LEVAN NELL NEIL and WordNet / ImageNet why aren't they working together? Can they be integrated with Iris?

Fynder Enlil

unread,

Apr 28, 2017, 9:17:19 AM4/28/17

to NELL: Never-Ending Language Learner

I have found a very interesting Artificial Intelligence project called Iris. ( https://iris.ai/ )

I think it would be great to integrate Iris.ai with Wolfram Alpha.

There are four other projects

http://levan.cs.washington.edu/ LEVAN

http://rtw.ml.cmu.edu/rtw/ NELL

http://www.neil-kb.com/ NEIL

and

http://www.image-net.org/ ImageNet

I think it would be great to see these coming together. Algorithmia also has facial recognition and OCR algorithms. A commercial product,

http://www.affectiva.com/ Affectiva has emotional intelligence, it knows what you are feeling.

Bring these products together with GoConqr and you have a formidable AI toolkit for Education. Especially if you have to assemble it yourself.

I am a student at Charles Darwin University and we have no Artificial Intelligence department here, we are a small regional campus.

So I was wondering is NELL compatable with these other projects? Can NEIL mine images for words that NELL finds and put the results on ImageNet?

Bryan Kisiel

unread,

Apr 28, 2017, 12:07:14 PM4/28/17

to NELL: Never-Ending Language Learner

Hi Fynder,

Of these, the NELL team has thusfar only explored a connection with NEIL.
Unofortunately, after some initial trials a few years back, the initiative
got left on the back burner. On the NELL side, the only outwardly visible
result is a fraction of the entities browsable on our website link to
their equivalent entries on the NEIL website. We do, however, have a
standing intention to add more "modalities" of learning to NELL, with
image learning being an obvious option that one would expect to provide a
significant amount of knowledge not ordinarily stated simply in text.

In the next year or so we expect to devote some concerted effort toward
integrating or linking in some fashion with a few other similar projects.
There aren't any detailed plans as yet, but thank you for the assortment
of suggestions. It's not yet clear how extravagant an effort it might be,
but reviewing these should be informative as we come into the planning
stages.

bki...@cs.cmu.edu

> --
> You received this message because you are subscribed to the Google Groups "NELL: Never-Ending Language Learner" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to cmunell+u...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

Fynder Enlil

unread,

Apr 29, 2017, 2:09:21 PM4/29/17

to NELL: Never-Ending Language Learner

Thanks for the Reply.

Another project worth inclusion would have to be

PlaNet

Under a corporate affiliation, the Search engine companies that have image search would produce a massive database.

Google Image guesses the identity of the object being searched.

Then there is this suggestion

http://spectrum.ieee.org/automaton/robotics/robotics-software/its-time-to-start-3d-scanning-the-world

A valid point that If we really want to interface with AI, It really does need to understand three dimensions.

It is like a Wikiscanapedia sort of project. I hope such would have more success than sensorpedia. A database of sensor data, indeed the ability to intergrate projects is a next big step for the field of AI.

There are some similar projects in robotics as well,

RoboBrain

https://arxiv.org/pdf/1412.0691.pdf

and

RoboEarth

http://roboearth.ethz.ch/

Activities in this areas seem to spin off well.

Both fields of Robotics and Artificial intelligence would benefit from a shared database of Locations and objects with three dimensions models.

I am surprised there is not more commercial interest in funding these projects, I could see google opening up its image search to any AI research, there are a few companies that love publishing their data online.

Even Uber is buying up AI companies ( http://geometric.ai/ )

That is why I was particularly interested in Iris.ai the field of education doesn't have a lot to invest in projects aimed at improving education.

I am also interested in the Rise of Academic Social Media. I think It would be great to see A central AI development for education and research into all fields that is publically accessible.

Integration seems to be the trend. Linkability. You make an AI with a visual database of the Worlds museums, three di catalog with links to all research into the period to recreations and simulations you may even include dramatizations and representations.

Visual familiarity with the Living organisms and associated sound is another area that could use an infusion of AI. Machine and Deep learning is exploding in Health. With companies like Enlitic using Deep Learning.

To what end?

For me, it is Artificial Consciousness:

Aleksander and Dunmall, 2003 ; Aleksander, 2005 have developed an approach to machine consciousness based around five axioms, which they believe are minimally necessary for consciousness:

1.Depiction. The system has perceptual states that ‘represent’ elements of the world and their location.
2. Imagination. The system can recall parts of the world or create sensations that are like parts of the world.
3. Attention. The system is capable of selecting which parts of the world to depict or imagine.
4. Planning. The system has control over sequences of states to plan actions.
5. Emotion. The system has affective states that evaluate planned actions and determine the ensuing action."

Source

Article Progress in machine consciousness

Lecun talks about the need for AI to Predict, to possess common sense and a world model, a simulation to base a prediction off.

I think Wolfram Alpha would have the Mathematical constants and Laws of Physics for such a world model, again, integration with AI projects is the next step.

Then there is ConceptMap 5.5

https://arxiv.org/abs/1612.03975

Thanks again for responding.