I view systems alignment as a complementary and very neglected form of learning that really should be picked up by neuro, psych, and ML. Previous lab work finds adults spontaneously do it https://t.co/QOVJwPPlgq and the environment supports it https://t.co
1/ As @BDRoads and @ProfData pointed out, learning word meanings is pretty difficult! Luckily, we can leverage the co-occurrence structure of concepts, that is manifested in words and sensory data https://t.co/vsmE7ZvU37
There's a surprising twist to this thought experiment. The environment supports this type of learning (https://t.co/luqaVUULQW), people do it in lab unprompted (https://t.co/QOVJwPPlgq) and children's early concepts reflect learning by system alignment (
Learning as the unsupervised alignment of conceptual systems https://t.co/Qmbt6RLAJs
The reason large language models (LLMs) succeed is likely due in part to what @BDRoads finds below, namely the structure in one system (e.g., linguistic) is repeated across other systems (e.g., visual). Our world turns out to be structured kindly across sy
Previously, @BDRoads found systems (e.g., language, vision, etc.) in our natural environment support unsupervised alginment. Our new result shows people exploit this cross system info, even during supervised learning tasks when it could be ignored. 3/3 htt
Previously, https://t.co/c0tJ1UwNNg, we found there is substantial information in our environment to support this unsupervised alignment of conceptual systems. Our current contribution finds that people use this information even when it's not required to c
@JhendersonIMB They would! (assuming with scenes you mean the co-occurrence statistics of objects) https://t.co/bsFlmsCuiC work by Brett Roads & @ProfData Related work by @michaelfbonner & @epstein_lab https://t.co/gbizffRcXI
Open Access UCL Research: Learning as the Unsupervised Alignment of Conceptual Systems https://t.co/gybdfsU1or
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
RT @ProfData: EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https:/…
EU/UK students seeking an AI/Ai-neuro PhD, please contact me as there are spots for this Fall through this programme, https://t.co/GOjvBZGLtb To get an idea of what we do, visit https://t.co/bO6LYuayKZ or read some sample work, https://t.co/WTudVHYOoc , ht
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…
RT @MarcCoutanche: This is VERY nice - drawing on multiple conceptual systems helps unsupervised learning in a way not seen for supervised…
RT @MarcCoutanche: This is VERY nice - drawing on multiple conceptual systems helps unsupervised learning in a way not seen for supervised…
This is VERY nice - drawing on multiple conceptual systems helps unsupervised learning in a way not seen for supervised learning
RT @ProfData: "Learning as the Unsupervised Alignment of Conceptual Systems" now has a view-only version from the publisher (see below) in…
RT @ProfData: "Learning as the Unsupervised Alignment of Conceptual Systems" now has a view-only version from the publisher (see below) in…
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…
RT @ProfData: "Learning as the Unsupervised Alignment of Conceptual Systems" now has a view-only version from the publisher (see below) in…
"Learning as the Unsupervised Alignment of Conceptual Systems" now has a view-only version from the publisher (see below) in addition to https://t.co/DjU0p7jTwE It's also made it to sci-hub for those so inclined.
New paper w @ProfData, "Learning as the Unsupervised Alignment of Conceptual Systems" is now publicly available as a view-only version via the following SharedIt link: https://t.co/Bos8OOYr0f.
RT @ProfData: We've shown that supervised learning can be solved by purely unsupervised means. The big challenge now is to devise algorithm…
RT @PMinervini: I was searching for papers on concept formation in humans, and @o_guest suggested this amazing read: "Learning as the Unsup…
RT @PMinervini: I was searching for papers on concept formation in humans, and @o_guest suggested this amazing read: "Learning as the Unsup…
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…
RT @PMinervini: I was searching for papers on concept formation in humans, and @o_guest suggested this amazing read: "Learning as the Unsup…
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…
RT @PMinervini: I was searching for papers on concept formation in humans, and @o_guest suggested this amazing read: "Learning as the Unsup…
I was searching for papers on concept formation in humans, and @o_guest suggested this amazing read: "Learning as the Unsupervised Alignment of Conceptual Systems", from @BDRoads and @ProfData -- https://t.co/34whJL3qS0, https://t.co/PqyM5qonv4 https://t.c
Analogical reasoning, distributional semantics, and similarity-of-similarities (RSA / 3rd-order isomorphism) in the minds of machines.
Interesting! https://t.co/PKDqOLZV3b
RT @chriswolfvision: Wow, Nature links to the arxiv version of a nature paper on its own website: https://t.co/mRZg2A1SYV https://t.co/ygVZ…
RT @chriswolfvision: Wow, Nature links to the arxiv version of a nature paper on its own website: https://t.co/mRZg2A1SYV https://t.co/ygVZ…
Nice!
Wow, Nature links to the arxiv version of a nature paper on its own website: https://t.co/mRZg2A1SYV https://t.co/ygVZHR0YxR
RT @chrismbird: This is a great study. Proof of an important and potentially v useful principle (see thread for details). Should provoke a…
RT @john_omalia: For anyone considering how to make deep learning more versatile, read this and consider using more semantic / conceptual m…
For anyone considering how to make deep learning more versatile, read this and consider using more semantic / conceptual mappings in your learning systems! My spin - break with tradition and explicitly embrace human biases for improved disentanglement i
RT @santoroAI: Very nice paper! Multi-modal alignment allows for unsupervised discovery of concepts/labels/meaning. IMO these ideas will ev…