Fucking classic startups claiming they can do mad shit with deep learning thats impossible
What an embarrassment "algorithm from mit and columbia"
>>9096462
It's not impossible, it's just really REALLY hard...
>>9096465
Nah...show me a deep vqa system that can do query understanding at that level and do that kind of multimodal relational reasoning.
Anon u jelly?
https://arxiv.org/abs/1506.05869
Read this paper. They are just implementing this I think.
>>9096489
That paper applies encoder decoder for conversation. The image above implies relational reasoning and qa understanding.
>>9096464
>that picture
cringe, do /sci/ users actually interact with other humans like that?
>>9096508
>that post
*cringe* do /sci/ users actually interact with other humans like that?
lol OP that gif is pure cringe.
>>9096472
It does not have to do all of that when it's able to provide a sufficient answer consistently, by other means.
>>9096510
lichrally shaeking
the point of startups isnt to do anything it's to get sold because you convinced someone else you could do something
>>9096462
I bet they hired thousands of pajeets to answer those queries so they can exit scam with investors gibs
>>9097727
I doubt even the likes of Google, Facebook, (NSA?) have the ability to use AI that can draw connections between [mathematical concept: proportionality] and [symbolic form] and [resemblance to fish]. Now that I think about it, how would it differentiate between actually looking like a fish, and just bearing resemblance to the fish symbol?
>>9098970
Furthermore, who says that the symbol even looks like a fish? What if people think that it looks like an insect's head?
>>9098970
It's basic fucking image recognition, you retard nigger.
>>9099036
The dialogue about the proportionality symbol has nothing to do with image recognition.