Artificial Intelligence Panel


#1

#2

I am assigned part-time to work at the US DoD Joint Artificial Intelligence Center (JAIC) so I have a high interest in AI for Solid.
I believe Solid will be a foundation for a user-centric ‘AI Assistant.’ Big Tech uses AI to provide assistance to the individual, by collecting data on us, which they also sell. One could reason Solid could do the same thing, in a private way. But it will be hard for a new app to ask the user to share personal data while keeping it private, so let’s discuss how.

I believe it is possible to anonymize data and also have strong user agreements, and then data (such as location, driving routes, purchase decisions, etc.) could be shared. But this will be hard to implement and might require some new standards. I’m working on standards for JAIC, so I am interested in any new ideas for new standards relating to Solid or AI, which I may be able to get the resources to develop. Let me know if anyone has any ideas on new standards. (I’m also involved in the ISO/IEC SC42 Artificial Intelligence standards committee.)

Jim Schoening


#3

@JimSchoening,

Perhaps you would be interested in joining this panel to develop proposals on this subject for the Solid specification?

I believe Tom is @tag42git on this channel (the other panellist)

Mitzi


#4

Hi @JimSchoening,

Yes please join the artificial intelligence panel.

I’m not an expert in AI, I just have an interest in it and its intersection with Solid. I made a related post at Virtual Assistants need a soul. I’m on github as sideshowtom

Tom


#5

Hi Jim,

Thank you for joining the artificial intelligence panel. I made a pull request for your review for the readme. It just includes some boilerplate and two placeholder projects: virtual assistants and neural networks, with the goals to identify and define standards needed for those. There’s no separate chat channel yet.

Tom


#6

Hi, GAN (Generative Adversarial Network) are pretty new & cool if you hadn’t come across them before.


I wasn’t worried about AI taking my job as a ui designer until I saw that lol.


#7

Thanks @jucole for that link about GANs. That looks really interesting. Please join the Artificial Intelligence panel if you can. At this point we are just collecting ideas for projects.


#8

hi @tag42git I don’t really know enough about ai to be any use on a panel; it would be interesting to look into ai for perhaps policing pods.


#9

Hi @jucole,

I’m definitely not an expert either. I think if we leave AI to the experts (and I’m not picking on experts, but they vary morally just like anybody else), that we will be in trouble, or bigger trouble than we already are. I hope you’ll consider joining. You won’t be required to do anything besides keep an eye on us :slight_smile:

Tom


#10

Hi again @jucole,

There is another Solid Panel you might consider joining, the Humans and Machines Panel. It doesnt have a repository yet but I think if there are people interested in joining then we could move it forward.

There is some overlap between the focus of it and another Solid panel. The focus of the Privacy and Individuals Rights Protection Panel is on “researching how the … specifications and implementations … can affect the privacy of individuals … can affect their rights … lead to any kind of harm to the individual …”, and the focus of the Humans and Machines Panel is on “establishing how to maintain borders between humans and machines (be they mechanical, electrical, digital or political)”.

These are similar, but I think there are a lot of cases where even though individual rights may not be restricted, the borders between humans and machines need careful attention from as many people as possible.

Take Wikipedia, for example. Probably nobody would consider it a human right, but it would be a very good thing to have it grow and maintain its integrity. It could be subject to influence from some kind of machine.

How Solid figures into the software of the future is not clear but its bound to influence it if not be the basis of it, so I think all this is kind of a big deal.

Tom


#11

Hi Tom (@tag42git) “humans and machines” is indeed a very interesting subject but unfortunately being slightly dyslexic means I would have less time to focus on my little social app that i’m trying to build in Solid.


#12

Hi guys, perhaps could be a source of inspiration or a tool to use , TensorFlow has a Neural Structured Learning part for learning from graph, structured data, knowledge graph .

Not tested yet but I’d be interested if someone has an idea how to do with one POD, with more PODs

https://www.tensorflow.org/neural_structured_learning


#13

Hi @Smag0,

Sounds really cool! It looks like their implementations are in either c++ or python, so to me this looks like something that maybe for performance (and privacy) reasons should be done on a (Solid) server. I think there is a javascript tensor flow library for the client but it probably doesn’t include this new stuff anyway.

So imho I think the first thing is to make a hypermedia driven API for the tensor flow library, maybe using the Hydra vocabulary. There are some people on the interoperability panel such as @elf-pavlik and @RubenVerborgh who have worked with Hydra and they may have advice on how to go about that.


#14

Hi again @Smag0,

I looked at your ‘Pod of the future’ in your Todo on your pod.

Once you have free text tagged with something like NLP Interchange Format then using SPARQL for NLP will open up a world of possibilities. A POS tagging example can be found at https://parts-of-speech.info .

The POS tagging has to be done with a trained neural network, something like … TensorFlow. An example is at https://github.com/mrahtz/tensorflow-pos-tagger.

Once you have the text tagged with part of speech tags, then the fun part of searching for meaning in it with SPARQL can begin. That will open up a world with your pod of the future app.