I gave a lecture at @Stanford CS 25. Lecture video: https://t.co/0mexY32eLI AI is moving so fast that it's hard to keep up. Instead of spending all our energy catching up with the latest development, we should study the change itself. First step is to identify and understand

Tencent presents WebVoyager Building an End-to-End Web Agent with Large Multimodal Models paper page: https://t.co/GVodaTNHEl The advancement of large language models (LLMs) leads to a new era marked by the development of autonomous applications in the real world, which drives

Interesting finding about the positive effectiveness of anthropomorphizing LLMs, something I have been trying to avoid when prompting. Somehow parallels how natural language google queries turned from being pretty useless to giving better results than just typing keywords.

"...research will be a living compendium of code, datasets, graphs and narrative content — remixable and always up to date. Open and freely accessible research will be more valuable and influential because it will be seen by LLMs."

"If scientific research is mostly read by machines, the question arises of whether it is relevant to package it into a single coherent narrative that is adapted to the limitations of human cognition." https://t.co/vdeqqdt1Mv

13/13 Thanks to our reviewers and colleagues who provided feedback, and to @coadapt_project and @virtualtimes_eu.

12/13 Finally, BCIs provide an opportunity for the discipline of human-computer interaction to sharpen its own understanding of embodiment. Looking back, we note that external aspects of embodiment got much more attention than neural phenomena. It is time to expand beyond.

11/13 While embodied aspects open up new research avenues, they also increase the research burden in some ways. Researchers should more clearly justify BCI use, should scrutinize prior designations of 'data artifacts', try to accommodate natural body movements during experiments.

10/13 How to operationalize this 'structuring cognition' aspect? Researchers can design and evaluate BCIs with longitudinal skill building in mind. They can opt for open-ended tasks to observe qualitative differences in outcomes instead of observing performance alone.

9/13 A more abstract implication is the application of systemic thinking that defines body-brain integration to BCIs. For complex tasks, BCIs are better viewed as entities that structure cognition as opposed to neutral means of accessing brain content.

8/13 For example, future BCIs can be informed by existing body-brain semantic mappings to be motor-congruent. BCIs can, counter-intuitively, be used in synchron with physical devices that give task-relevant tactile feedback or help users assume certain body/hand postures.

7/13 Acknowledging this integration is crucial if BCIs are to widen in application scope and go beyond proof-of-concept. We identify various ways embodied cognition can provide guidance and lead to novel research directions for BCI research in human-computer interaction.

6/13 Yet we know that the role of human body goes beyond transmitting thoughts. Our bodies are deeply integrated with, and facilitate cognition in the brain. Gesturing during speech eases lexical recall, walking facilitates thinking, body postures support mental imagery...

5/13 This is precisely the issue that concerns embodiment. At first sight, it is tempting to consider BCI deployment as purely technical: "Only if we could read the mind perfectly and directly interact with the environment using brain signals, bypass our limited-bandwidth body".

4/13 BCI research for able-bodied, in contrast, needs to improve over existing muscle-operated devices (e.g., touchscreens, keyboard). How BCIs will compare with body-mediated interaction requires understanding their relationship.

3/13 Why now? We see an increased interest in deploying BCIs for the general population. This is a departure from the traditional BCI focus on users with severe physical disability for whom brain signals can be the only means for communication and thus easy to justify.

2/13 Main point: The recent wave of interest in brain-computer interfaces (BCIs) requires paying attention to a relatively neglected area within this domain, the extent cognition is grounded on the wider body, namely the question of embodiment.

Our paper "Revisiting Embodiment for Brain-Computer Interfaces" is out. https://t.co/5kGorpy68E preprint (open access): https://t.co/0XoHH1xAAr A thread on our main arguments and take-aways #BCI #HCI 1/13 🧵

her şeyin özeti

Language abilities != Thinking. Or why LLMs such as ChatGPT can eloquently spew complete nonsense. Their grasp of reality is very superficial. https://t.co/rT2XhJB72G This piece in the Atlantic comments on a paper by the MIT Cognitive Science crowd