Podcast: Play in new window | Download
Subscribe: iTunes | Android | RSS
Justin and Jason discuss the latest with Math Academy – Jason’s class, the district program and the software, temperpedic beds, the problem of living in the middle of a “fun” area of town, why Jason doesn’t want stuff and what he does want, Justin’s new “super router”, why getting a storage unit is generally a mistake, Phil’s hack for how to emotionally detach from books he needs to get rid of, the evolutionary advantage and repercussions of human curiosity, Duolingo and granular learning, why Jason thinks learning a foreign languages is rarely worth the time, Justin’s new work-at-home situation, how website design work is drying up, why future actors will have to compete with CGI versions of actors of the past, Justin’s idea to create AI from the DNA level, the path and prospect of strong AI, Jason’s desire to create a synthetic hedge fund using Collective2 and the prospect of a Trump presidency.
The discussion around curiosity and learning was pretty interesting. Kind of explains why some of us going down these link spirals sometime.
Hey Justin, I think what you were describing when you were discussing an AI based on DNA is basically neural networks which are evolved using genetic algorithms.
It seems very similar to what you were describing, genetic algorithms use encodings which are comparable to our DNA, then use the idea of survival of the fittest and replication to adaptively find a solution using that encoding. They’re a general search algorithm so they can be applied to any problem, but they’re often applied to neural networks to evolve a network.
Not sure how familiar you are with them, but I’ve recently wrote a book on them if you have any questions.
—
Also, an interesting theory I’ve heard about conscious is how it might be something which evolved to give a social animal a better understanding and appreciation of pain etc. We could program a robot so that when a button is pressed the robot is in a state of pain. We might even program the robot to react a certain way when it’s in pain. But we would understand that robot isn’t actually feeling anything. In this example, that “pain” is really just an input or a state which changes the output of the robot. Most people would have no problem putting this robot in “pain” because they understand it isn’t feeling anything “real”.
The only difference between us and this robot is consciousness. We understand that pain isn’t just a state or input for other humans. For that humans it’s really a “physical” thing. Pain to a human is as real as a tree or a rock is. So in short consciousness is the reason we care that someone is hurting. We know this because people don’t care much about hurting things that they don’t believe are conscious, we’re quite happy to pull legs off of spiders for example.
So perhaps conscious is just where we evolved to perceive certain non-physical experiences as “physical” so we can appreciate those experiences more? I’m not sure if I buy it personally because I’m not sure if consciousness is the simplest or only way to do it, but it’s the theory which makes the most sense so far to me. It’s undeniable that there is a difference between understanding your in pain as opposed to being in a physical state of pain.
Hey Justin,
On the notion of DNS/body based AI. Or bottom-up as Jason described. This is my favorite science fiction story exploring that concept: http://subterraneanpress.com/magazine/fall_2010/fiction_the_lifecycle_of_software_objects_by_ted_chiang/
Hi!
Long time listener here. I think I have been listening for 3 or 4 years at this point. Today when I listened to the podcast I had a revelation. I realized that I had the two of you switched around from the picture. This means that everytime I’ve listened to you speak on the show I’ve visualized the wrong person. It is surprisingly hard to wrap my head around it now and visualize the correct person instead. The weird thing is that I have these memories of all the things you’ve talked about over the things, but now that I’ve come to this realization I have to reevaluate the memories. It’s a very peculiar sensation!
Anyways – just want to say that both of you are doing a great job with this podcast. It’s still interesting to hear you guys talk and it’s nice to have Phil on the show more often. He’s a breath of fresh air.
16GB SDCard… Sure Justin will fill it up, just find the convenient topic, bio hacking maybe 🙂
um, Jason, not Justin… Peter you made me confused lol
Scott Adams (Dilbert creator) has been doing an amazing series on Trump and his use of persuasion techniques. Start here: http://blog.dilbert.com/post/126589300371/clown-genius
It’s really been fascinating to watch this political season from the lens that Trump is a master persuader — you will constantly see that he uses emotion and identity over logic and reasoning and it is working. 90% of people are irrational and Trump is exploiting this with amazing technique. Combine that with his disruptive “growth hacks” for getting over a billion dollars worth of earned media and the general entertainment factor of Trump and it’s been a fun ride.
There are also some rumblings of Mark Cuban running as Trump’s VP. The idea of someone with a strong entrepreneurship and tech background holding a high position in government is very appealing to me.
@Peter & Tolga – Jason is the American with blonde hair (Uber, Catalyst, Math Academy, Operation Superhero, etc) and Justin is the Brit with not so much hair (Modern Teacher, Pluggio, Lyte, former musician and occasional dieter).
@Lee – Wow, thanks for that. So encodings are all you need to recreate the entire system? Cool.
Re consciousness. This may sound nuts, but I have the idea it is some kind of etheric state that is created by and experienced by the physical brain but itself it is not a actually the physical brain. For example if the brain was a reel of film & projector… conciseness would be the movie that was projected by running the film through the projector… and I also wonder if that principe of in the moment consciousness could be applied to ai.
@Jon – Thanks I will check that out!
@Jason – Funny dude 😉
Just able to get around to listening.
The automation discussion is interesting. One thought that I had about the web design market aspect raising the lower bound (nice observation Phil!) was that the middle of the market may get eaten out as “good enough” moves towards zero (or rather, the new lower bound). Marginal improvements over that lower bound over won’t be perceived as being worth it, while the impressive work will command a premium.
As the discussion covered putting smaller and smaller pieces of knowledge onto a screen, I began thinking about encapsulation, both as a programming concept, but also as a general knowledge concept. For example, when we talk about a function, Jason, Justin and I’m assuming Phil would fully understand, but the concept would need to be explained to someone learning programming. We have encapsulated the various discreet pieces of knowledge that comprise the concept of “function” in our understanding of “function”.
The language learning part of the discussion, especially Phil’s description of Duolingo’s use of gamification and streaks has been an interest of mine. I was having a discussion with my cousin a couple of weeks ago about how working streaks into your life (assuming in a beneficial manner), has a complication in that you can track your steps for the day, and how many times you’ve performed an an action on a web site, but, those aren’t integrated. You can game any of these individually, but collectively, it’s harder.
You almost need a API to handle streak data that can accept input from various locations, could be different applications, from meetings on your calendar, your GPS indicates you’re at the gym, check in’s on GitHub, etc. You’d then be able to analyze all of the events and derive the streaks and maybe even predict trends based on correlated that appear from the data. Of course, the biggest issue here is walled gardens, especially from device manufacturers.
Thanks again for the conversation guys!
The founder of Duolingo is Luis von Ahn and his story is fascinating. A lot of smarts going into Duolingo. Tim Ferris (I think also an investor in) interviewed him. http://fourhourworkweek.com/2016/01/26/luis-von-ahn-duolingo/ I could have sworn I got it from you guys, but just in case… For reading long form articles Spritzlet is indispensable. I’m sure you’ll love it http://spritzlet.com/
Great show as always, and interesting discussion on learning.
DuoLingo is fantastic, I’m using it to pick up some Russian at the moment. And kudos to Colby for learning languages out of sheer interest at his age. Bra jobbat med svenskan, Colby! 🙂
However, I think DuoLingo works best at beginner levels, while it’s not as effective if you’re already at an intermediate to advanced level. At least that’s how it is for me.
And sure, if you try to apply a ROI to learning languages, it’s rarely worth it unless you explicitly need it in your work. But hey, that’s probably the case with most things we spend time on outside of work (and some of the things at work as well). Look at all the content we consume daily – how much of that is actually well invested time from a business point of view? Personally, I just enjoy learning about artificial intelligence, space exploration, coding, languages and a shitload of other things that will never really pay off. 🙂
Travis of Uber has a TED Talk up: https://youtu.be/pb–rJGgVIo. It’s a chance to see a show recurring character speak for himself. Will he mention his Wii Tennis ranking? How he missed out on hiring the CTO he wanted?