TLDR AI 2024-09-11
Apple A18 chips π±, AMD unified GPU architecture π, GPT-fabricated scientific papers π
Reach over 5 million tech professionals with TLDR (Sponsor)
Get your product featured right here in TLDR! Reach software developers, AI/ML engineers, executives and other tech professionals reading TLDR every day. TLDR offers 10 interest-based newsletters to get you directly in front of your target audience.
Learn more about running your first campaign with us.
AMD announces unified UDNA GPU architecture (7 minute read)
AMD announced at IFA 2024 that it will unify its RDNA and CDNA architectures into a combined UDNA microarchitecture, aiming to better compete with Nvidia's CUDA ecosystem. This strategic move seeks to streamline development and bolster AMD's position in AI and HPC markets. The transition to UDNA is a pivotal step, with full-scale implementation expected beyond the upcoming RDNA 4 generation.
GPT-fabricated scientific papers on Google Scholar (20 minute read)
Generative AI tools like ChatGPT are increasingly producing fraudulent research papers, infiltrating databases like Google Scholar alongside legitimate studies. These papers, often on controversial topics like health and the environment, pose significant risks to scientific integrity and public trust. Enhanced vigilance and more robust filtering in academic search engines are essential to curb this growing issue.
Apple announces its new A18 and A18 Pro iPhone chips (3 minute read)
Apple unveiled the A18 and A18 Pro chips at its βGlowtimeβ event, promising significant CPU and GPU improvements over the A16 Bionic. The A18 Pro boasts enhanced memory bandwidth and upgraded image-processing capabilities. Both chips support advanced AI features, with the A18 Pro notably improving on-device model performance and thermal design for better gaming experiences.
π§
Research & Innovation
Can LLMs Generate Novel Research Ideas? (28 minute read)
This longitudinal study assessed 100 NLP researchers' abilities to write novel ideas and then review those ideas. It found that LLMs can generate more novel ideas, but these ideas were slightly less feasible than the ones produced by human research scientists.
Superhuman Automated Forecasting (21 minute read)
The Safe AI Institute has released research on a system that can outperform human experts at forecasting.
AdEMAMix Optimizer (32 minute read)
This paper from Apple proposes an alternative to traditional exponential moving average optimization that uses contributions from older gradients to dramatically improve learning convergence.
Concept Sliders (16 minute read)
Concept Sliders are a powerful way to control the output of diffusion models. Work has been done to make them work with the new Flux suite of models.
Minifying HTML for GPT-4o: Remove all the HTML Tags (6 minute read)
Stripping HTML to plain text reduces costs with minimal performance loss in GPT-4o for data extraction tasks. Tests on the Mercury Prize dataset showed that GPT-4o performs well without HTML structure and that GPT-4o mini is a cost-effective choice for unstructured questions. For structured extraction tasks, consider testing both versions to balance cost and accuracy.
Waymo Giving 100,000 Robotaxi Rides Per Week But Not Making Any Money (3 minute read)
Waymo is providing over 100,000 paid autonomous rides weekly in San Francisco, Phoenix, and Los Angeles, doubling its May numbers. Despite its growth, Waymo remains unprofitable, with Google's experimental division incurring a $2 billion operating loss. High costs for vehicles and city mapping, combined with public hesitancy, continue to challenge its path to profitability.
Get the most interesting AI stories and breakthroughs delivered in a free daily email.
Join 920,000 readers for
one daily email