Podcasts

Podcasts are like an electric campfire, people gather around and tell stories

2020-08

Jim Keller @ Lex Fridman Podcast #70

  • Jim Keller yra ko gero vienas žymiausių žmonių moderniųjų mikro procesorių sferoje. Juokaujama, jog visa mikroprocesorių rinka yra Keller'is šokinėjantis iš vienos kompanijos į kitą ir bandantis nukonkuruoti savo paties sukurtus produktus. Žinoma tai tik juokelis, naujos architektūros kūrime dalyvauja tūkstančiai žmonių, tačiau užtenka truputėlis paklausyti jo kalbant podcast'e, kaip jis bando supaprastinti iki abstrakcijų išties sudėtingas koncepcijas ir tuomet supranti kiek daug insight'o šis žmogus turi.

  • Mr. Keller approaches the usual microprocessor design loop, where a company develops a new design from scratch and then looks at the most fundamental way of adding performance. Usually, he says, easy 10% performance increments can be found by simply looking at a design and increasing execution units - increase a buffer here, increase a cache over there, put in another add processor on this part of the pipeline. However, he also speaks of how this process in itself is limiting, inasmuch as doing this often will eventually guide processor designs towards a bottleneck and the diminishing returns problem, where any more additions made to the design don't seem to increase performance - mostly just adding complexity, area and power requirements, and generally convoluting a given design. Mr. Keller says that that is the point where teams should see that the design in itself is already limiting - and that that's when you do something from scratch. Fast advances in performance in the world of microprocessors depend on 3-5 year cycles of building a new processing design from scratch. His work in the microprocessor world usually comes in at this point - Mr. Keller has established himself as one of the foremost leaders of microprocessor design teams when looking at a clean-slate design, with his stints at AMD (collaborating on their K7, K8, and K12 designs alongside the HyperTransport bus), Apple (for their A4 and A5 microprocessors, which are now in the iterative phase), Tesla (with their HW 3.0 designs for autonomous driving) and AMD's Zen.

  • I imagine 99% of your thought process is protecting your self-conception, and 98% of that is wrong.

  • A lot of things need to be "rearchitected every decade, or even faster".

2020-05

George Hotz @ Artificial Intelligence Podcast

  • geohot yra pasipūtęs, temperamentingas ir stačiokiškas trolis, tačiau užtenka išgirsti jį kalbant apie savo stichiją ir iš karto pasimato jo ambicingumas, didelis intelektas bei itin gilus temos perpratimas. Šitu aspektu jis man primena Martin Shkreli. Daug kas linkę susidaryti greitą neigiamą įspūdį, tačiau tokių žmonių skill'ai kalba patys už save.

  • Man patinka jo nematerialistiškas požiūris ir biurokratijos engimas. Jis yra labai competitive, mėgsta išties didelės kompleksijos problemas ir pagrindinė jo varomoji jėga - laimėti, "parodyti kitiems" išsprendžiant jas pirmesniam. Išties gerbia žmones už jų realius skill'us ir vienas pagrindinių jo troškimų - būti žinomam būtent tokių, išties talentingų žmonių tarpe.

  • Podcast'e patiko kontrastas tarp kalbančiųjų: Hotz'as greitas ir chaotiškas, Lex'as tuo tarpu santūrus, lėtas ir apgalvojantis. Vedėjas išlaikė nepatiklumą ir paspaudė pašnekovą tose vietose, kuriose jų požiūriai išsiskyrė ir dėlto diskusija tapo įdomesnė. Norėčiau, jog tokia dinamika būtų buvusi ir tuomet, kai to paties podcast'o svečias buvo Musk'as, nes tuomet jis daugiausiai kartojo savo jau atkaltas supaprastintas formuluotes. Manau jog Lex'as būtų tas vedėjas, kuris sugebėtų pokalbį pakreipti į daug smulkmeniškesnę techninę pusę ir paspaustų Musk'ą pakalbėti apie daug gilesnius inžinerinius niuansus, ko ir pritrūksta per visus jo interviu.

  • Elon Musk was looking for someone to build a vision system for Autopilot. We talked about a contract. If I could deliver something that meets Mobileye level performance tomorrow I would get $12 million. I would lose $1 million for each month I couldn’t deliver.

  • In life, I usually get away with saying stupid things. Elon is very sharp. He immediately called me out on it. Most people don’t even notice.

  • Tesla is going to win level 5. Tesla is gathering data on a scale no one else is. They are putting real users behind the wheel. The incremental strategy is the only strategy that works.

  • You only have to mess up once to be caught as a criminal. That's why I'm not a criminal

  • What have you learned about the design of systems? You learn to not take things for what people say they are. You look at things for what they actually are. I understand that is what you tell me it is, but what does it do?

  • A huge success for driver-assist programs would be if they could fix drivers that are drunk, distracted, and asleep. [The causes of a large percentage of fatalities.]

  • The long term idea for Comma.AI to make money: Be a car insurance company. I have the best data set to see who statically are the safest drivers. We see you driving unsafely we won’t insure you. This causes a bifurcation in the market because the only people who can’t get Comma insurance are the bad drivers. Geico can insure them. Their premiums are crazy high. Our premiums are crazy low. We would win car insurance.

  • I don’t want attention from the masses. I want attention from people who I respect. I respect people who have skills. I would like to be known among people who have real skills.

  • When I say that life is work what I mean is that work gives my life meaning. I don’t mean that every minute of the day you should be working. I don’t think that is the best way to maximize results. If you are working 12 hours a day you should be working smarter - not harder.

  • The meaning of life: I'm just trying to figure it out what the game is about, then reduce the uncertainty about what the reward function is and try to maximise it.

2020-04

Stephen Dubner (co-author of the Freakonomics book series and host of Freakonomics Radio and podcast)

  • MMA - it’s high-level problem solving with dire physical consequences

  • We’ve been interested in people beating the shit out of each other since the beginning of time.

  • Getting fired, dumped, or losing a game is important because the feelings of failure make the feelings of success sweeter.

  • Feelings of failure are really critical for your motivation. If you’re afraid of failing, but you go out and fail anyway, you learn that fear won’t kill you. You can view failure positively or negatively, the choice is yours.

  • We’re so often married to our ideas like our ideas are a part of us and we’re losing if our ideas that we’ve been discussing are incorrect.

  • You have to be willing to have unpopular opinions to make new discoveries. The world doesn’t progress very much unless there’s a guy like Steve Jobs or a Copernicus…poking people in the eye with a stick…you know very often they’re discredited or hated for their whole lives.

  • Most nuclear plants were built in the 1970s and new ones haven’t been built because people are scared of meltdowns–However, science has improved dramatically. One analogy: compare the safety of a car built in the ‘70s with a Tesla. Things have gotten much better. The same is true of nuclear plants.

2020-02

  • Garrett Reisman - buvęs NASA astronautas, astronautikos inžinerijos profesorius ir SpaceX konsultantas pasakoja apie savo patirtis 2 savaites gyvenant vandenyno dugne, apie tai, jog Battlestar Galactica serialą jis žiūrėjo būdamas tarptautinėje kosmoso stotyje ir apie tai, kaip buvo nusivylęs, kai pirmą kartą pamatė žemę iš kosmoso.

2019 - 09

  • John Carmack programuotojų, inžinierių ir developerių tarpuose pristatinėti nereikia, jis jau ir taip yra legenda. Tikrai planuoju paskaityti Masters of Doom, o iki to laiko - interviu pas Joe Rogan'ą, kuriame kalbama apie AI, VR, raketas ir turbo boosted ferrarius.

  • He's the kind of guy who gets passionate about something and he gives it his all and enjoys sharing his knowledge.

2019 - 07

Internet History podcast'as apie pirmojo iPhone sukūrimą

2019 - 06

  • Naval Ravikant - vienas iš angel investorių AngelList kompanijoje

If you want to see who rules over you see who you not allowed to criticise.

Working for things as rewards is a real trap there a lot of people follow into.

Netapsi turtingu vien tik pardavinėdamas savo laiką. Kiekvieną kartą, kai pradedi charginti daugiau už savo darbą, prasideda lėtas upgrade loop'as ir tavo išlaidos taipogi padidėja. Savo pajamų prieaugį panaudoji gyvenimo kokybės nedideliam pagerinimui, galbūt pasidarai remontą, susitvarkai ar nusiperki automobilį, perki butą, persikraustai į geresnį rajoną ir t.t. Ir gan greitai pradedama vėl sieki didesnio charginimo, nes niekas pernelyg nepasikeitė. Kitaip tariant - jei nori užsidirbti, turi tiesiog daug dirbti. Jeigu nori praturtėti - turi daryti kažką kito. Dėl šios priežasties net dideles pajamas gaunantys žmonės, kaip pvz teisininkai, gydytojai ir pan taipogi nėra tokie jau išties turtingi, t.y. neturi sutaupę milžiniškų sūmų savo sąskaitose. Tai ypač pasitvirtina, jei turtu laikyti asset'us, kurie tau uždirba pinigų.

There are two great addictions - heroin and the monthly salary.

Specialisation is for insects.

Žmonės randa vieną sritį, kurioje jiems sekasi ir dažniausiai lieka joje visą likusį gyvenimą. Dėlto, kad patogu, dėlto, kad taip nepasirodys neišmanėliais ir dėl baimės pradėti viską iš naujo. Romėnai, graikai turėjo visą ciklą - eini į mokyklą, dalyvauji kare, vertiesi veslu, tarnauji senate ir galiausiai tampi filosofu. Kitaip tariant turi daug skirtingų "karjerų", išbandai save keliose visiškai skirtingose srityse. Tarsi gyvenimo arka. Dabar dažniausiai taip nebegyvename, specializuojamės.

We each have two lives. The second one begins, when we realize we have only one.

2019 - 05

  • Jamie Metzl - futuristas, knygos Hacking Darwin autorius

One of the biggest stories of the 21st century will be how the US - China rivalry plays out and the playing field will be revolutionary technologies and China has a national plan to lead the world in these technologies by 2050 they're putting huge resources they have really smart people they are really focused on.

China is a country run by engineers... America is a country run by lawyers.

  • Lex Fridman - mokslininkas dirbantis MIT ties Human-Centered AI & Autonomous Vehicles

Once you become this thing that doesn't need emotion doesn't need conflict it doesn't need to be inspired it never gets lazy it doesn't have these things that we have built into us as a biological system like if you looked at us as we're operating software it's it's not good software right it's software designed for cave people and we're just trying to force it into cars and force it into cubicles but part of the problem with people and their unhappiness is that all of these human reward systems that have been set up through evolution and natural selection to have these instincts to stay alive they're no longer relevant in today's society so they become become road rage they become you know extracurricular violence it became depression they become all these different things that people suffer from

2019 - 04

2019 - 03

  • Alex Honnald - kur kiekvienas interviu džiugina jau vien tuom, kad pašnekovas vis dar čia, vis dar gali atvykti. Alex'as yra žymiausias Free Solo kopėjas, kuris vienintelis įlipo į garsiausią JAV kalną be jokios įrangos. 22 metus vyrukas vaikšto tokia slidžia riba, kur viena vienintelė klaida, jam kainuotų gyvybę. Podcast'e pasakoja apie tai, kad kopiant įsitempimas turi būti tarp 3,5-7, nes kitu atveju yra didelė tikimybė padaryti klaidą. Apie tai, kaip kopikai pasiilsina rankas kabėdami žemyn galva ir apie tai, kad turi būti pakankamai įsitempęs, kad neprarastum savo general alertness, tačiau tuo pat metu ir pakankamai atsipalaidavęs, jog jaustumeisi komforto zonoje ir nenuvargintum savęs. Yra puikus filmas apie tai, kaip vystėsi visa ši kopimo kultūra, pavadinimu Valley Uprising.

  • I don't like holidays because I don't like being told when to have fun