Gary Marcus

I have expressed my position on this many, many times (including in the recent book "Architects of Intelligence"). But you still seem misunderstand it every time and to insist that we disagree, when we don't actually disagree that much. I'm tired of wasting my time.... 1/2.

... but here we go again 1. Whatever we do, DL is part of the solution 2. Hence reasoning will need to be compatible with DL 3. That means using vectors instead of symbols, and differentiable functions instead of logic. 4. This will require new architectural concepts. 2/2

Ok, now we are getting somewhere; in some of your previous tweets you seemed to ok with symbols and I didn’t understand how to square that with the rest of your view. (1/2)

Agreed on 1 & 2. But I see a false dichotomy in 3. An ASCII encoding is a vector that serves as symbol, as is an embedding if it used as value to bind a variable to. Systems like @LittleBimble & @egrefen’s do that. [continues]

And such systems (see also @_rockt, https://arxiv.org/pdf/1511.06279.pdf … etc) include what I think is essential: operations over those variables. That’s what I am fighting for, as explained in recent Medium post. You seem to exclude that.

Except that you are not actually fighting for it. You are merely inventing non-existing enemies and then claiming that others are fighting them on your behalf. You are not helping.

Can we organise a proper live-streamed debate? These endless Twitter rants are too hard to keep up with!

I'm all for doing this on Facebook ; -)