I think that isn't quite right though. There's another factor at play which is the pareto relationship between the skills and knowledge you need and the time takes to get them: 80% of the most important information and skills can be acquired in probably 20% of the time. And then sometimes that final 20% of skills either isn't necessary or can be bridged in a different way or you develop new skills and knowledge entirely which the top people do not have.
The CEO of AI21 said, “there are perhaps 200 people in the world who are simultaneously excellent researchers and engineers.” And this is why there are only a few companies with the capacity to build SOTA models.
I've been following his blog for years up until he got hired at anthropic and even though I knew he was very impressive I didn't know the scale of his brilliance. I use his blog as a way to get motivation to work/code 😂
As an engineer I can say that vision, enthusiasm, learning attitude, ownership and empathy is also a quality which distinguishes an engineer. Vision, enthusiasm and learning attitude is pretty personal. Empathy for the every stakeholder which includes yourself, fellow engineers, PMs, leadership and customers. What it means is that you write better code which could understood by you and any one reading it. Adding proper logs, telemetry which would be helpful in debugging and reporting numbers to PMs, leadership and also helps analyze the product feedback. This video, I have bookmarked as it is a great video to understand the mindset of great engineers.
My guess is that mathematical maturity takes the longest to build, as does learning general principles of research. Once you're good at both you can get to the top pretty fast by only focusing on what's driving the progress.
@@takamotoyagami4222 I'm assuming he means that the goal of Q* is to understand how to conduct research and get good at maths, in the hope that scaling that will result in a world-class AI researcher as an emergent property
I desperately want to be a part of the AI revolution, but it's amazing how much depends on having time and focus. I'm a single mom, and it's excruciatingly painful how little Flow time I have for myself. (I've programmed since 1979, focusing on graphics and rendering, dabbling in CV. I loved noneuclidian geometries since I was a teen. The linear algebra skills - particularly visualizing at higher dimensions than 3 - comes naturally to me. The first model I trained was so exciting and fun! I'm desperate to go farther. It's depressing that I can't contribute my gifts and hone my abilities. So jealous of these men - not for their abilities, I have those, but the privileges of free time and uninterrupted focus.)
As (relatively) new father (who also derailed his career a few years ago) I deeply empathise with what you're saying. I am in pretty much a similar boat. Young child, lots of stuff to do around the house, other half is main bread winner. Of course, much harder for a single parent. I've come to accept that I had my chance of being young and independent and now I'm grateful the hours i get to try to get back into something like this.
I'm top of my field in my profession, literally can't go higher without moving completely into full time leadership...that said, i'm not a fan of talking down others as lacking drive or initiative and hence that's how you did well. Because you can easily spin that around as failed leadership. Not everyone is a 22 year old kid with nothing else to do, some are parents, others struggle with personal issues, others simply aren't motivated...but could be.
I loved Sholto's remark about mentorship. It is life-changing. I'm on the lookout for a mentor in AI research. If you have any recommendations or might be open to mentoring, let me know. Highly appreciated!
Next level of growth. Are you in? 1. CodeCraft Duel: Super Agent Showdown 2. Pixel Pioneers: Super Agent AI Clash 3. Digital Duel: LLM Super Agents Battle 4. Byte Battle Royale: Dueling LLM Agents 5. AI Code Clash: Super Agent Showdown 6. CodeCraft Combat: Super Agent Edition 7. Digital Duel: Super Agent AI Battle 8. Pixel Pioneers: LLM Super Agent Showdown 9. Byte Battle Royale: Super Agent AI Combat 10. AI Code Clash: Dueling Super Agents Edition
Or hes really excited about what hes doing. Doing adderall is not conducive to productive software development. Eventually your brain shuts down and you just start writing nonsense.
Indeed. Neither of them has done anything original in the field of artificial intelligence, or even relevant. There are people who are lucky in life. These 2 are an example of this.
@@SholtoDouglas-w6i you did great! I had a mentor that used to say that when somebody is doing something new is deserving just clapping for 6 times and feedback for improvement only from the 7th on. I was inspired by the 3 of you. Thanks for showing up.
"It's amazing how quickly you can be a world-class in something cause most people aren't trying that hard"
really inspiring word!!
I think that isn't quite right though. There's another factor at play which is the pareto relationship between the skills and knowledge you need and the time takes to get them: 80% of the most important information and skills can be acquired in probably 20% of the time.
And then sometimes that final 20% of skills either isn't necessary or can be bridged in a different way or you develop new skills and knowledge entirely which the top people do not have.
So they were outstanding software engineers who put in a ton of work on a new field.. not surprised but they inspire me the fck out
They were lucky to be on the cutting edge of the hyped Gen-AI field. Some researchers have been doing research for 30 years unnoticed.
i think the issue is doing it alone , anthropic is a team....Genius researcher dont wanna collaborate
The CEO of AI21 said, “there are perhaps 200 people in the world who are simultaneously excellent researchers and engineers.” And this is why there are only a few companies with the capacity to build SOTA models.
Why so few?
I KNEW Tristan Hume will be mentioned somewhere. Big fan of his work. Dwarkesh, if you could please talk to Tristan on the pod, we will be grateful 🙏
Not familiar with Tristan's work but have met him a few times and he's a cool dude
I've been following his blog for years up until he got hired at anthropic and even though I knew he was very impressive I didn't know the scale of his brilliance. I use his blog as a way to get motivation to work/code 😂
As an engineer I can say that vision, enthusiasm, learning attitude, ownership and empathy is also a quality which distinguishes an engineer. Vision, enthusiasm and learning attitude is pretty personal.
Empathy for the every stakeholder which includes yourself, fellow engineers, PMs, leadership and customers. What it means is that you write better code which could understood by you and any one reading it. Adding proper logs, telemetry which would be helpful in debugging and reporting numbers to PMs, leadership and also helps analyze the product feedback.
This video, I have bookmarked as it is a great video to understand the mindset of great engineers.
My guess is that mathematical maturity takes the longest to build, as does learning general principles of research. Once you're good at both you can get to the top pretty fast by only focusing on what's driving the progress.
That's the alleged goal of Q*
what do you mean by that?
@@whatsanimesh
@@takamotoyagami4222 I'm assuming he means that the goal of Q* is to understand how to conduct research and get good at maths, in the hope that scaling that will result in a world-class AI researcher as an emergent property
this, the limiting factor almost always comes down to ones core understanding of the math
@@vishaljain4915cause existing is math
I desperately want to be a part of the AI revolution, but it's amazing how much depends on having time and focus. I'm a single mom, and it's excruciatingly painful how little Flow time I have for myself.
(I've programmed since 1979, focusing on graphics and rendering, dabbling in CV. I loved noneuclidian geometries since I was a teen. The linear algebra skills - particularly visualizing at higher dimensions than 3 - comes naturally to me. The first model I trained was so exciting and fun! I'm desperate to go farther. It's depressing that I can't contribute my gifts and hone my abilities. So jealous of these men - not for their abilities, I have those, but the privileges of free time and uninterrupted focus.)
As (relatively) new father (who also derailed his career a few years ago) I deeply empathise with what you're saying. I am in pretty much a similar boat. Young child, lots of stuff to do around the house, other half is main bread winner. Of course, much harder for a single parent.
I've come to accept that I had my chance of being young and independent and now I'm grateful the hours i get to try to get back into something like this.
"manufacture luck". makes so much sense
Great chat & insightful comments
Very inspired by their story, and I have to say this is among the top content that I have seen recently
I'm top of my field in my profession, literally can't go higher without moving completely into full time leadership...that said, i'm not a fan of talking down others as lacking drive or initiative and hence that's how you did well. Because you can easily spin that around as failed leadership. Not everyone is a 22 year old kid with nothing else to do, some are parents, others struggle with personal issues, others simply aren't motivated...but could be.
I loved Sholto's remark about mentorship. It is life-changing. I'm on the lookout for a mentor in AI research. If you have any recommendations or might be open to mentoring, let me know. Highly appreciated!
If you go ham [in a niche field].
Put these two next to connor leahy and let them have a thorough discussion!
this defiinitely is a good interview, just honest personal advise
Next level of growth. Are you in?
1. CodeCraft Duel: Super Agent Showdown
2. Pixel Pioneers: Super Agent AI Clash
3. Digital Duel: LLM Super Agents Battle
4. Byte Battle Royale: Dueling LLM Agents
5. AI Code Clash: Super Agent Showdown
6. CodeCraft Combat: Super Agent Edition
7. Digital Duel: Super Agent AI Battle
8. Pixel Pioneers: LLM Super Agent Showdown
9. Byte Battle Royale: Super Agent AI Combat
10. AI Code Clash: Dueling Super Agents Edition
Very insightful
I really love this podcast it speak to me a lot... It's the best
Very insightful and inspiring** video. lol my only issue is that quote wasn't from LeBron, it was Kobe who said that :)
That is an amazing shirt
it is a cool shirt, right?
Wtf they’re good looking too. That’s not fair
life isnt fair, stop bein gay
Rapid fire testing
No salt, but they're not leading AI researchers though, right?
Who is the one on thw White t shirt?
10pm to 2am interesting
I’m just going to hazard a wild guess here, but he probably wasn’t also working at a nearby bar to pay his rent.
Ya, but at the same time, thousands of people who also weren't working at the bar to pay the rent are spending money at the bar .
@@kodreaming This is highly relevant.
Just because you're working on the trendiest thing, doesn't mean you're better than any other software engineer out there
Yeah I don't know why everyone is hyping this up in the comments. Extremely misplaced opinions
Ai is not yet ai...
why do i get the vibe he's on adderall
coke
or noo, i wront, he is on lemonade and chips like all ml nerds)
Or hes really excited about what hes doing.
Doing adderall is not conducive to productive software development. Eventually your brain shuts down and you just start writing nonsense.
that’s what AI means. Adderall Intelligence
bc they all are
Too bad Claude is so bad at math compared to chatgpt with Wolfram can not even do latex properly
So they got lucky is all
Indeed. Neither of them has done anything original in the field of artificial intelligence, or even relevant. There are people who are lucky in life. These 2 are an example of this.
@@raul36 Can't believe this sh*t showed up on my feed ffs.
probably bullshit.. america always needs heros and a posterchild for something. who is really going to challenge whether they don't know AI.
Both of these people would sound much more intelligent, if they didn't say "like" every 5th or 6th word. It's not difficult!
I genuinely didn't know how much I did this until the pod - and I'm now extremely watchful for it. Trying!!
@@SholtoDouglas-w6i you did great! I had a mentor that used to say that when somebody is doing something new is deserving just clapping for 6 times and feedback for improvement only from the 7th on. I was inspired by the 3 of you. Thanks for showing up.
@@SholtoDouglas-w6iwhat advice would you give to someone who is starting a grad program in ML research
Because that is so important 🙄
the thing is they are way more intelligent AND successful than you so they can say whatever they like ;)