Pretty please, everyone - stop having interviews and talks without budgeting for transcription. Video is agonizing.
Agreed! And please: no summary, no link.
(In this case, that would mean telling the general pattern of the answers, if there is one, and paraphrasing a few particularly insightful points from the interviewee.)
Transcription is a very good candidate for being crowdsourced - everyone can just to a little bit, especially if you start from speech recognition output.
Google is working on that:
"Speech Recognition in YouTube Lets You Search for Spoken Content"
Just pointing out
I did not manage the interview, just made the introduction; but I will convey your input to the reddit guys
despite lack of transcript the linked reddit thread has some metadata, you can read the question and jump right to the answer. This is not very agonizing.
'stop having interviews' - are you serious? seriously? I mean the reddit guys put a lot of work into this; just ask yourself whether you'd rather have the information available or not. It's like telling any free lecturer to not speak unless they give you the lecture notes first.
just ask yourself whether you'd rather have the information available or not
It's a negotiation thing. Wishing you didn't have the information is unreasonable, but preferring probability p of getting no information and probability 1-p of getting the interviews with transcription/summary/whatever over almost-certainly getting the interviews without transcriptions may be quite reasonable.
It's like minimum wage! You can interpret it as "people who can't earn at least this amount aren't allowed to have jobs", but you can also interpret it as "employers must pay all of their employees at least this amount".
It would help if Youtube made it possible to play videos at maybe 1.4x or 2x their regular speed. It's still not as fast as reading, but less slow. I yearn for the day when all of Youtube's videos use the HTML5 video tag, and we can do this at the browser level.
Or listen to the interview while you're playing Super Mario World romhacks (or whatever). That's my overly-specific method for enjoying interview videos.
It would help if Youtube made it possible to play videos at maybe 1.4x or 2x their regular speed. It's still not as fast as reading, but less slow. I yearn for the day when all of Youtube's videos use the HTML5 video tag, and we can do this at the browser level.
I am confused (cf Grice) by that statement, because it's all true about the particular video: if you are in the beta, then it uses the video tag and offers (at least to me) a control to switch to 2x. Even if it is flv, you could download the video and run it in a standalone player. A standalone player also lets you get past the 2x offered by youtube, my chipmunking software (quicktime) doesn't work well past that. I think there are better algorithms, though. (ETA: almost all youtube videos have mp4 downloads, even if they aren't available as html5. I don't actually have a flv codec.)
I used to multitask (and still do with the news in the morning), but I find I get a lot more out of the talk if I pay attention. And running it fast makes that a lot easier, since it's all or nothing.
This is a decent article that summarizes Peter Norvig's 2007 Singularity Summit talk, which made it sound like Google is definitely not working on AGI. http://news.cnet.com/8301-10784_3-9774501-7.html
This interview implies to me that at some point in the future he will assign Google researchers to work on the broad problem of AGI, even though at present Google has no employees dedicated to AGI.
Pretty much, with the addition that they are working on the prerequisites for perhaps other reasons, but I suspect with the AGI potential in mind. Do not remember which one of the 3 CEOs said this, but they know that the ultimate search output is NOT pages of ranked links; it is the answer to your search query. This smacks of an AGI-hard problem.
I think that's probably Larry Page - here:
"Google CEO brags about Google Artificial Intelligence"
(Sorry this is mostly a link instead of a post, but I think it will interesting to the FAI folks here)
I helped arrange this interview with Peter Norvig:
http://www.reddit.com/r/blog/comments/b8aln/peter_norvig_answers_your_questions_ask_me/
I think the answer to the AGI question 4 is telling, but judge for yourself. (BTW, the 'components' Peter referred to are probabilistic relational learning and hierarchical modeling. He singled these two in his singularity summit talk)