Podcast: Play in new window | Download
Subscribe: iTunes | Android | RSS
Justin and Jason discuss the Morning Brief launch strategy, brand concept, and development roles, the difficulties of launching the Math Academy diagnostic in time for the summer program, why Jason had to develop an animation studio for his content team, the underrated “fast zombie” series Black Summer, the sad reality that while everyone in the media seems to have an opinion on the Wuhan lab-leak hypothesis, few are familiar with the basic facts, and likewise for the UAP phenomenon.
I love the Julia language! I think it’s the future, for data science at least. Its package ecosystem is still quite weak compared to Python though. As long as you shuffle off expensive operations to compiled packages python works for most applications, but I’m looking forward to writing more stuff in Julia down the road.
Elixir has also been picking up some steam in ML with Nx, the Erlang JIT and now Livebooks If you’re used to Jupyter notebooks, Livebook is very snappy in comparison.
https://dashbit.co/blog/announcing-livebook
Along a similar line, Jeremy Howard talked about the ecosystem size / speed tradeoffs on Lex Friedman’s podcast. They’re actually writing some fast.ai libraries in Swift now: https://youtu.be/XHyASP49ses
My guess is that Python won’t go anywhere for a while but more and more new things will be done with Julia, Elixir, Swift, Rust and other younger languages over the next decade.
If there were a smart contract with the necessary oracles to make the bet, I would.
You guys were looking for a UFO skeptic directly interviewing UFO eye witnesses and pilots. Mick West has some really good and detailed interviews with many of those folks. For example:
https://www.youtube.com/watch?v=s4-D_v6sWWo Patrick “PJ” Hughes (aviation technician onboard the USS Nimitz in 2004)
https://www.youtube.com/watch?v=uwZU6RiTEAw Alex Dietrich (the second pilot present at the Nimitz “Tic-Tac” encounter)
I have to say personally I am not even a little bit convinced by the UFO evidence, and the latest evidence doesn’t even really seem any better than stuff from 20 years ago.
“Secret and Saucy”, “Did you press record?” – you guys sound like a real professionals sometimes 🙂
Nice episode, enjoyable as always, keep recording !
Joe beat me to the ecosystem comment. The thing with python is that it doesn’t just have a rich ecosystem for data science, it has a rich ecosystem for other operations ranging from web development to making desktop apps. In the data science space, this was a huge initial advantage of R–the ecosystem (and package management). If I want to doing something intensive, then I can offload it. Prime examples are tensorflow and pytorch. I have done cuda calculations and there are still situations where handtuning is worthwhile, but often, developer time just wins. One thing that I might suggest is that there could be a superior solution to grid search depending on your problem space. If it’s say continuous optimization, I’ve found that the DREAM algorithm (efficient MCMC) works brilliantly. If it’s deciding where to measure, then Bayesian optimization is really nice.
For Justin, for me I would love to see summaries of articles rather than just links. One other thought is that a number of us may sign up because you have provided good value over the years–but may not be your best/most representative audience.
Jason–the math academy diagnostic sounds awesome!!!!! I bet there are extremely few people in the education space who are looking at knowledge graphs. Eclectic interests rule!
Oh–one more thing. I agree with Jason on betting. It sharpens the mind and science is famous for some bets. When it comes to the Covid origin story, the question “You hear hoofbeats, is it a horse or a Zebra” comes to mind. I am a physicist–not a biologist, however, there was an interesting talk using topological data analysis by a physicist at a conference I attended virtually earlier this year. It seemed to show that the virus is very consistent with natural origins. This is also consistent with most of what I’ve seen in the scientific literature. There does not seem to be convincing evidence that this was in any way bio-engineered as a weapon. Even the gain of function angle seems rather speculative. It’s fairly common for there to be a long gap between identifying a virus and conclusively finding its reservoir/origin. It would seem that the main question is whether or not it’s purely zoonotic or if in the course of study, it was leaked from a lab inadvertently. This puts us very much in hall of mirrors territory. On the one hand, you have lots of speculation from those that think this is a possibility and mentions of sick scientists (who are unnamed). On the other hand, you have the director of the lab vehemently denying it. More interestingly, you have Danielle Anderson who was recently doing research in that lab (her home base is in Australia where she works now) who claims that she did not see this sickness and would be under far less pressure than anyone in say China. In short, I don’t see the case as being particular strong from a causal angle either way, but if I had to bet, I would put money towards the zoonotic idea and the Chinese government being secretive.
Alright then, I’ve got a second bet to place!
Justin WILL do some ML dev… just as soon as there’s a Laravel service to wrap it up nicely and abstract away a few details 😎
A few comments/questions about MorningBrief:
1. I found the signup process confusing – I tried login but there is no register, not unless I use a social login. Then I did “Play with tech demo” – it asks my name, not sure why as it doesn’t add to the experience except saying “hello matthew”, and doesn’t ask for email address, so how do I register or get on the beta?
2. It shouldn’t ask how many links you want. Make a default and let the user change it later.
3. During the “play with tech demo” process you get asked 3 things – name, # of links, and the tags you want to follow. All that matters of those 3 (for the demo) are the tags. Don’t ask the 2 questions that don’t matter; it just gives users that much more opportunity to disappear.
4. It seems I can pay $100 for lifetime access but why would I pay without even experiencing it? How can I try it?
Curious,
– what crawling tech are you using?
– what service/library are you using for your content analysis?
-Matt
@Matthew – What the… tough crowd! 😉
Ok, imma try to answer your concerns.
– We’ are only selling lifetime access right now to early adopters. You’re right that we need a link to the subscription page on the login page, that will help guide people. Right now you can get to the sales page directly (“Get Lifetime Access”) or by completing the demo (clicking the ‘next’ button). I’ll add a message on the login page!
– Asking for the number of links you want up-front is a demo of the actual product, that’s why it’s there.
– You might be right about the name, but I’m undecided. I’m tracking conversions and seeing where people get hung-up. I’ll circle back to a decision about that. There’s also something really nice about the paid journey where if you already entered your name all that info is there and you start receiving briefs as soon as you pay.
– The tech demo IS the brief. You don’t get anything different if you pay. That’s what goes in your inbox every day. So, if someone doesn’t like what they see they will probably not be interested in the product.
– Re crawling tech and content analysis, we discuss that in the show.
@William – Ok ok, I know you are going to try to do at least 2 full comment blocks here moving forward. 😉 I hear you on the abstract we’re thinking about that.
@Mark – Damn right, laravel and php all the way baby! Did you hear that php is thinking about introducing immutable read only properties… talk about forward thinking 😉
Oh BTW, we already have it – https://www.swoole.co.uk – I’ve heard it can be used fo ML.
Justin, Jason – long time listener here, I find much inspiration from your casts. It is encouraging to listen to other folks facing the same challenges in various forms. Please continue to cast, I find them quite meaningful and thoughtful/provoking.