Bjarne Stroustrup: Deep Learning, Software 2.0, and Fuzzy Programming

This is a clip from a conversation with Bjarne Stroustrup from Nov 2019. New full episodes are released once or twice a week and 1-2 new clips or a new non-podcast video is released on all other days. You can watch the full conversation here: • Bjarne Stroustrup: C++...
(more links below)
Podcast full episodes playlist:
• Lex Fridman Podcast
Podcasts clips playlist:
• Lex Fridman Podcast Clips
Podcast website:
lexfridman.com/ai
Podcast on Apple Podcasts (iTunes):
apple.co/2lwqZIr
Podcast on Spotify:
spoti.fi/2nEwCF8
Podcast RSS:
lexfridman.com/category/ai/feed/
Note: I select clips with insights from these much longer conversation with the hope of helping make these ideas more accessible and discoverable. Ultimately, this podcast is a small side hobby for me with the goal of sharing and discussing ideas. I did a poll and 92% of people either liked or loved the posting of daily clips, 2% were indifferent, and 6% hated it, some suggesting that I post them on a separate KZbin channel. I hear the 6% and partially agree, so am torn about the whole thing. I tried creating a separate clips channel but the KZbin algorithm makes it very difficult for that channel to grow. So for a little while, I'll keep posting clips on this channel. I ask for your patience and to see these clips as supporting the dissemination of knowledge contained in nuanced discussion. If you enjoy it, consider subscribing, sharing, and commenting.
Bjarne Stroustrup is the creator of C++, a programming language that after 34 years is still one of the most popular and powerful languages in the world. Its focus on fast, stable, robust code underlies many of the biggest systems in the world that we have come to rely on as a society. If you're watching this on KZbin, many of the critical back-end component of KZbin are written in C++. Same goes for Google, Facebook, Amazon, Twitter, most Microsoft applications, Adobe applications, most database systems, and most physical systems that operate in the real-world like cars, robots, rockets that launch us into space and one day will land us on Mars.
Subscribe to this KZbin channel or connect on:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Support on Patreon: / lexfridman

[@lexfridman]

This is a clip from a conversation with Bjarne Stroustrup from Nov 2019. New full episodes are released once or twice a week and 1-2 new clips or a new non-podcast video is released on all other days. If you enjoy it, subscribe, comment, and share. You can watch the full conversation here: kzbin.info/www/bejne/q4Xbg3lrlsxlbaM (more links below) Podcast full episodes playlist: kzbin.info/aero/PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4 Podcasts clips playlist: kzbin.info/aero/PLrAXtmErZgOeciFP3CBCIEElOJeitOr41 Podcast website: lexfridman.com/ai Podcast on Apple Podcasts (iTunes): apple.co/2lwqZIr Podcast on Spotify: spoti.fi/2nEwCF8 Podcast RSS: lexfridman.com/category/ai/feed/

[@perpetualmoto]

Thank you for these conversations!!

[@FaruhPaerhati]

These are awesome talks, thank you for sharing these Lex

[@salifouaffoh9626]

"All that AI stuff is on top of c++". That's the point

[@fvw94]

I dont think that is true...

[@salifouaffoh9626]

@@fvw94 what i think he means by saying that is the need of Performance especially by python calling a bunch of C/C++ native code with libraries like numpy through cython. Python is like an Interface to quickly and bug safely write Code when the hard work is done behind the scene.

[@user-xs9ey2rd5h]

Tensor flow is build in c++ for instance

[@fvw94]

@@batkomahnoX100 A compiler is sth that translates a code to machine code. Sayin that they are written in C++ is incorrect

[@fvw94]

@@salifouaffoh9626 u talk about performance, but I havent heard Cuda yet

[@mavhunter8753]

I think the big take away from this video is that, it takes all kinds to make this world, and one thing shouldn't necessarily replace the other. I couldn't agree more.

[@mikeg9b]

A fuzzy program called Bjarne Stroustrup created C++.

[@furkansarihan]

I laughed

[@awsomeguy563]

Wow this is such a good example. kudos

[@sapienspace8814]

42

[@leodu561]

My daily dose of C++ love

[@sapienspace8814]

Even in Avionic design, failure is categorized in probabilities. Catastrophe happens when standards of certification are lowered (e.g. flight controls at Level D instead of Level A: 737-MAX).

[@EnginAtik]

“C++ versus AI” is not a good question. Current AI systems not having rigorous theory complete with mathematical proofs etc in certain applications is probably what Stroustrup is driving at.

[@CirclesOfMotion]

AI (LcZero) win C++ (Stockfish) in chess already. Also neural network can approximate any function therefore there is no things that C++ can do and AI can't with given accuracy. So all this talk about "fuzzy" nature of AI sounds as nonsense for me. "Fuzzy" nature of AI appearing in tasks where other paradigms just give up.

[@EnginAtik]

@@CirclesOfMotion TensorFlow was written in C++. So it is not AI versus C++. Fuzziness is not necessarily the problem either. AI solves a practical problem but we don't understand the solution. We can come up with experimental stability margins and other measures about the AI's solution and approve it for certain applications but we don't have much of a clue if and when something fails. The other extreme is, sooner of later best move for every position in chess will be tabulated (64-digit octal number of positions) then the best chess program will be looking up the next move from a big table. Obviously such a program does not need much intelligence and has no knowledge of chess but it never makes any mistakes. The whole chess game becomes a function that takes and integer value and produces an integer value. So the issue is not AI either: if we can tabulate the output for every input combination for any problem then everything is a static lookup. In AI's toolbox we have a set of handy tools to create this lookup table which boils down to a nonlinear curve fit. We can extrapolate the output using this curve for inputs that are not in the training set. On the other hand, we humans have limited computation and memory resources and don't have the luxury to do any tradeoff between them. This makes us model the physical world by building on the first-principles and If the model we build stands tall on these foundation we call it cognition/understanding etc. Why we have this urge to understand the outside world is a subject of psychology.

[@BeibinLi]

​@@CirclesOfMotion What is AI built upon? Low-level binary code? Or, is LcZero written in C++ or CUDA? They are just two different worlds... Not an apple-to-apple comparison.

[@itsbro9527]

@@CirclesOfMotion Stockfish is better now!!! Check the rankings...

[@mandaputtra]

Programming aren't just this language can do that this languange can do this, its how you achive something, to achieve something you need tools. The hammer on AI maybe a Python, because PHP cant still be used, there is no tools around it for now. But if I make tools for PHP for doing the machine learning, or that fuzzy stuff maybe PHP can be used now.

[@brunofelix5434]

I Love C++ ❤

[@hamoudrodriguez2702]

What does he say at 5:35 "all this AI..." I couldn't get it.

[@alexandremarcil498]

"all of this AI stuff is on top of C++..."

[@hamoudrodriguez2702]

@@alexandremarcil498 thank you!

[@FilipPavlovicZen]

@@alexandremarcil498 Thank you. I also couldn't get that and neither could KZbin's captions AI.

[@user-hh2is9kg9j]

There is a hot Porn actress called Ria Rodriguez. is she a relative?

[@hamoudrodriguez2702]

@@user-hh2is9kg9j no, never even heard of!

[@carlossegura403]

A biological system depends on both "fuzziness" and a well-defined structure. For example, the human skeleton is a fixed geometrical form (core). While organs, brain, skin, etc. are fuzzy/dynamic.

[@mapleandsteel]

Bjarne Stroustrop sounds like a logical positivist. Is he?

[@abdelhaksaouli8802]

Describing biologicals system as MESS is kind wired especially when this unreliable mess is the source of inspiration to something that work 99.8%

[@incelstate2466]

Will somebody please address what happened to his hair?! Pleas for the love of god where did Bjarne Stroustrup's hair go?

[@incelstate2466]

It is missing even on his youth pictures

[@atlmember4045]

Have you never seen a bald man?

[@ansismaleckis1296]

he lost his hair when debugging a c++ compiler

[@sapienspace8814]

Fuzzy

[@benjaminrohling6365]

you could make a 3D learning matrix and also create 3D layer CPUs.

[@benjaminrohling6365]

practically cubes instead of dyes

[@Houssein56]

Ai shouldn't take care of self driving car 😂 , comparing this discussion with the one with karpathy proves how much someone can disagree with his future self

[@minecrafthowtodude]

2:30 haHAA, how you think most comp sci students are getting through their degrees? We know basic algebra, and time/space complexity.

[@viadharmawheel]

Yep, RoboTaxi is now coming out in 2021, LOL...

[@leeamraa]

Apples to Oranges!!!

[@mavhunter8753]

I think AI is only really good for dealing with repetitive tasks that computers would be more effective at. The thing that drives biological systems is millions of years of evolution, AI simply doesn't have that.

[@Actrl51]

AI could potentially shrink millions of years of iterations and learning into a couple of minutes.

[@1luarluar1]

yes, but that exactly the point...AI does not need million of years, but a fraction of that time...is a milestone that will change radically what we thought about evolution..in a few years Ai is able to drive cars...it took us millions of year to get there...AI is already able to win games that requires human creativity and intuition...so, it will be far better than us in every field...we received the gift of intelligence (we were just mud and water)...now it is time to donate this gift to someone\something else...that's the way it works...we are not the end of evolution, that's for sure.

[@user-yk1cw8im4h]

Raul Gubert It didn’t take us millions of year to able to drive cars, if thats the case nobody would be able to learn how to drive within their lifetime...

[@1luarluar1]

@@user-yk1cw8im4h The '' system man'' took millions of year to be the way it is (of course now in a couple of months are we able to drive)...''the system man'' did not fall from the sky in one day..it became like it is now after millions of year of evolution....the system AI is just, let's exaggerate, 100 years ....I hope you understand what I mean...cheers!

[@awsomeguy563]

@@user-yk1cw8im4h I don't think he's talking about an individual human, but the human species as a whole. The "system" that allows us to learn how to drive cars has been developed over millions of years. An individual human uses that system to learn how to drive in days. The system being our ever evolving brains.

[@sedaahmadova8798]

Переводите, не понимаю

[@sedaahmadova8798]

@@casanova2829 а я не знала что учить надо

[@zerberus1097]

@@sedaahmadova8798 учи английский

[@1luarluar1]

I don't agree of what he is saying...programming is great and absolutely necessary but then there are things you cannot do with programming, you need a different system to achieve what you have to do....programming for a self driven car is not plausible otherwise we would already have self driven cars...AI is a different approach to what we want to achieve and it is working! we are very close to get a reliable autonmous car...when the system is fully capable, I'm sure we will have a fraction of the car accident we have now...but there are plenty of other problems AI has already fixed...definitely there are risks, but that's the power of this new approach...no one is scare of programming, that's because it pales in comparison with what AI can do...

[@gnewb9658]

clearly you dont understand how AI works. The base layer of AI is fully programmed, more often than not in C++, the language Bjarne created.

[@KIXEMITNOG]

C++ IS THE KING

[@Daniel_WR_Hart]

Machine learning systems are like pets, they're messy fuzzy things

[@accumulator5734]

I love Bjarne but that example of a self driving car asking for help while your reading a book makes him look way out of touch, but to his credit he also did imply he was not interested in the topic. At this point in time your not suppose to be reading books with a self driving car lol, ur suppose to be paying attention. But as the neural networks improve then eventually the car is suppose to be safer than even the driver.

[@njanirudh]

Lol Then wtf is it called fully autonomous car if we need human guidance all the time?

[@a2xd94]

None of the autonomous vehicle conversations revolve around ‘what is the % of accuracy of a human vs a machine doing equivalent roles in this system’...but rather whether a regular human has enough trust the machine with their life. Literally. Sadly yet understandably, that conversation is not as easy as it seems and that part is entirely up to humanity’s viewpoint on technology. The more positive and trusting it is, the faster technology which inevitably will enable autonomous driving will advance.

[@r.a.monigold9789]

Myopia is self focusing, then. Have you had much experience with contemporary assisted cruise or autopilot driving? Perhaps a bit of reading or even "YouTubing"? When you do, be prepared to be pleasantly surprised...

[@weignerg]

This guy whistles way too much for me to focus on what he is saying.

[@rosshoyt2030]

Yep that's just how Bjarne is. However, he invented C++ so if you want to learn about that, you need to learn to tolerate the whistle. Lol

[@Almatrass]

Why did Bjarne even take on this interview? He must be insulted to turn up to have such an incompetent guy ask the most ridiculous questions

[@gnewb9658]

i was looking for this comment, lex saying he works in machine learning and implying Bjarne is not familiar with it jesus christ lex xD

[@ZerofeverOfficial]

The first AGI is going to make Stroustrup look like a toddler. Should have asked him how he feels about that. Seems like he feels threatened.

[@stifflery]

How he feels about what? The AGI, that doesn't exist? It's nothing more than an abstract concept for now. By the way whenever or if it comes out (highly unlikely in our lifetime), isn't that going to make ALL of us look like toddlers?

[@jjsullivan5196]

You gotta check your hype on this one. He's just drawing the boundary between artifacts of software engineering and ML research. Not every software engineer needs to know/work in ML systems (data literacy is always a plus though), and vice-versa. I used to be really into the kind of work that goes into ML, but I burned out when I realized that ML isn't going to get you concrete actions and outcomes. The process of engineering is much more interesting to me, rather than model optimization.

[@ThePandaGuitar]

AI is overhyped and the kids are wasting their time going down that rabbit hole. I don't want no ML algo to fly my airplane.

[@undertest1708]

I like your show Lex, but Bjarne lost my respect when he bashed into Bitcoin.