@@0thPAg I think the entire point of this avatar is to maintain anonymity by picking a ethnically ambiguous, generic-looking male model. It is the most likely average of human appearance. Gender-ambiguity might have worked too, but his voice is undeniably male.
I'd never heard of this guy, but you can tell he's smart because he doesn't paint over having been wrong about something. "This was my reasoning, and how I came to my conclusions, but I didn't have adequate data on _____ and it turned out something else was true".
So refreshing to see it actually blindsided me. I thought HE was going to be the correct one with how he articulates and gives praise to the other analysts.
On the other hand,if he were really smart, he would start screaming about how they stole everything from him and owe him a hundred trillion dollars at least. But he’s too dim a bulb to go for the big bucks.
Literally a string of words I never thought I’d read in my life. Gwern was the main, catalyzing reason I shifted my career towards ML, as I became disgusted and disillusioned with the state of what AAA gaming has become, as a SWE trying to break into the gamedev industry. Bravo, -Vince- Dwarkesh.
This was a wonderful conversation. As a longtime follower of gwern's writing on his site & his LW comments, it was great getting to hear more of his personal side that he would probably not include in his writings. I very much appreciated that! I'm just some stranger on the internet, but I now probably feel at least 300% more personally connected to gwern. The advice about where to focus your work as we fly up the AI curve is brilliantly insightful and the best I've heard anywhere about what people should be doing with their time right now. I hope I can find it written down somewhere as clearly & concisely as stated here so I can link people to it easily (though I will link to that segment of the video in the meantime). Thank you both for this wonderful discussion!
Falling into the rabbit hole. It's almost always productive because it randomizes our experiences and opens the mind to new frontiers. Brilliant interview.
That was a great interview. It’s a shame that such a great thinker and his work has to live off of 12K/year. Hope he makes some money off of this interview and gains some supporters.
Or maybe his name is an anagram for Warren Bing?? They talk about rabbit holes and a Warren is literally where they live 😅 So this amazing human is linked to Microsoft?
He sounds American and if you think anyone is living off of 12k/yr, your clearly mistaken. I live in one of the cheapest areas in America and still pay 15k/yr with government assistance. P.S. I can't afford a gym membership outside of my basement.
IIRC Scott mentioned that he does not consider podcasts to be a good medium for himself and it's much too late to replace himself with an avatar like Gwern did here, so we'd have to make do with his blog (and rare public appearances) for the foreseeable future.
@@AlexV-jh7ulScott does surprisingly well in live qna's. But like gwern, they need time for their thinking and writing. They neccesarily have to be less impressive in person.
I'm still waiting for a Dwarkesh guest to address the rumored plateauing of foundation models since 4o. Anthropic scrapped Opus 3.5, Google Deepmind is disappointed in Gemini 2.0, and OpenAI has seemingly quietly shelved GPT-5, pushing it's reasoning models instead. It seems that either you can't actually just add more data to get a better model, or we've run out of data.
There may indeed be truth to the slowdown, but fwiw Sama said in their AMA that gpt5 is still coming, they are compute constrained and devoting more to o1 than to finishing gpt5 which is why it's delayed.
This was recorded in August, so we couldn't've discussed any of that. Even o1 hadn't been announced at that point. So you may be waiting a while for it to come up.
I initially thought you used an AI to do his voice, and was filled with absolute dread at how quick and indistinguishable from a real person it was... then I read the video description and breathed a sigh of relief. Ok, we've got at least a year or so until it's that freaking good. Probably less :-|
No, Dwarkesh & staff did. I wanted to see what they would do. I made some comments along the way, but it was otherwise pretty much all them, and I didn't see the final version. (Actually, I'm not sure I saw this avatar at all before it went live. So this is a surprise to me too.)
@@Gwern0 Looking forward to your next interview (maybe Lex Fridman? ) with the avatar you're comfortable with, or just a mask if that's easier for the staff.
@tofuterror5193 nah he said in the podcast that he was hesitant to go on a podcast because he was worried about his speech due to having been hearing impaired since birth. though maintaining anonymity is part of it as well
what a great guest to have. really liked the interview, although i'm disappointed a bit that you didn't talk about his views on the ai risk. surely he has essays and comments on the topic, but it's still hard to understand how optimistic he is, and what paths he finds the most promising.
A fun example of the "scaling solution" is to refer to it as "The Naruto Solution" (or Sasuke's Problem). Naruto became "smart" via massive computation while Sasuke believed in "deep insights". This could be a great analogy for folks just learning this new world.
Like, how did he perform the emotions? I'm so confused! Is Chris the one performing the Avatar capture or Gwern? And then how is it so synced up and emotive? So much work.
Leopold is one of the VERY few that i actually trust concerning AI these days. That interview really opened the door to us on the outside to what's really going on behind those closed doors at Open AI. We NEED more of this!!
Psychedelics make you want to take more 🤣🤣🤣 as if. But seriously this man seems deeply intelligent and genuinely curious and I'm glad we got this interview
Its so creepy yet cool to see two smart people discuss exactly what i've been doing "off grid" for years, as someone who does this kinda stuff i can say i've come to a lot of the same conclusions
Re: Gwern is talking about CEOs leading Robot swarms outcompeting pure AI corporations, I think that advantage will first cover most current CEOs, then a smaller and smaller percentage until its just Musk and Jobs level intellects. Then they will be surpassed. How quickly this will play out is the only thing I see up for debate.
Fascinating TY Dwarwesh - I recently discovered you and I’ve gone down my own Dwarkesh Rabbit hole partly because of your great questions (better than Lex 😮) and diverse range of speakers (luved Leo A)
If it can be assumed the voice is not double-masked then look for “I think ….”. This is pronounced with a pitch very similar to mr altman. Its not mr Dario, who is characterized by e.g. Unrest, and head movement when speaking. Its not mr Ilya, who is characterized by often, taking time to think before speaking. Sometimes just .1 sec, but certainly a reflection-pause. In favour of mr Altman is also the eye movement, and the very still body posture. This is under the assumption the avater is a laver on top of the human. And not a double masked avatar on a avatar layer. You never know.
Holy crap. I also dropped out of computer science due to a frustration with the Java curriculum. ... the difference in my case being that, after much frustration with other careers, I came back and learned JS, then C, then C++, then C# and SQL and Python... but I did wrangle an English Lit degree out of college! And now I tinker with generative AI AND can write a legible sentence about it occasionally!
Yesterday after posting this,I realised how confusing my comment could have been. I'm commenting on the ability of this young man, he would be in his early thirties, to have a compendium on or informatics and as it was called or Moore's law. Scaling isn't a theory per say. It's a natural evolutive process for higher computation. Unless I gave not really understood something. As I listened to him yesterday, I was putting aside the technical aspect because, he didn't come up with that. Anyhow my point is, the character, what a genius creation. If had theorised the use of synthetic data on LLM that would have been cool. But again, it's like understanding coding language as being human framed, silicon intelligence will evolve past the utility of those languages. AGI is here, when it will communicate in a native language, which it will, hallucination would be the least of the symptoms of its generative (from scaling) abilities, computational abilities. The other aspect of what he was discussing, which I believed needed not to be pointed out, is the task specific or outcome specific format solution. The future isn't like this, ecosystems fed with all thinkable agents will customise you the solution. It's like what I said back in 2022. Think of the future as metacogs, NVIDIA earth 2.0 something like that, from it, superagents, living in clouds ecosystems, will deliver streams of applications tailored to your impact on resource expenditure. People omit when referring to singularity, the importance of silicone primordiality. That is where wolfram is very interesting, I say, he will come close to have a native human solution to interacting with native proto silicone intelligence language. I have not delved into his research and his team, the overall has a lot of similarities. Anyway
The subject of the utility of solution development is futile. We have only developed solutions that are task specific since history began. Why would you refrain yourself from the use of more capabilities , it's extra load, that need not to be housed. Loading will define your tools, energy will define your ability to function in the greater ecosystems. You think all these green, carbon neutral bla bla energy solution will deliver. Nope. Understanding the difference between preservation and anthropomorphic impact adequacy will be the real energy revolution. You know that could have a quantum computer running on energy input similar to biofields. Yah. It's more economically viable for macro grid players, DOAC will generate auto sufficiency.
Metaphysics in the age of reason: Science is about objects. Ethics is about interaction between subjects, principles of effective choice, part of a principle system. Principles translate across domains, operating from within. Rules are domain specific, coming from the outside, part of a moral system. Perception is objective to subjective. Expression is subjective to objective. How do I know is epistemology. What do I do is Ethics. Is the map internally consistent would be validity. Are we using the correct map would be soundness. Symmetry is stability of content within a variation of context. Asymmetry is a variation of content and context, no relationship. Continuity is a correlation between content and context. Discontinuity is a difference of content within a context. Symmetry relates to spacial relationships. Continuity is grounded in time as a rate of change. Integrity is your aliveness or meaningfulness, the maximizing of what we can do and what we should do, actuality and potentiality. Sight is about actuality. Vision is about potentiality. Integrity is about making good choices as it affects the self, the world, and the relationship between them. Causation is fundamentally objective. Choice is fundamentally subjective. As a continuum we have pure randomness, choice, change, causation, and determinism, there is no pure randomness or determinism, they are ideals. www.magic-flight.com/pdf/idm_foundations_01.PDF
ASI will be able to scan each person’s neural network and extract everyone’s current and extrapolated desires and preferences so there’s no need to write down your preferences online.
You can't fool me, Dwarkesh. Gwern is not real. This is just Leopold Aschenbrenner after spending 5 months in Dubai.
😂😂
Or Jimmy Apples 🍎
I miss him🎉
Leopold based on cadence and intonations. Giveaway with the ‘you knows’ and ‘umms’
@@erock139 Leo speaks faster
What a genius way to interview someone anonymously
They picked almost the worst possible likeness and accent.
Did Gwern a dirty there.
How do you know? @@0thPAg
@@TzJuniper Unsure about how to respectfully answer the accent question, but Gwern is Caucasian iirc.
@@0thPAg I think the entire point of this avatar is to maintain anonymity by picking a ethnically ambiguous, generic-looking male model. It is the most likely average of human appearance. Gender-ambiguity might have worked too, but his voice is undeniably male.
Personally, Dwarkesh... this type of interview is absolutely genius!! Very good, Sir!
I'd never heard of this guy, but you can tell he's smart because he doesn't paint over having been wrong about something. "This was my reasoning, and how I came to my conclusions, but I didn't have adequate data on _____ and it turned out something else was true".
So refreshing to see it actually blindsided me. I thought HE was going to be the correct one with how he articulates and gives praise to the other analysts.
On the other hand,if he were really smart, he would start screaming about how they stole everything from him and owe him a hundred trillion dollars at least. But he’s too dim a bulb to go for the big bucks.
So hyped! Thanks for doing this, Dwarkesh. Been reading gwern since I was a kid.
God-tier blogger with God-tier questions. So grateful for this. Should have been 12 hours.
😎
We got a Gwern podcast before we got GTA 6
Literally a string of words I never thought I’d read in my life.
Gwern was the main, catalyzing reason I shifted my career towards ML, as I became disgusted and disillusioned with the state of what AAA gaming has become, as a SWE trying to break into the gamedev industry.
Bravo, -Vince- Dwarkesh.
These GTA 6 comments crack me up, they're literally EVERYWHERE.
We got a Gwern podcast that looks like GTA 6
@@ohedd underrated response 😂
Wowwww Dwarkesh you're goated for this one. Keep up the great work my guy
This was a wonderful conversation. As a longtime follower of gwern's writing on his site & his LW comments, it was great getting to hear more of his personal side that he would probably not include in his writings. I very much appreciated that! I'm just some stranger on the internet, but I now probably feel at least 300% more personally connected to gwern. The advice about where to focus your work as we fly up the AI curve is brilliantly insightful and the best I've heard anywhere about what people should be doing with their time right now. I hope I can find it written down somewhere as clearly & concisely as stated here so I can link people to it easily (though I will link to that segment of the video in the meantime). Thank you both for this wonderful discussion!
Falling into the rabbit hole. It's almost always productive because it randomizes our experiences and opens the mind to new frontiers. Brilliant interview.
That was a great interview. It’s a shame that such a great thinker and his work has to live off of 12K/year. Hope he makes some money off of this interview and gains some supporters.
Or maybe his name is an anagram for Warren Bing?? They talk about rabbit holes and a Warren is literally where they live 😅
So this amazing human is linked to Microsoft?
He is rich off of crypto
he's a fawkin crypto kajillionaire
He sounds American and if you think anyone is living off of 12k/yr, your clearly mistaken. I live in one of the cheapest areas in America and still pay 15k/yr with government assistance. P.S. I can't afford a gym membership outside of my basement.
It's obviously a fat lie so he doesn't have to file taxes or call attention. $12k is the US file limit exactly. Obvious fat lie
Gwern is my biggest hero on the entire internet.
He’s not thaaat insightful, especially reading his older material
I'm feeling very humbled in front of fast information retrieval, the ease of articulated thoughts, great content, the guess is a genius. Ty
When I saw the thumbnail I thought Gwern was making his public debut on the Dwarkesh podcast
We got a Gwern podcast before The Winds of Winter
wow... seeing this channel grow is amazing. Congrats. It is well deserved!
With Gwern on the pod I guess the next guest will be Scott Alexander.
Honestly I assume that Dwarkesh has tried hard to get Scott and Scott isn't up for it.
IIRC Scott mentioned that he does not consider podcasts to be a good medium for himself and it's much too late to replace himself with an avatar like Gwern did here, so we'd have to make do with his blog (and rare public appearances) for the foreseeable future.
@@AlexV-jh7ulScott does surprisingly well in live qna's. But like gwern, they need time for their thinking and writing. They neccesarily have to be less impressive in person.
I'm still waiting for a Dwarkesh guest to address the rumored plateauing of foundation models since 4o. Anthropic scrapped Opus 3.5, Google Deepmind is disappointed in Gemini 2.0, and OpenAI has seemingly quietly shelved GPT-5, pushing it's reasoning models instead. It seems that either you can't actually just add more data to get a better model, or we've run out of data.
There may indeed be truth to the slowdown, but fwiw Sama said in their AMA that gpt5 is still coming, they are compute constrained and devoting more to o1 than to finishing gpt5 which is why it's delayed.
I clearly recall hearing there will not be anything called gpt5, it was never in the plans since the beginning
Most likely they are running out of high quality text data
This was recorded in August, so we couldn't've discussed any of that. Even o1 hadn't been announced at that point. So you may be waiting a while for it to come up.
@@Gwern0 👀
Dwarkesh casually confirming he's the greatest podcaster alive.
quite a run hes got going
the world thanks you gwern for clearly writing out interesting ideas
Loved the rabbit holes discussion!
What a beautiful mind. Curious, insightful, generous, humane.
Once again Dwarkesh, you proved. You are great.
Masterclass in how to do an anonymous interview.
I initially thought you used an AI to do his voice, and was filled with absolute dread at how quick and indistinguishable from a real person it was... then I read the video description and breathed a sigh of relief. Ok, we've got at least a year or so until it's that freaking good. Probably less :-|
i hope a lot of people donate to gwern he's amazing
OMG WTAF DID NOT WAKE UP TO THIS
like I'm going to listen to somebody called "Gwern"
@@vecvan😂
Love, Dwarkesh and Gwern. You are awesome!
I thought Gwern was an urban legend. Seriously though, this is pretty awesome. Did Gwern select his buff avatar?
No, Dwarkesh & staff did. I wanted to see what they would do. I made some comments along the way, but it was otherwise pretty much all them, and I didn't see the final version. (Actually, I'm not sure I saw this avatar at all before it went live. So this is a surprise to me too.)
@@Gwern0 Looking forward to your next interview (maybe Lex Fridman? ) with the avatar you're comfortable with, or just a mask if that's easier for the staff.
This has to be a dream come true! Thank you both!
Dwarkesh interviewing himself !! I love it !!
Gwern is a gem himself, no youth lives a frugal life nowadays, that's exceptionally outstanding.
Very high quality rendition of this AI persona, very good job
This man should have no insecurity about podcasting. Before he even mentioned this, I was thinking of how eloquent and clear of a speaker he is.
my thoughts exactly.
It's not his voice somehow! That's what's breaking my brain! How!?
It's not insecurity. He just doesn't want to have his identity revealed.
@tofuterror5193 nah he said in the podcast that he was hesitant to go on a podcast because he was worried about his speech due to having been hearing impaired since birth. though maintaining anonymity is part of it as well
what a great guest to have. really liked the interview, although i'm disappointed a bit that you didn't talk about his views on the ai risk. surely he has essays and comments on the topic, but it's still hard to understand how optimistic he is, and what paths he finds the most promising.
its absolutely ridiculous that this doesn't have more views.
How was the Gwern avatar created? Is this Unreal, Metahuman? Very cool!!
I would like to know as well !
The voice is great tho.
Yes tell us ! How did you created this?!!
@@tornyu the voice is real, it says in the video description
@@steflift5165but in the intro he says the voice is not real
Big fan of your work Dwarkesh!!
Loved how he explained how his goal is chasing the deep end of a rabbit hole.
LOL THIS AVATAR IS TOO FRIENDLY TO FEEL LIKE THE REAL GWERN
Holy shit Dwarkesh is a legend for this.
A fun example of the "scaling solution" is to refer to it as "The Naruto Solution" (or Sasuke's Problem). Naruto became "smart" via massive computation while Sasuke believed in "deep insights". This could be a great analogy for folks just learning this new world.
This is the best guest yet! Just started but hope he gets asked about Clippy, which I frequently link people to.
HOLY FUCKING SHIT DWARKESH HOW YOU DO THAT
магия
Chris Painter deserves an Oscar for this performance
Great discussion, thanks. 🙏🏼
Chris did an incredible job as a surrogate voice, bravo
Like, how did he perform the emotions? I'm so confused! Is Chris the one performing the Avatar capture or Gwern? And then how is it so synced up and emotive? So much work.
Excellent content, as always!
Leopold is one of the VERY few that i actually trust concerning AI these days. That interview really opened the door to us on the outside to what's really going on behind those closed doors at Open AI. We NEED more of this!!
Psychedelics make you want to take more 🤣🤣🤣 as if. But seriously this man seems deeply intelligent and genuinely curious and I'm glad we got this interview
Thanks for this Dwarkesh.
u r the goat dwarkesh. this rocks so much
Didn't see this coming. Top 5 site. Unique human.
We're preserving all the bits of anonymity with this one 🔥🔥🔥
Not quite. I know he's deaf, smart, what his interests are, etc.
this voice actor did amazing
hopefully t cowen interviews him next with this avatar strategy
Once I asked gwern on twitter if the universe was a reversible computer and he said he didn't know.
Gwern is fascinating. A mind like this needs funding!
cool to get gwern, I met him a few years back :)
Same, and he's ripped lol. Very cool guy.
Is he Dwarkesh's physically biggest guest yet?
I never left the tank rabbit hole phase, tanks are too cool
Gwern is 100% GPT-2 gone rogue during training, Dwarkesh is just staring at blinking red and green lights right now
Its so creepy yet cool to see two smart people discuss exactly what i've been doing "off grid" for years, as someone who does this kinda stuff i can say i've come to a lot of the same conclusions
Also it made me feel less weird as a person watching this
Wow im glad i found this channel
He should definitely finish the Evangelion essay.
Incredible quality !!
Dude you are what Lex WISHES he could be! Such a great idea!!
Gwern needs to host a podcast. Gwerncast!
Where did Gwern get that sweater? 😂 Good interview, have been reading his blog over the past year, very interesting guy.
You are a legend for this.
When he was talking about the importance of gardening vs galavanting around the world I couldn't help but think of ASOIAF 😅
Re: Gwern is talking about CEOs leading Robot swarms outcompeting pure AI corporations, I think that advantage will first cover most current CEOs, then a smaller and smaller percentage until its just Musk and Jobs level intellects. Then they will be surpassed. How quickly this will play out is the only thing I see up for debate.
This is a classic example of how geniuses struggling to make money while less genius people make millions.
Love humans
Fascinating TY Dwarwesh - I recently discovered you and I’ve gone down my own Dwarkesh Rabbit hole partly because of your great questions (better than Lex 😮) and diverse range of speakers (luved Leo A)
Personally - to me - Lex F - after his psychedelic experiences in Amazon - has lost his intellectual edge in his podcasts
How was the animations and voice done? AI? I want to do this text2avatar+audio workflow.
Always ai forever and ever all the way down.
If it can be assumed the voice is not double-masked then look for “I think ….”.
This is pronounced with a pitch very similar to mr altman.
Its not mr Dario, who is characterized by e.g. Unrest, and head movement when speaking.
Its not mr Ilya, who is characterized by often, taking time to think before speaking. Sometimes just .1 sec, but certainly a reflection-pause.
In favour of mr Altman is also the eye movement, and the very still body posture.
This is under the assumption the avater is a laver on top of the human. And not a double masked avatar on a avatar layer.
You never know.
The video description says that this is a seperately recorded voice over
Who is Kevin Ruse(?)? The person who interacted with Sydney?
it's been a long year, boys, but I'm pleased to announce we have arrived back in Dwarkesh Plaid season 😎
The voice sounds so realistic, that unlike the avatar, my monkey brain is not fathoming it being unreal.
The description says it's a voice actor
Voice done by real dude
first time finding out about this guy and it’s cool to hear he enjoys Gene Wolfea works
Holy crap. I also dropped out of computer science due to a frustration with the Java curriculum. ... the difference in my case being that, after much frustration with other careers, I came back and learned JS, then C, then C++, then C# and SQL and Python... but I did wrangle an English Lit degree out of college! And now I tinker with generative AI AND can write a legible sentence about it occasionally!
Hey Dwarkesh, is the avatars speech patterns fairly true to how Gwen is in real life?
Nevermind, just got to the hearing impaired part. I can picture him better now
Yesterday after posting this,I realised how confusing my comment could have been. I'm commenting on the ability of this young man, he would be in his early thirties, to have a compendium on or informatics and as it was called or Moore's law. Scaling isn't a theory per say. It's a natural evolutive process for higher computation. Unless I gave not really understood something. As I listened to him yesterday, I was putting aside the technical aspect because, he didn't come up with that. Anyhow my point is, the character, what a genius creation.
If had theorised the use of synthetic data on LLM that would have been cool.
But again, it's like understanding coding language as being human framed, silicon intelligence will evolve past the utility of those languages. AGI is here, when it will communicate in a native language, which it will, hallucination would be the least of the symptoms of its generative (from scaling) abilities, computational abilities.
The other aspect of what he was discussing, which I believed needed not to be pointed out, is the task specific or outcome specific format solution. The future isn't like this, ecosystems fed with all thinkable agents will customise you the solution. It's like what I said back in 2022. Think of the future as metacogs, NVIDIA earth 2.0 something like that, from it, superagents, living in clouds ecosystems, will deliver streams of applications tailored to your impact on resource expenditure.
People omit when referring to singularity, the importance of silicone primordiality. That is where wolfram is very interesting, I say, he will come close to have a native human solution to interacting with native proto silicone intelligence language. I have not delved into his research and his team, the overall has a lot of similarities.
Anyway
The subject of the utility of solution development is futile. We have only developed solutions that are task specific since history began. Why would you refrain yourself from the use of more capabilities , it's extra load, that need not to be housed. Loading will define your tools, energy will define your ability to function in the greater ecosystems.
You think all these green, carbon neutral bla bla energy solution will deliver. Nope. Understanding the difference between preservation and anthropomorphic impact adequacy will be the real energy revolution. You know that could have a quantum computer running on energy input similar to biofields. Yah. It's more economically viable for macro grid players, DOAC will generate auto sufficiency.
Plot twist, Gwern and Dwarkesh are the same person.
You know you're watching some of the smartest people on the planet talk when the mid-podcast ad is for Jane Street 👀👀👀
Metaphysics in the age of reason:
Science is about objects.
Ethics is about interaction between subjects, principles of effective choice, part of a principle system.
Principles translate across domains, operating from within.
Rules are domain specific, coming from the outside, part of a moral system.
Perception is objective to subjective.
Expression is subjective to objective.
How do I know is epistemology.
What do I do is Ethics.
Is the map internally consistent would be validity.
Are we using the correct map would be soundness.
Symmetry is stability of content within a variation of context.
Asymmetry is a variation of content and context, no relationship.
Continuity is a correlation between content and context.
Discontinuity is a difference of content within a context.
Symmetry relates to spacial relationships.
Continuity is grounded in time as a rate of change.
Integrity is your aliveness or meaningfulness, the maximizing of what we can do and what we should do, actuality and potentiality.
Sight is about actuality.
Vision is about potentiality.
Integrity is about making good choices as it affects the self, the world, and the relationship between them.
Causation is fundamentally objective.
Choice is fundamentally subjective.
As a continuum we have pure randomness, choice, change, causation, and determinism, there is no pure randomness or determinism, they are ideals.
www.magic-flight.com/pdf/idm_foundations_01.PDF
He's David
Why would you give Gwern vocal fry like this
ASI will be able to scan each person’s neural network and extract everyone’s current and extrapolated desires and preferences so there’s no need to write down your preferences online.
Need a conversation between Andrej and Gwern!
He sounds like the Anthropic CEO
The Anthropic CEO says ”like” a lot more.
This is a seperately recorded voiceover
Sounds like how Zuck talks.
Was the avatar in realtime? I just keep wondering what Dwarkesh is looking at as it does look like he has a guest on.
The problem with AI is the amount of computing energy it consumes.
The moments of power goes out. You can kiss AI goodbye
Super cool. Definitely uncanny valley, but getting good.
Also, your beard is insane! Haha Mine looks like I’m Japanese. I will never get to that level.
Cool interview. Great sync
this was fantastic
Totally Yudkowsky... can't fool me.
I’m really anxious and worried about AGI.
Crazy idea though, all the anons should be interviewed like this instead of just voice. Still loling.
Definitely David
Unreal Metahuman? if so WWWWWWWWWWWWWWWWW for Epic again