Can we control the output of neural generation systems?
Can we build a better chatbot using this control?
And what makes a good conversation anyway?

Check out the blog post for our NAACL paper:

https://t.co/UAAUxSeA0L

#AI #NLProc #DeepLearning #ConversationalAI @facebookai

The Straddle is my favorite trade these days. But more often than not, one needs to adjust to make a good return from it. There are many ways to do this depending on your view, psychology and imagination. Will discuss a few that I use

The Straddle is a delta neutral strategy and it is important to keep deltas in check during the course of the Trade and not let them run away from you. Of course Volatility is the elephant in the room- but not going there as that will require a separate discussion

The first and most common method is to go 'inverted.' That is to say to shift the untested side to generate additional Premium and increase the runway to Breakeven. So if you start with a Straddle at 700 and price goes to 740, you can move the PE up to say 720.

How much should you move the untested side? While again this is a matter of skill, experience and artistry, the rule of thumb is to cut deltas by half. Remember, prices will move and you have to allow sufficient room for them to fluctuate

The second method is to shift the entire Straddle in the direction of the move. So if you have a 700 Straddle and Price goes to 740, you can shift the Straddle to 720. If you time this right, it can generally be done at virtually no cost or minimal cost.

1) FL absentees, early voting:

*Dems as expected ("souls to the polls Sunday") had a good day and FINALLY take the early walk-in vote lead by 6,100 votes.

Rs total: 1,151,593
Ds total: 1,092,547

Rs +59,046 combined lead. This is very good for this point in the campaign.

2) At this point in 2016, Rs only led by 40,000. I could be mistaken but I think yesterday was the last day polls were open on Sunday, so the final "souls to the polls" Sunday.

3) As of yesterday, four red counties were yet to report. That would likely boost the lead back to 70k

4) I don't think the FL Rs in their wildest dreams expected to lead this late in the campaign.

5) Likewise in NV:
Clark (with its union "encouragement" for members turned out big yesterday and added net 6,000 voters.

However---big however---neither Washoe nor the rurals . . .

(5) contd. . . were open yesterday. My guesstimate is that would slice that number in half.

While this helped, the NV Ds are still looking at a significant--to potentially massive--shortfall from their 2016 numbers coming out of early voting.

6) Gotta make sure you compare days to days. Calendars and voting days differ.

One thing that goes unremarked about Wikipedia: the vast majority of the content was written over 12 years ago—which means (given turnover) that the editors dominant today were not directly involved in that creation.

Put another way, the editors who built the dominant nodes in this network...

...have little overlap with the ones who made this much more recent managerial flowchart.

Internet time runs at hundred-fold speed—the difference between the people who painted what's in the Uffizi, and the people in charge of keeping those paintings from deteriorating. Very different tasks, and (one presumes) very personalities as well. @PaulSkallas?