Jim Keller: Most People Don't Think Simple Enough | AI Podcast Clips
ฝัง
- เผยแพร่เมื่อ 9 ก.พ. 2020
- Full episode with Jim Keller (Feb 2020): • Jim Keller: Moore's La...
Clips channel (Lex Clips): / lexclips
Main channel (Lex Fridman): / lexfridman
(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/
Jim Keller is a legendary microprocessor engineer, having worked at AMD, Apple, Tesla, and now Intel. He's known for his work on the AMD K7, K8, K12 and Zen microarchitectures, Apple A4, A5 processors, and co-author of the specifications for the x86-64 instruction set and HyperTransport interconnect.
Subscribe to this TH-cam channel or connect on:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Support on Patreon: / lexfridman - วิทยาศาสตร์และเทคโนโลยี
![[TH] 2024 PMSL SEA GF D3 | Summer | มาร่วมเป็นพยาน กับวินาทีแชมป์!](http://i.ytimg.com/vi/JCYvEsFlwx8/mqdefault.jpg)
2:19 - "If you constantly unpack everything for deeper understanding you never get anything done. If you don't unpack and understand it when you need to, you'll do the wrong thing." Loved this, well said.
Right, that's where experience helps. Michael Clarke is ground up redeveloping the Zen5 design following Jim's philosophy.
11 months later... whelp, we're there.
Reminds me of Bruce Lee. "If you have one to the extreme, you'd be very unscientific. If you have another to the extreme you are all of a sudden a mechanical man, no longer a human being"
honestly there is a point where digging any deeper causes too much impatience and unfulfillment within me that I 'naturally' say "right let's use this knowledge and make something work"
everybody's limit is different though and that makes it challenging to work with many other people when developing or fixing something, there are always those who want to go further and wider, and others just want something that works for now
"An idiot loves complexity, a genius loves simplicity" - Terry Davis. God I wish he were alive to be interviewed!
Thing about terry was you needed to extract the moments of his genius from the psychotic ramblings about the CIA.
"Rewrite more often!" - Less about perfection, more about action, more synchronization to end user needs
Business types hear that and think it's less about action. The difference.
It’s dangerous because you need to have someone who knows how to rewrite. In software, there is usually the second system effect. Jim is saying the second system is faster (when he writes it?), but they usually fail. Most programmers either hunt for an elegant all purpose solution or become architecture astronauts and lose the lessons of the messier straightforward code.
loving Jim's Logical mind, he is so exact and to the point .
The fun stops when you're about to rebuild an AI from scratch and it says:
"no, don't worry, I got it"
Luke Passos paaahahaha
The AI is going to say. You're using up valuable resources but I have a solution for that.
The fun begins!
When you hit the build button, but all that happens is your laptop speakers say: "I'm sorry, I'm afraid I can't let you do that... Dave".
user:
>mkdir AI_2
computer:
"say no more, fam"
4:52 Lex: "So... where does the x86_64 standard come in? Or, how often do you..."
Jim: "I was the co-author of that spec back in '98."
Completely badass.
And by co-authored he means, "I wrote it, and a colleague rewrote it to make it easier to understand for other people" :P
I wasn't even surprised seeing his other accomplishment
x86 was already 20 years old in 1998, and that was 25 years ago. I wonder at what point we'll ditch x86 and just start over...
@@rockets4kids Well he is working on RISC-V and his stated opinion was every 3-5 years.
"If you really want to make a lot of progress in computer architecture, you should write one from scratch every 5 years"
I'll keep it in my frontal lobe cortex in my next SW job, sir.
Also they found a new way to produce Graphene, called Flash Graphene.
Here are 2 videos on it.
th-cam.com/video/hzm5AMPFMqs/w-d-xo.html
th-cam.com/video/GzDrnoGdLO4/w-d-xo.html
Graphene as a super conductor.
th-cam.com/video/HSn57YVDvHg/w-d-xo.html
Also check out Archer Exploration also known now as Archer Materials, a company from Australia that uses Graphene to make Quantum Computers for room temperature, for our homes.
Check their youtube channel and website, the "Proactive" news channel for the latest updates on them, and when you do a search for that, be careful not to get the wrong Proactive channel, type in something like Proactive Archer Graphene.
I have the videos under my Graphene playlist, and Quantum Computing playlist, be sure to check each playlist description for articles and more info on this.
Archer Materials mentioned in a video interview, "integrating into modern electronics, for room temperature, wide - spread, allows open access to the consumer markets".
And with new materials, this is the key to many "emerging" technologies, Quantum Technologies, and so new materials change everything, like Graphene and beyond that with new materials.
The Graphene Flagship found thousands of new materials made from Graphene, recently.
Here's an article.
www.techradar.com/news/dummy-40-ways-graphene-is-about-to-change-your-life
Also the Graphene Flagship Roadmap.
graphene-flagship.eu/project/roadmap/Pages/Roadmap.aspx
Be sure to check pages back in this news link.
graphene-flagship.eu/news
Graphene Flagship Products.
graphene-flagship.eu/material/GrapheneApplicationAreas/Pages/Graphene-Products.aspx
Graphene Flagship Application Areas.
graphene-flagship.eu/material/GrapheneApplicationAreas/Pages/default.aspx
Graphene and vehicles.
th-cam.com/video/JjNrQ3wf3kM/w-d-xo.html
Skip to (2 : 02) in this video about vehicles with Graphene, and then watch it from the start.
Be sure to check out my other channel's Graphene playlist and watch all videos from top to bottom in that order, the link is located in my other comment above.
Please also check out the rest of the other channel and the "About" tab, the link is located in my other comment above.
th-cam.com/video/vlApBbhZoAE/w-d-xo.html
That melted my noodle a little also... how applicable is that other facets of our lives...
i don't have a deep understanding of Tensor flow because i wrote my own version from the ground up 👍
@@thefootballpunnedit so true brother
@@lineage13 you're a beast.
I did create a tiny Android app with TF. It was super simple.
I've always expressed this as "Architecture beats optimization"
Architecture IS optimization
@@ADogNamedElmo I can't be as optimization depends on something that already exist in order to happen.
@@paulovinicius9940 I think you misunderstood what I said
@@ADogNamedElmo Is but isn't. Architecture doesn't necessarily perform optimally and sometimes has no real point other than personalization. Although I do suppose that personalization in essence is equivalent to optimization for the individual.
Love that quote. Laying out architecture effectively is significantly more optimal than trying to optimize the hell out of something that is fundamentally imperfect
because they're praised for seeming smart, and we equate complexity with intelligence
Great way of putting it, might steal that one.
I wonder if this has anything to do with group evaluations in schools dragging people into positions they are not capable of handling...
Being smart involves taking a very complex thing and extracting a pattern from it, a.k.a. pattern recognition. So it is partially right.
@@Moreoverover ants are smart, I agree
@@nathankayhan4358 i believe that. one time i worked with this girl who used all these big words and spoke for a long time when making a simple point, i remember a couple times when i sat there knowing her point but waiting for her to stop dressing it up
00:00 start
00:38 difference from being able to follow a recipe and seeing the layers of complexity behind the recipe
01:28 When you get to be an expert at something you're hoping to achieve deeper understanding not just memorize a large set of recipes to go execute
02:15 if you constantly unpacked everything for deeper understanding you never get anything done , if you don't do deep understanding when needed, you will do the wrong thing.
03:30 the need of refactor and redesign to achieve a new local maximum since the previous one has plateaued. (every 3-5 years in a non-trivial project)
06:00 a project starts, improves and then improves through diminishing returns. Then refactor/redesign. The starting point will be lower than the past peak, but the peak will be higher.
06:30 Short term oriented business is not going to like redesign.
My wife makes this point about surgery frequently...she performs surgical procedures that didn’t exist when she was a resident.
Im curious on what you mean by this 🙂 Can you elaborate?
What?
tf 😂
you sound like a lucky man
@@yominorwemajor7235 Means don't trust surgeons words if anything goes wrong.
No beating around the bush, the practical objective reality, pure and "simple." (literally simple).
Einstein once said, everything should be as simple as possible but not simpler. Jim Keller's thinking style is just that.
Thanks to Lex and Jim Keller for the video.
This principle has a name! Occam’s razor ^^
Àl0
True.
@@DonArmadillo
I love how Jim is refering to human beings as "them"
What are you referring to?
@@vasileturus Us?
@@variator7466 I was asking more to get the timestamp. I agree we should use "us" :). God bless your mind.
@@variator7466 most people on this planet are trash, look at how they are blindly going with this propaganda rona bs virus, I don’t wanna be associated with such a dumb and logic lacking society.
@@Archonsx I hope the irony of your comment isn't lost on you :)
Lex, I admire your tenacity and strong mindset! The community is growing, keep it up with the amazing work! I'm not kidding, I consider you as a mentor, even tho we never met. I guess this is the real magic of social media.
This guy looks like a cross between Mark Hamill and Dax Shepard.
😆😆😆
Lol. I couldn't figure out why he looks so familiar. Now I know.
I see a little bit of Kiefer Sutherland in there as well
I see a little of Kevin Costner in him too
🤣🤣 dammit
Thank you for hosting this discussion and sharing it with us. I hope you can keep finding such wonderful guests.
Lex, you and your guests have an undefinable knack of explaining something in a way that the average person can understand what is being explained- without actually understanding the entire topic. That is truly amazing. Thank you
This whole conversation about starting new project at lower level than your optimization point reminds me principle of deload and progression in strength training. One step back, two steps forward.
Actually strength training was the first thing that convinced me and made me truly realize the KISS principle. Now I try to apply it for everything in my life
The "Understanding vs Recipes" is so incredibly brilliant.
3:45 the rewrite is half as complicated and takes half the time. Rewrite more often! Great tips.
I love that idea: every 5 years, do a rewrite. That’s the best way to make things better and simpler over time, even though it’s hard in the short-term.
This guy is one of the smartest guys I've ever heard
Excellent video, thanks for posting, much appreciated. Love your work Lex : )
Thank you for taking the time to do this.
There is something called "deep learning" which means that you not only memorize the "recipe" but also understand the relationship between ingredients, how they connect and affect different variables (ingredients) within the recipe.
Deep Learning is rather about building predictive models than about explanatory models.
damn this podcast really gave me some serious insight on how companies like Intel or AMD work. Not only that, you can seriously tell how smart the guy is just from how he says things confidently and how on-point his answers is. Respect for Jim, he's basically what brought us the best in computing as in today, along with his team of course.
Jim Keller's Steve Jobs quote is basically what Apple has done since the A6 SoC. The starting point of a new architecture is lower than desktop CPUs, but the ceiling is higher. For years, Apple put big powerful SoCs in their iPhones, often overkill for the applications of the day. All the while developing CPU design expertise subsidized by massive iPhone sales. Then in 2020, wham, the Apple M1 comes out, seemingly out of nowhere, and crushes Intel's best chips at lower power consumption.
Amazing interview, Jim Keller is an absolute legend! Thank you, Lex
Interesting fellow. He's right. The desire to dig down and understand what's actually happening is an attribute of true intelligence. Also, the ability to reduce things to their simplest terms. Too many people over-complicate things, believing that as a sign of intelligence. One of the best compliments I ever received was a technician that told me, "I like working on things you've programmed because the programming is easy to follow, simple, but always efficient and effective".
Sales come from making things look/sound simple and express them to people who don't understand it... that's why engineer are terrible in sales :p and psychology people great at it
I guess that’s the beauty of humbly taking it back to 1st principles
Great clip from a great podcast!
Very intriguing points: Running projects in parallel as a strategic approach
Wow going to watch this full talk. 🙏
These podcasts are great, keep em going!
Jim Keller is a great listen, glad you had him on!
If the whole software industry would listen to this man, we just might get something done.
money speaks louder than progress
This is an LGBT phobic statement, and it's probably racist as well.
yes. Frequently, money is only made if progress is controlled and conveniently postponed.
@@BeHappyTo
youtube works? no?
So... we need a new javascript framework?
One of the best videos, it’s what a lot of people have lost around the world. Details to basics. Losing what real living has to offer. Traversing through constant change rapidly. Major changes creating havoc. It happens on every level. The dangers of young not having a clear path.
Case in point: The Von Neumann Machine. It was supposed to be a placeholder architecture, implemented in the way it was ONLY because of the limitations of the technology of that time.
Fast forward to 2020.....
Taken another way, we're stuck with Ptolemaic Epicycles, until the model totally collapses and it no longer has any utility.
We need a Copernicus.
I made this comment, but i rewrote it before posting it. It was way more efficient this way, the last attempt was two lines.
People like Jim Keller are the best people to learn from. They are out in the real world DOING amazing things, figuring out problems, solutions, getting shit done. Professors, philosophers, they don't ever need to deliver something, all their ideas are conceptual, never practical. Those who cannot do, teach...
I love his no nonsense approach. You can tell his understanding is deep, he knows it all inside and out. Amazing role model for engineers.
yeah tell that to multi-billion dollar "motivational" industry..
No really although I understand how you might be led to believe that. Professors of all kinds including philosophy professor are required to deliver regularly. Publish or perish is the saying
@@ouimetco A publication is not the same thing as delivering on a real world practical need or function. There are countless publications and journals out there that don't lead to anything.
@@thepunisherxxx6804 yes and there are countless they do lead to greatness. Witness CAS9 and crispr
i love youtube for all the resource it provides. thanks for all this.
I keep re-watching this. Jim Keller is just so interesting.
Understanding the abstraction rather than relying on black box magic
What's meant by this ?
@@morgenthau5986 Understanding something that's hidden away behind some framework and/or UI instead of relying on said framework and/or UI.
Wonderful interview. This guy is brilliant!
I heard this guy on Joe Rogan now I can’t get enough 👍
I don’t see an interview with Rogan
lol everyone is searching "Joe Rogan Jim Keller" now... (I did it too)
I guess the OP meant "Joe Rogan Lex Fridman" .. Right?
Someone correct me if I'm wrong. I wanna watch both of them haha
Joe Rogan interviewed Lex Fridman. Jim Keller was not on Rogan´s show.
@@alansmith4655 I noticed that. It seems like it's the go to mic for podcasts, others like Eric Weinstein use it too
@@quantumhealing341 It's probably for the best. Joe's brain would have exploded...
unbelievably relatable. honestly unlike anything ive ever heard for how spot on this is, thank you.
blarg derp Right? Always amazes me when someone brilliant comes along and explains something so simply, eloquent and mind blowing
This was really insightful.
Congratulations on 1M Subscribers !!
David Deutsch wrote a whole book about this issue (The Beginning of Infinity). For tens of thousands of years, humans made very little progress in generating real knowledge, because they did not generate good explanations (explanations that are hard to vary) for natural phenomena, but relied on trial and error to generate rules of thumb (recipes) without any underlying understanding. Only since the scientific revolution has good explanation seeking really taken off.
great clip to share... I shared this exact timestamp already and now they get it cut perfect
It makes sense to me, since I am usely like that, but I also get hardly something done, and my hardwork gets ignored too. Gave me bad depression in crucial teen years. Though I learned to not give effs, I will hope to work hard now
Jim Keller: Humans are a mess
Me: reporting for duty Sir!
Beautifully put
This is profound. On many levels. Thank you.
Love it, this is the difference between asking how or why.
These videos are getting me hyped on the podcast
Jim is a really smart dude - cheers for this mate!
The brain likes to make things more complex than it has to be.
laziness and clutter
Buckminster Fuller said the
Same thing :
In , ' Operating Spaceship Earth '
Chapter 2 Origins of specialization . Fascinating.
It's good to keep these things in moderation.
Finding a good balance is hard.
I agree doing something well the first time is going to take way longer than doing something meh and then re-iterating,
since knowing more about the subject at hand is hard without actually tackling it in the first place.
Extremely accurate and applicable to so many areas of life. Well said.
CPU Jesus Jim Keller is a fascinating individual - the man dug ditches in college and went on the be part of teams that revolutionised personal computing.
Have him on again to give us an update on what's happening in the tech space.
Great content, been talking about exactly this lately and came up to the same conclusions!
@s__n_Ghs_w_J_g_r_v_ well ofc my good sir
Fascinating point about overcoming design plateaus and diminishing returns.
i feel so honored!! to hear a man who created my joy as a kid :)!
Amazing podcast, keep on going
This is the best TH-cam video I've watched in my life; Period! because these guys: "Gets it"!!!
Probably one of the few people on earth who knows how actually a computer is functioning from top to bottom.
Damn, this talk hit the nail on the head
great talk, fascinating insights
I would love to see have Ben Goertzel on sometime Lex! I think it would make for some interesting conversations
I think the point is that when you are building something each tech doesn't have to have deep understanding of his component, chips etc. However, the person overseeing the project has to see the bigger picture, He must understand the capabilities of each component and how they come together to accomplish the overall mission.
Searching for deep understanding has always held me back. The relationships between objects and systems has always been much more interesting to me than the objects and systems themselves. Very few people seem to care or even acknowledge that these relationships exist, they just want to tell you more facts about the object. I think the invisible nature of these relationships makes them difficult for many to perceive and easy for them to ignore.
@Thelondonbadger A good idea is only as good as the thousands of follow up ideas needed to make that idea into something tangible and fully functional.
This guy throws an easy 98 across the plate with a really unique command.
2:50 - Friedrich Hayek, nobel prize winner in economics, noted in his book "The Sensory order" that there is a fundamental limitation inherent in our brains, which is that our brain is an apparatus of classification, which means that it cannot classify anything more complex than itself. In other words you can describe the principles and patterns of our brains, but never anything in detail. There are quite simply too many variables. This is one of the primary reasons why it is impossible to predict the market in detail.
Awesome explanation, carpentry work is like that, plans anyone can build, repair is a different game.
as an amateur carpenter, I still cling to recipes, but at the margins I'll make adjustments out of necessity.
Great interview with straight answers!!!!
This is a good description of how autism works... I've got ASD. I'm really good at understanding... terrible at recipes. Most of higher education is based around teaching you lots of recipes and then testing your memorization of those recipes. So those of us with learning disabilities have HUGE problems getting degrees and certifications... but then, once out in the real world, we're often the subject matter experts in our field, despite having the lowest qualifications in the company. It's a weird situation to be in. I've tutored other students in subjects that I personally flunked out of. The class teaches you how to pass the upcoming test, not the subject itself.
A recipe/process is necessary exactly because not everyone gets to be an expert in every field due to resource constraints, so I think these two can not be compared in this way.
Back to the basics! That's my philosophy when people start complicating things.
Oh my god. This is so insightful!
Question for anyone who can shine a light: When he says people aren't thinking simple enough, what does that mean in the context of his explanation of deeper understanding vs. recipes?
Congrats on 1 million subscribers. KUTGW.
Seek to understand, not simply to learn/remember.
I was taught to never rewrite but everything he says on that point makes total sense.
I know who this guy Keller is. He's a freakin' genius. AMD benefited 40x in the last 5-8 years, basically in large part to him.
Seems the people who would be (generally) the greatest asset to an organization are those who can quickly recognize when it's appropriate to use "the recipe" and when it's best to re-engineer/re-design. Most people I know seldom look beyond the recipe.
This is actually genius and the secret to a well lived life. I will have to tell Jim.
As a software developer, I like to apply the "start over every n years" to my career. If I don't make tangible progress (salary, skill set, work/life balance etc) in a year (max 2), something has to change. Something being me/my mindset.
I also see the need for a re-write on most projects I have worked on over the years. So many places end up with monolithic systems that have grown over the years and everyone is scared to unpick it cos the guy who designed it moved on years ago. It's funny how most dev teams call them selves agile but can not/will not entertain a fundamental rethink of solutions to problems that really get simpler over time (due to experience and technology advances). I guess this decision rests with the higher ups tho.
Excellent interview ! I see that Jim is now President and CTO of "Tenstorrent" Co. I wish him the best !
Nice Video!
Wow. This is really great
This is significantly wiser than one may think
He says exactly what I have been thinking
This is sorta obvious and you can see this on brain imaging when you compare the difference between an amateur at something and a professional. When attempting a task the amateur doesnt know what their doing, they feel more anxious and excited, more regions of the brain light up, they don't know what to focus on. Professional does the same task, they know what to focus on, only a single part of the brain is lit up, as if all the energy of the brain is focused on the one pathway perfect for that task, they're calm, focused, tunnel visioned. etc.
It's not something you canjust do from jump. It takes trial and error.
1:19 I half expected him to say "...omelette sandwich. That's a good idea. I'm hungry."
watched this 8x in a row.
I didn't know who this was and so the name drop of "co-authored x86-64 spec" was like oh- now I need to watch this whole episode 😂
Love this guy!
Lex you are one of my favorite interviewers!