Today we have @SarahGriff90 discussing her recently published paper 'Early language competence, but not general cognitive ability, predicts children’s recognition of emotion from facial and vocal cues'.

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

This post got a little delayed since the paper's publication, but the intervening time has given me an opportunity to reflect.

So this post has some extra thoughts regarding more recent NLG developments - e.g., pretrained models like GPT(2), and the suitability of beam search.

👇 Thread. New deck getting published this week: "Consumer startups are awesome, and here's what I'm looking to invest in at Andreessen Horowitz." If you want to read it, subscribe to my newsletter here: https://t.co/262t8eh0wf

1/ A lot of new consumer technologies have been introduced to US households in the last 100 years. But it's taken many of them - like the telephone - more than 50 years to get to the majority of the US. Why is that?

2/ We had to literally teach people how to use phone. Which end goes to your mouth, which goes to your ear. Say "hello" when people call. The motivation of consumers to talk to their friends has always been there, but we had to teach the behavior

3/ If you compare phones to the latest technologies, there's been a huge shift. Things are being picked up much faster.

4/ Even while there's been all this innovation recently, physically speaking, we are still the same human beings from 100,000 years ago.