Meeting minutes
take up item 1
https://
Difficult to define unanimously, first definition by IEEE
1990 and creation of computational Intelligence (CI) society in 2003
take up item 2
3 pillars neural networks, fuzzy systems and evolutionary computations
take up item 3
Jim Bezdek subset of AI
relies on soft computing methods
nick Cercone 2022 web intelligence
Applications: web mining, semantic web, Natural Language processing
fake news identification, chatbot classification,...
take up item 4
Using Hugging face JS in your applications
transformers and more
Joshua's bridges gap between web dev and Machine learning
hugging face AI community
love open source (libraries)
JavaScriptification of AI
modules Inference (calls ), hubs users interact with hugging face hub, agents library
28 different tasks in inference speech recognition, translation for free
text to image,
create models and files to repositories,...
Agent JS you can ask the LLM draw me a picture, it will generate and run the code (stable fusion for cat)
transformers JS 6000 downloads (high use)
supported tasks: text classification, code completion, text-to-text
image classification, object detection, segmentation (text, vision, audio, multimodal)
speech-to-text (open AI model)
transcribe audio file whisper web, real time
text-to speech in the future
embeddings, ...
Applications: what can you do with tools: talk to a youtube video
browser extension
semantic image search engine
search using natural language across a DB
doodle dash
take up item 5
Protocol for CI as a service
use cases telerobotics, videogame AI,
useful for robotics , videogames,...
simulation training AI systems
evaluation of AI system
discussion sensors control signals
Time synchronization
remote procedure calls
robot to robot communication
extensibility (new types of sensor data..)
Smatnic (future opportunities)
Conclusion control robots and avatars, existing protocols can be extended ...
take up item 6
<Vagner_NIC_br> Adam's presentation link: https://
Standardization in AI/CI (ISO level)
Interoperability
to make models work together at syntactic and semantic level
Trustworthiness legal implication
EU governments require compliance
Way data are managed is important
consider legal implications
Internal mechanics
models spread over different components
how to make communication efficient
propagating learning experience over nodes
Metrics
Measuring the way we use these resources
competition between models
different ways to look at performance
Should be shared
should be accountable
sustainability
metrics should take energy demand into account
what is the impact carbon footprint use and training
CI in health
areas of interest and approach to the human way of thinking
BIp4Cast project
information from doctors, sensors devices,..
Collect info about a person on their mental health
indicators of depression , mania,...
modelling mental disorders
<Vagner_NIC_br> Victoria Lopez' presentation: https://
all the sensors we are using to get informartion help define a new treatment
updates between devices but difficulty to have data in communication,
solution raw data standardization
can data analysis predict a crisis ie bi polar disorder
problem of interoperability of devices
privacy question for devices
anonymisation is done previously with patient data
take up item 7
bi polar disorder: the doctor receives an alarm
based on euthymic state chart
when is the line going up ?
Doctor is alerted for consultation
prevention is key
computational intelligence as a service? How can I trust the results I get?
CI is to provide a means to get info quickly
CI models are testing the efficiency
it is a question of testing, posing the problem correctly
unpredictability of models however form of intelligence (human as well)