Agreed @tdietterich that touch is part of grounding. It’s a very precise and high fidelity grounding in a highly localized way (typically within square centimeters or at best square feet scale), whereas vision is a much more expansive grounding of the world that can extend very
RESEARCH
-
CSAIL Researchers Open AI Research Collaboration Opportunities
By
–
The CSAIL researchers welcome collaboration. Please contact @mitqinzy and @yaqubhai for more info, including resources.
-
CapSpeech: Guidelines for Expressive Text-to-Speech Systems
By
–
CapSpeech provides guidelines for data construction & model design to develop prompt-guided expressive text-to-speech systems across use cases like cinematic narration, game character voiceovers, conversational chatbots, emotion therapy, & more.
-
CapSpeech: AI Text-to-Speech with Controllable Voice Design
By
–
MIT CSAIL, JHU, & USC researchers have developed "CapSpeech," a text-to-speech framework that generates voices w/controllable timbre & speaking style via text prompts.
— MIT CSAIL (@MIT_CSAIL) 5 juin 2025
Imagine designing a voice just by describing it. CapSpeech can customize age, timbre, accent, emotion, & more:… pic.twitter.com/Dw0i6JUG19MIT CSAIL, JHU, & USC researchers have developed "CapSpeech," a text-to-speech framework that generates voices w/controllable timbre & speaking style via text prompts. Imagine designing a voice just by describing it. CapSpeech can customize age, timbre, accent, emotion, & more:
-
CapSpeech: Open-Source Pretraining Dataset and Models
By
–
CapSpeech is a fully open-source project for academic use, offering a large-scale pretraining dataset, high-quality fine-tuning subsets, & pretrained models.
-

Reference to recent paper on AI issues and problems
By
–
Reminded me of some of the issues highlighted in @joelbot3000
’s recent paper: https://
arxiv.org/abs/2501.13075 -
Qwen unveils new high-performance multilingual embedding standard
By
–
Qwen silently dropped the new standard for embeddings on the Hub!
– 0.6B, 4B and 8B versions (probably would use only the 0.6B)
– 32k context length – 100 languages – SOTA on MTEB, but like real SOTA, with 10 points margin on the second bests https://
x.com/tomaarsen/stat
/tomaarsen/status/1930579927020994694
… -

Increase human input ratio for valuable AI output
By
–
Just increase the ratio of human input to (valuable) ai output. @swyx
-
AI in Science: Insights, Debates, and Gaps
By
–
I tested this prompt on the topic: “The Role of AI in Scientific Discovery” LLMs returned: – 5+ sub-topics
– Top cited papers
– Major debates (e.g., human vs. machine creativity)
– Knowledge gaps for future work No fluff. Just academic-grade insight. -
LLMs and the Automation of Academic Research Tasks
By
–
Let’s break down what academic researchers actually do: 1. Identify gaps in knowledge
2. Collect, organize, and compare sources
3. Extract key arguments, methods, results
4. Synthesize findings into clear insights
5. Write with clarity and structure LLMs can now do all 5.