What happens when you purchase a robot for household tasks, but the bot fails? MIT framework helps non-technical users learn why using counterfactual explanations, then fine-tunes an ML algorithm so the robot can perform the task correctly: http://
bit.ly/3Q019Iw
@mit_csail
-
MIT Framework Explains Robot Failures Using Counterfactual Explanations
By
–
-
Universities Need Large-Scale Research Cloud for ML
By
–
"We [must] empower universities with a large-scale research cloud to enable us to better study and understand ML." CSAIL Director Daniela Rus w/
@BostonGlobe
: https://
bit.ly/3OlqAD4 -
MIT Researchers Use LLMs for CAD Design and Manufacturing
By
–
From text to reality: MIT researchers find new ways to use LLMs to help in design & manufacturing.
— MIT CSAIL (@MIT_CSAIL) 28 juillet 2023
These models can convert text prompts to CAD, generate manufacturing instructions, and search for optimal designs: https://t.co/XKCbRMagZ2 pic.twitter.com/fL1PV1h9GdFrom text to reality: MIT researchers find new ways to use LLMs to help in design & manufacturing. These models can convert text prompts to CAD, generate manufacturing instructions, and search for optimal designs: https://
bit.ly/3Ydvq8T -
Essential Data Science Cheat-Sheet: ML, Probability & Statistics
By
–
This 10-page data science cheat-sheet covers machine learning, probability, statistics & more: http://
bit.ly/2mmrIMB (
@MaverickLin v/
@KirkDBorne
) -
MIT Hacker Discusses 54 AI Innovation Questions
By
–
Asking an MIT hacker 54 questions: https://
bit.ly/3q4unLL -
MIT Algorithm Protects Sensitive Data in Machine Learning Models
By
–
Sensitive data encoded within ML models can be extracted by malicious agents. MIT algorithm enables users to potentially add the smallest amount of noise possible to ensure that this data is protected: https://
bit.ly/3NKm3ss -
MIT Neural Network Cuts Robot Task Planning Time in Half
By
–
Imagine you've asked a robot to make you coffee. The difficulty: there are too many actions involved.
— MIT CSAIL (@MIT_CSAIL) 23 juillet 2023
MIT system cuts this planning time in half w/a neural net that predicts the probability that a task plan can be refined for more feasible robot motion: https://t.co/v6W1VcjT4I pic.twitter.com/M5ewUb8577Imagine you've asked a robot to make you coffee. The difficulty: there are too many actions involved. MIT system cuts this planning time in half w/a neural net that predicts the probability that a task plan can be refined for more feasible robot motion: https://
bit.ly/3PXobzH -
MIT Dataset Enables Precise Semantic Image Captions for Accessibility
By
–
Training ML models w/MIT’s new dataset empowers them to create precise, semantically dense captions, while illustrating data trends & intricate patterns. It could help improve accessibility for those w/visual disabilities: https://
bit.ly/3D6VwQZ -
MIT combines neural networks with classical simulation for robot control
By
–
MIT researchers combine classical numerical simulation with neural networks, generalizing large-scale multiphysics scenes and potentially closing sim-to-real gaps in robot manipulation & general control in complex physics: https://t.co/hEkGyMEPQY
— MIT CSAIL (@MIT_CSAIL) 21 juillet 2023
Code: https://t.co/sUfAvqSwJM pic.twitter.com/R2C5EZygnkMIT researchers combine classical numerical simulation with neural networks, generalizing large-scale multiphysics scenes and potentially closing sim-to-real gaps in robot manipulation & general control in complex physics: https://
bit.ly/3O6HfdB Code: https://
github.com/PingchuanMa/NC
Law
… -
MIT AI Tool Generates Novel Proteins Using Diffusion Models
By
–
MIT computational tool uses generative AI to craft previously unseen proteins by incorporating frames into diffusion models. Each frame aligns w/the properties of protein structures, enabling the tool to potentially enhance biomedicine & drug delivery: https://
bit.ly/3JWCHnH