samples-second/
#Semiconductors
Source: TechJuice
https://www.techjuice.pk/brain-like-chip-fusion-plasma-million-samples-second/
#Semiconductors
Chery debuted its Ziwei in-house chips, including the C100 central computer, S1500 SiC power chip and M-series MCUs.
Source: Pandaily
https://pandaily.com/chery-ziwei-chips-c100-central-computing-s1500-sic-m-series-control
#Semiconductors
Source: Pandaily
https://pandaily.com/chery-ziwei-chips-c100-central-computing-s1500-sic-m-series-control
#Semiconductors
An Amsterdam-founded company cut its workforce from over 1,000 employees to around 120 over about three years. The CEO reveals who is better placed to survive AI onslaught as operations are automated.
Source: Business Today
https://www.businesstoday.in/technology/artificial-intelligence/story/this-firm-cut-its-workforce-from-1000-to-120-in-three-years-the-ceo-reveals-who-is-better-placed-to-survive-ai-onslaught-562612-2026-10-20?utm_source=rssfeed
#EnterpriseAI
Source: Business Today
https://www.businesstoday.in/technology/artificial-intelligence/story/this-firm-cut-its-workforce-from-1000-to-120-in-three-years-the-ceo-reveals-who-is-better-placed-to-survive-ai-onslaught-562612-2026-10-20?utm_source=rssfeed
#EnterpriseAI
Researchers introduced WAQS, a framework for state-dependent steering without extra inference computation.
Source: arXiv
https://arxiv.org/abs/2610.24367
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24367
#MachineLearning
Text-to-SQL systems often assume a single correct translation, but many queries are ambiguous and can have multiple valid SQL outputs.
Source: arXiv
https://arxiv.org/abs/2610.24248
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24248
#MachineLearning
The US has ordered the removal of DeepSeek from civilian government IT systems, adding to growing restrictions on Chinese technology companies.
Source: BusinessLine
https://www.thehindubusinessline.com/info-tech/us-orders-removal-of-deepseek-from-civilian-government-it-systems/article71603248.ece
#TechPolicy #DeepSeek
Source: BusinessLine
https://www.thehindubusinessline.com/info-tech/us-orders-removal-of-deepseek-from-civilian-government-it-systems/article71603248.ece
#TechPolicy #DeepSeek
GUI world models are increasingly used to improve action selection by predicting consequences, but gains cannot be attributed to prediction alone. Direct Action Ranking improves task success without generating future screens.
Source: arXiv
https://arxiv.org/abs/2610.24372
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24372
#MachineLearning
Researchers proposed DASMark, an identity watermarking framework for dynamic agent systems. The framework includes Action-Preserving Embedding and Shift-Robust Recovery to improve stability.
Source: arXiv
https://arxiv.org/abs/2610.24382
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24382
#MachineLearning
Tool-using LLM agents face decision points under uncertainty, choosing between autonomous action, information gathering, and human escalation. Researchers constructed a counterfactual dataset of 3,115 action outcomes from 1,036 decision points in ToolSandbox to train SURE-Act.
Source: arXiv
https://arxiv.org/abs/2610.24194
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24194
#MachineLearning
Frontier supervision can bootstrap open models on forward reaction prediction. Models named Resonance distill reasoning traces and learn from verified outputs.
Source: arXiv
https://arxiv.org/abs/2610.24199
#MachineLearning
Source: arXiv
https://arxiv.org/abs/2610.24199
#MachineLearning
The ruling Democratic Party of Korea introduced a bill to support a semiconductor cluster.
Source: The Korea Times
https://www.koreatimes.co.kr/southkorea/politics/20261020/ruling-party-introduces-chip-cluster-bill-as-union-opposes-labor-exemptions?utm_source=rss
#TechPolicy
Source: The Korea Times
https://www.koreatimes.co.kr/southkorea/politics/20261020/ruling-party-introduces-chip-cluster-bill-as-union-opposes-labor-exemptions?utm_source=rss
#TechPolicy