Sam Altman's been very busy.

It's not quite the things that we were looking for.

We're expecting OpenAI and Microsoft went big.

Elon Musk and Mark Zuckerberg, well, this may not be great for them.

Meanwhile, OpenAI loses some employees over some concerns while Elon Musk subpoenas the board member that orchestrated Sam Altman's firing.

Things are heating up.

Let's dive in.

And just as I'm getting ready to publish this video, on top of everything else you'll hear today, Reuters publishes that Apple renews talks with OpenAI for iPhone Genitive AI features, according to Bloomberg News.

They wanna use OpenAI's technology to power some new features that are being introduced to the iPhone later this year.

OpenAI could potentially be integrated into the next iPhone operating system, the iOS 18.

It looks like they're also considering Google's Gemini.

Apple scrapped its car project and pivoted to AI, based on some of the research that's been coming out.

The results are a bit of a mixed bag.

We've looked at Apple's research into their AI agent system called Roam.

I'll link the video down below if you wanna see it.

It's a tiny model that's capable of operating on device, on the iPhone, for example, and able to understand the various references and context that appears on screen so that you're able to interact with this AI through the phone itself.

It doesn't have to be hosted in the cloud.

You don't have to be connected to the internet.

But, of course, the more advanced chatbot capabilities, the higher reasoning, the more complicated tasks, you do need a much stronger agent.

And while the gap between GPT-4 and the other competitors has grown much smaller, there are rumors that some companies are getting an advanced look at the next generation model, whether that's called GPT-5 or something else we don't know, but that's what they're referring to it as, saying that the GPT-5 will be better at reasoning and have a lot more agentic capability.

Nothing is confirmed yet.

There's no agreement that's been reached as far as we know, but it does seem like this is gonna be Altman's year.

As I'm recording today's video, this hits my newsfeed.

While it doesn't change anything that we've talked about, it certainly solidifies the horse that we're on.

This morning, the Department of Homeland Security announces the establishment of the Artificial Intelligence Safety and Security Board.

Members include Sam Altman, Dario Amadei, Jensen Huang, Satya Nadella, Sundar Pichai, and many others.

That's, of course, OpenAI, Anthropic, NVIDIA, Microsoft, and Google.

I believe I also saw Adobe on there.

We have Fei-Fei Li, a very well-regarded and respected researcher in AI.

We have people from Adobe.

Notably, the two people missing from the list are Mark Zuckerberg and Elon Musk.

The intrigue, as they say, Mallorca said he deliberately chose not to include social media companies, including Meta and X in the group, though these companies also have substantial AI operations.

The other thing about Meta and X, or Zuckerberg and Musk, is that they both open-sourced powerful AI models.

Grok and LLaMA 3, very powerful open-source, and there is still a lot of discussion about whether or not open-source AI models are a potential danger.

People are complaining that these people have a shared objective of creating a moat around AI by outlawing open-source AI.

Certainly that helps the bottom line of companies like Google and Microsoft specifically, OpenAI, and Delta Airlines is on there as well, for some reason.

There's a number of OpenAI employees or board members that are back in the spotlight.

I guess let's start with Roon.

Roon has been kind of vaguely referred to as some employee at OpenAI, potentially somebody that's working on pretty big things.

It's all very secretive.

At some point, Sam Altman actually posted, also Roon is my alt, basically implying it's his secondary anonymous account, but Roon has denied those allegations saying I'm the main.

The takeaway here is that people at OpenAI enjoy trolling people, which I cannot blame them for.

But recently it got a little bit more weird because as we hear from Sam Altman a little bit less now, Roon's posts and kind of stepped up in intensity.

He's hinting that big things are coming.

Here he's saying, I don't care what line the labs are pushing, but the models are alive, intelligent, entire alien creatures and ecosystems, and calling them tools is insufficient.

They are tools in the sense that a civilization is a tool.

However, soon after that, Roon unplugs his account, shutting it down.

Drama ensues, people don't know what's happening.

Roon did clarify that he's just shutting the account down for a while, nothing is wrong.

He just needs to detox from the site a bit.

Meanwhile, Jensen Huang himself delivers the first NVIDIA DGH200 in the world, hand-delivered to OpenAI and dedicated by Jensen to advance AI computing and humanity.

One little known technical fact about these chips is that the ones signed by Jensen Huang himself run 20% faster.

I don't have a source on that, but it seems to check out.

Meanwhile, Sam Altman is over at Stanford in the NVIDIA auditorium, interestingly enough, talking about some new things that are coming down the pipe.

A few things that are announced, for example, that ChatGPT will now have memory across chats.

Your conversation will be stored so that ChatGPT becomes more helpful over time, continually remembering details and preferences.

And of course you can review and delete specific memories, wipe the slate clean or just turn off the feature entirely.

And Sama said that they were releasing a few things before deploying GPT-5.

It's Sam Altman, of course, but there's a lot of speculation about when the next big thing from OpenAI will drop.

OpenAI is potentially having some summer events, whether that's in-house or something similar to the November Developers Conference where they might announce some new stuff.

Meanwhile, there's some things that are heating up with the various employees, board members, and insiders at OpenAI.

Elon Musk, Sapina's ex-OpenAI board member, toner, Helen Toner.

You remember her, she was likely the person, the main sort of driver behind the coup, the firing of Sam Altman from OpenAI, Ilya's departure.

It seems like she might have been either trying to scrap OpenAI completely or create sort of a fire sale to sell it to Anthropic, potentially believing that Anthropic would place more care when it comes to AI safety.

Super quick note here that I just wanna quickly get out of the way before we continue because this is something that keeps coming up over and over again when talking about certain things, and it's most fiercely debated when it comes to AI safety.

We have people that want to pause AI.

This is like the pause button, right?

That's the symbol, some of them taken up to kind of show that, hey, we want to pause AI development.

It's unsafe.

And on the other side, we have sort of the AI, let's accelerate this thing.

And there are various degrees of it, of belief how fast we should go.

Should we go defensively?

Should we go at full speed?

People like the newly appointed CEO of Microsoft AI, Mustafa Suleyman is saying AI is gonna be excellent, we should develop it, but he seems to be strongly against certain capabilities like self-replication, kind of fully autonomous.

I guess a different way of saying it is he wants to have it developed, but up to a point to never allow certain features.

I'm paraphrasing, but that's kind of what he said in a recent TED Talk.

But whenever we talk about AI in this channel, AI safety specifically, I try to be unbiased, I try to give a fair representation to both sides as much as I can, and I get yelled at by both sides.

Each side is sort of upset with me for not explicitly saying I'm on their side.

But I think my entire point here is that while there are people on the AI side that really believe in what they say in terms of AI being a global positive force and improving humanity, there's also some people that are just saying it because they think it will lead them to making lots of money.

I think this is obvious, but what might be less obvious is that on the AI safety side, on the AI doomers, AI pause AI, it's the same exact thing.

You have smart people that are actually concerned and actually working on AI safety.

But at the same time, there are other people that say they are concerned with AI safety that are also seeing that message as a way to get them to acquire more money or power or influence or whatever.

It's important to understand that there's motives here besides just what's good for everyone as a whole, and also that the top AI researchers do not agree on these things at all.

Yann LeCun, of course, is a proponent of open source.

He tends to not be too worried about AI risk.

Jeffrey Hinton is worried about AI risk and in fact just recently made a statement saying that open sourcing AI is dangerous.

Mustafa Suleyman, co-founder of Google DeepMind, thinks that AI is gonna bring about incredible benefits and just improve humanity in many different ways, but is very scared of certain X risks, specifically mentioning things like self-replication, autonomous decision making as potential danger points.

Elon Musk is concerned about AI in the past being more concerned with the X risk, but now seemingly more concerned with making sure that the software is open source.

You recently open sourced Grok and is now on track to raising six billion from Sequoia and others.

His Grok-1.5 with vision came out as a very strong vision model.

Specifically when it comes to understanding the physical world, which makes sense, because of course Tesla has tons of data regarding the physical world, driving around on the streets, et cetera.

Sam Altman saying he's not at this time concerned with that risk of runaway AI, of it escaping and doing various nefarious things, right?

He is concerned with things like misinformation and other things that are more, they're not determining a scenario, but there's other risks that society will have to Figure out how to navigate around as AI and AGI, whatever you wanna call that, as it sort of rolls into the world.

Demi Sassabis, the other co-founder of Google DeepMind, currently working under the alphabet umbrella, so kind of part of Google, also has some concerns that he's voiced to Elon Musk in private, and also he's also potentially sounds like he's concerned about powerful AI being used for profit rather than benefiting humanity at large.

In fact, there's this chart that kind of shows the various people and how soon they think AGI will be here, and also how likely it is that there's some X-risk or all future value, how likely it is to be destroyed by misaligned AGI, so if we don't get AI safety right, is it possible that at least all of humanity is wiped out?

And these are, of course, kind of just approximations based on what these various people have said, and different people might put these opinions and people in various different places, but I think the point is no one knows.

There's not a consensus.

People might be going with their gut or just trying to appease whatever in-group they're in or whatever gets the most likes, but all that is to say that I'm fairly optimistic, but I tend to be somewhat fairly optimistic in general, so that might just be how I'm wired.

But with all that said, it is important to understand that there are people that are sounding the alarm on AI safety.

Here's Daniel Cocotayo, and I apologize if I'm mispronouncing that, so he's a philosophy PhD student, worked at AI Impacts, center on long-term risk, then OpenAI.

And then he quit OpenAI due to losing confidence that it would behave responsibly around the time of AGI.

AI Impacts talks quite a bit about AI risk being an existential risk to humanity, giving a lot of supporting arguments for that.

Right, then we have Center on Long-Term Risk, that he worked for.

Our goal is to address worst-case risks from the development and deployment of advanced AI systems, so it looks like they went from an organization that is worried about AI risk to another organization that is worried about AI risk to then working at OpenAI.

Recently, we saw OpenAI fires two researchers for leaking information who were believed to be allies of Ilya and seemingly having ties to effective altruism, the same organization that Helen Toner belongs to, the same organization that Sam Bankman Freed belongs to, and so now Elon Musk managed to subpoena ex-OpenAI board member Helen Toner, and Toner has until May 21st to comply with the subpoena.

One of the things that was very frustrating to see during that whole fiasco with Sam Altman getting fired, with Helen Toner being behind the coup, apparently as far as we could tell, was that they weren't very clear in communicating what the issue was.

I think the quote was Sam Altman wasn't consistently candid.

Very little information was kind of emerging from that, what actually happened, right?

That famous saying, what did Ilya see?

Like, what happened that kind of spooked everybody?

Toner is an expert on AI safety and governance, but this is one thing that I'm noticing is that that term AI expert is being thrown around when talking about people that are concerned with AI safety.

When people start talking about AI safety, they're labeled an AI expert.

We must shut AI down, AI safety expert, and in case you're confused, this red arrow just to, so you understand he is the expert, red arrow pointing to his face, and here's of course Sabine Hassenfelder.

The headline is AI experts make predictions for 2040.

I was a little surprised, and here's her surprised face, mega deaths?

And yes, she does make a mega death joke in the beginning.

I was hoping she would bring up the band and she did.

In a previous video, somebody said that while I try to be neutral about my opinions and try to not have a bias, that my opinions still come through, my bias still comes through in the videos.

And actually, as I'm recording this, I'm just realizing that when I took screenshots, the progress bar at the bottom of these screenshots kind of gives away who I am and am not listening to.

Sabine, 100%. Elijah Yudakovsky, yeah, I mean, I listened to him.

I heard him out.

I tend to like him.

He's funny, he's entertaining.

And this one, why this top AI guru thinks we might be in extinction level, something, something.

Again, what makes someone a top AI guru?

And what is a guru?

Is guru above or below expert?

And then Gary Marcus, I just skipped that one.

But coming back for a second to Rune, apparently he had a little podcast interview.

I think there was actually a few that he did, again, did not drop the pseudonym, the anonymous profile pic, but this is one of the interviews that people are pointing to as an interview with Rune.

Here's a few quick points that he makes.

I would say definitely the AI companies are super secretive, like OpenAI.

Like they're pretty secretive.

And like they don't, they sort of reveal what they're working on, but it's on a timeline where the technology they reveal is like a year or two behind like what they have behind closed doors, right?

They definitely remind me of like a secret Manhattan Project type, right?

And also on the same scale of like world impact, right?

If what they make becomes real.

What do you think are a couple of the companies in that space right now that you could see becoming like generational huge tech companies?

I think OpenAI is already or is well in the process to becoming like a generational company.

I think that Anthropic, which was a OpenAI sort of spin-off, is a bunch of people who were there in the beginning, but you know, had differences of vision or philosophy or whatever.

They're also gonna be super important because they're very good at what is called like model interpretability, meaning like the introspection of models, right?

We have these giant language models.

How do we understand what they're thinking?

Like they've gone further down that line of research than everybody else probably.

I am hopeful about Adept, which is like this, I like few of these very smart people from Google Brain starting this company to make language models like able to use computers.

Like in a general sense, use Photoshop and use Excel and use like tools of that nature.

And I don't know if they can blow it off, right?

Like we, it's too early to say if they're gonna be important or not, or if OpenAI will beat them to that.

But in terms of art, by the way, the stability and Midjourney are probably like, you should keep an eye on them at least.

My buddy Corey has this company Spellbrush and they do like game asset generative design.

I think that stuff is really cool and about to pop off probably.

I think it's gonna pop off sooner in like Asia and stuff than it is here.

Cause I mean, honestly, the Japanese and Korean markets, they play a bunch of these weird games where it's kind of expected to have these, like sort of keep pumping out new characters.

And like those characters are like ripe for being plucked by like generative art and generative AI and stuff like that.

That's interesting.

How is like China doing on the AI front?

Be honest, like I actually don't know.

I don't, frankly, I don't know either, but like I have spoken to a bunch of people I don't know and they kind of don't, they're not like scared or anything, like they don't think that, they don't have like language model technology on our scale.

I'm not really sure why, but like the majority of the best scientists in the world come to the US, that's probably part of it.

They just don't have that.

They don't have any language models on the scale of like GPT-3 or like, you don't see a lot of crazy demos coming out of them, out of their like big companies.

I just want to like ask you.

Quick note, this was November, 2022.

Obviously the landscape is very different in a lot of different ways.

A lot more models out of China, the company that he was talking about, Adept working on AI agents.

And then him saying that OpenAI might beat them to it and that the technology that we're seeing is years behind what's developed internally.

Again, end of 2022, they're talking about agents, AI agents.

That was certainly, I would say before that term started being more commonly used.

Some like popcorn questions if you're down with that.

I'm guessing you've met like some pretty interesting, successful people.

You've tweeted about hanging with Vitalik and things like that.

These really brilliant outlier smart people, do you think they're happy or like what drives them?

I'm assuming they're talking about Vitalik Buterin, the creator of Ethereum.

Interestingly enough, he donated to some of the AI safety causes.

An amount that ended up being somewhere, I think close to a billion, maybe 700 or 800 million that funded a lot of these efforts, including some of the things behind effective altruism.

But from what I've been reading of Vitalik's writing his blog posts, it doesn't seem like he's necessarily 100% aligned with that idea.

He believes in AI being able to provide a lot of good, but also thinks that there's, I think he stated like a 10% X risk.

We should definitely take great care to develop it slowly and carefully.

Develop AI, but do so very defensively, very carefully.

I think a lot of them are without a doubt, like happier than the average person.

They have to be because if they weren't, they wouldn't be able to take like the stresses of the job.

You know, like normal people, when they have like a minor setback in their job, like it ruins their mood, right?

Like it really fucks them up.

Meanwhile, these guys deal with things that are like like horrible.

Like when they make a mistake, it's like really bad, right?

And yet they're able to keep going and like keep tracking.

I mean, there's obviously a lot of differences.

I mean, Vitalik is very, very different than like Sam Baldwin or something.

Just very high tendency for action or like bias for action.

Like they'd rather do something than not.

They see some like vague, far off threat and they start preparing for it rather than like just letting it be.

The other question around that is like, do you need to be unsatisfied to always be like hungry to like create the next big thing or can you just be super curious and motivated out of a place of like happiness and excitement?

I think that really depends on the person.

Like there are many people that there's strivers because of like some deep like some emotion that we don't generally associate with happiness.

Some people just really want like revenge, like to prove their enemies wrong, you know?

And they actually become wildly successful just because of their drive for revenge.

The people I've met like on the other side of the spectrum, like most of the people in Silicon Valley seem to be driven more by like, yeah, like actual, first of all, curiosity, like you said.

They're not satisfied percent, but they are like happy people.

Vitalik, for example, just seemed contemplative, like very intellectually curious and like interested in how he can change the world through like ideas alone, like kind of scholarly almost.

What makes you think that like the world might be actually way better?

I think, first of all, like computers keep getting better, which means like we are able to automate more boring stuff and do more crazy like consumer experiences for people, right, than we were before.

And like that trend line as it continues, it's generally good, I think.

Some people may disagree.

Some people think that phones and computers have been bad for the world, but I really don't think so.

I think the internet in particular has been like, pretty incredibly beneficial for humanity generally, right?

Like you always hear the sob stories and you never hear the things that are like good about the internet.

Like you never hear about how, I don't know, like Facebook connected, like long-lost relatives and things like that.

Obviously my main thing is just like AI, right?

I think in the near future, we are going to become like a much more wealthy, like much more friendly, like sort of smoother civilization than we were before because we're just gonna automate an insane amount of boring, laborious, like kind of painful work and make everything faster and better in like a step change kind of way, right?

Shortly before going offline, Rune had a bit of a debate with this person, Connor.

Connor is of course the top AI guru that thinks we might be in an extinction level event due to AI.

This is lesswrong.com, a website largely around AI safety, meaning that in general people here will worry about AI ex-risk.

And this post is called Read the Rune, which is clever, and I'll link it down below.

Very interesting read.

The intro Rune, a member of OpenAI's technical staff, and they're saying he's one of the few candidates for a worthy opponent when discussing questions of AI capabilities development, AI existential risk, and what we should do about it.

Again, so lesswrong.com, Eliza Yudkovsky, Connor, they're all proponents of there is an AI existential risk.

Here they're presenting a debate that happened online over X about this specific thing, right?

He comes out with one of his inflammatory statements.

He's saying, things are accelerating, pretty much nothing needs to change course to achieve AGI.

Worrying about timelines is idle anxiety outside your control.

You should be anxious about stupid, mortal things instead.

Do your parents hate you?

Does your wife love you?

People fire back and are kind of upset with his statements.

They're saying it's time for governments to step in.

Rune replied to all of that, saying a bunch of pause emojis harassing me today.

My post was about how it's better to be anxious about things in your control.

And they're like, shame on you.

Also, he deleted that previous post because the chance of miscommunication is high.

People misunderstood what he was saying.

And here's a statement that kind of went a little bit viral.

Political culture encourages us to think that generalized anxiety is equivalent to civic duty.

I'm translating that as like the people that are constantly talking about AI risk and complaining online, that them kind of just being anxious about these things is some sort of a noble cause that they're pursuing.

Here's a debate, a clip from a debate on AI safety between on the right we have Connor.

He's of course a proponent of either pausing AI development or regulations to slow it down.

He'll explain his side of it.

And on the left we have the person that's described as sort of the EAC leader.

Somebody that's really pro-development, pro-acceleration, et cetera.

Here's just a few kind of back and forth on what they believe.

You want a market of institutions, based.

Like first of all, based, awesome, loved.

But now I'm also making the point, that's hard.

You need to do policy, you need regulation, you need to work with current institutions.

You need to prevent all the market failure modes.

How do you legislate around information?

But we need to test our hypotheses of what works.

We need A-B testing.

We can't have one global policy because then it's really hard to A-B test if you have like one central node prescribing things for the entire planet, right?

It's fine.

And I think this is based and you should go work for the Charter Cities Institute or whatever, right?

Like this is based, great.

I'm not denying any of this.

If we could have, I would think it was great if we just had like some countries where just like fucking everything is legal and they just try and like all medic sins are legal there and we'll see what happens, based, right?

You know, maybe not all of them, but like, you know what I mean.

And I think stuff like this is probably not good.

All right, we've established this.

Institutions are dysfunctional, super agree.

If we give more power to dysfunctional institutions, that doesn't make things better, why would it?

If we could improve institutions, I think we agree this would be based, this would be great.

And you propose the best way to improve institutions is just through, let's basically, market-based evolution.

Disruption.

Disruption, market-based evolution, et cetera.

Fair enough, I think this is a reasonable position to hold.

Not saying you're wrong.

I'm making a stronger case for agency, I think, than you are.

Or I'm thinking, I think the market is so absurdly inefficient that you can just do better by just literally thinking about it and just trying.

But I don't have to defend this point if you strongly.

All we're pushing for is a bit of humility that our models might like not have as high accuracy as we estimate them to have, right?

And then...

I mean, of course, this is a fully general piece of advice.

At some point, you should stop updating.

At some point, you should not get even more uncertain.

Most people are overconfident.

I think this is a very fair thing to say.

But at some point, you have to stop updating downwards. So my, at least, going back to our hierarchical control structure, as you go up in the hierarchy, there's less and less bits of information you have about the future at larger and larger scales.

It's very hard to predict the whole macro system.

If you're talking about the whole macro system, there's maybe a few statistics that you could predict, right?

And that's just a few bits.

You should be lighter and lighter touch in your ability to control, like your top-down feedback, in a way that's kind of proportional to how many bits you have about the future, right? And so that naturally yields this sort of hierarchical control system, right?

But like banning, having very precise description, prescriptions that are legislated at like a national level or international level, like I'm just saying we should really have very high certainty that this is the right move before we do something like that.

And I think right now, there's too high uncertainty and we should be cautious about making a move we regret, especially given that legislations are so hard to walk back.

Right, but I am the one living in fear.

Well, joking, but...

It's all risk reward, I don't know.

I mean, yes, but like it seems like you're just extremely terrified in a sense around like doing something wrong in this direction to a degree, which seems kind of unreasonable to me.

Like I agree...

I mean, we have a strong prior of like giving too much power to government and very few individuals, how that can really backfire.

And I think there's a gap right now of mass uncertainty where people are scared and they're afraid, they're high uncertainty about the future.

And that's instrumental for some people to consolidate power, crystallize it, and it's very hard to disrupt that.

And we want to make sure that that does not happen.

And that's why we're fighting tooth and nail to be the counter force to that happening right now.

I'm not claiming, I know on this date with this thing, this will go wrong in this way using this technology, which will lead to an unrecoverable state.

I'm saying if you keep just randomly rolling the dice over and over again, with no plan to ever stop rolling or removing the bad faces of the die somehow, then eventually you roll death.

Eventually you roll the X-risk.

This is, I can say with extremely high confidence.

I think the dice is not identically distributed through time, right?

Like...

But it will be if you don't...

We take calculated risks.

Like if you don't do policy, if you don't improve institutions, if you don't make better ways of...

We should do all that.

I just think like right now it's far too early and it's far too instrumental to select few to over legislate while we're in period.

Well then when is the right time?

When is the right time?

When we have a better prior on how the market...

When do we know?

When do we know?

Like when do I wake up and say, all right, now's go time, when?

Well, I mean, right now, I don't think it's right now because there's just too much things are moving dynamically.

Like if you're gonna...

But things will only get more dynamic in the future.

When does it stop?

When does the buck stop?

When do you act?

Well, I mean, it's not going wrong.

I mean, GPT-4 people thought would melt the world or something and everything...

Again, when do you act?

What is the sign that when you see on the sky...

It's a continuum.

There's no discrete point.

It's like...

You don't have a plan.

You don't have a plan.

What you're saying is you have no plan, you have no threshold, you don't know what to do.

You say we should just wait and see what happens, bro.

Well, no, we should play by ear is what I'm...

Only install policies that change things for short time scales because the situation is so dynamic.

But my point is it will only get more dynamic.

There will never be that God comes from the heavens and all right, guys, now's the time.

Why not?

You don't have a plan.

You don't have a model.

Maybe I'm right.

I mean, I think every corporation is adapting in an online fashion to the landscape.

Sure, but if I say now is go time, now is the time for big regulation.

The time is for that's a different question.

You don't even have a way to evaluate it.

You don't even have a function to evaluate.

You just said play it by ear.

You don't even have...

I do think that my point is like current institutions have such a slow clock rate that...

I claim now...

Even if you have a good thesis, a good tangent right now, a model, a first order model of how the future will go from like the local temporal tangent, like that first order model is gonna be super inaccurate in the future.

And if you have a very low clock rate of adaptivity, hyperparameter setting that you impose now and it's like crystallized forever, that's gonna be highly suboptimal, right?

Like it's just, it's gonna be negative, right?

But then at some point you have to act.

Like you're just making excuses.

This is just cope.

This is just cope for not acting now.

But I mean, every company is acting and adapting to the situation and incorporating AI.

Yes, but if they're making a mistake, if everyone is making a mistake and AGI is around the corner and it will kill everybody, if this is true...

No, well, we disagree on that.

Sure, fine.

But you don't have a plan.

I don't care.

Like you don't have a plan.

You said you don't have a plan.

You don't have a plan.

There is no plan.

The only thing you're talking about is only to play it by ear.

This is not right.

No, I think that the system adapts, right?

Like there's planning happening.

The lower levels of the hierarchical cybernetic control tree adapt on a faster time scale and are adaptive to the changing landscape of technology today.

But life can also just die.

Like life can just end.

Life isn't fair.

A meteorite can just hit the planet, and it's big enough.

And then that's just it.

I mean, it did, and then we bounced back better than that.

If it was big enough to just liquefy the planet, it would just be over.

And then that's it.

But I just don't think that a constant rollout of intelligence being progressively incorporated in our techno capital machine is gonna be that discrete event that causes mass disruption.

I don't care if it's discrete or not.

This is irrelevant.

No, I think it's very relevant because a complex system failure is when the system is maladapted to a sudden change in the environment.

And the point is, if you keep things malleable and changing on a fast enough clock rate, as you idiomatically slowly change the effective landscapes of constant rollout of new capabilities, then the system can Morph to incorporate it and incorporate this new reality and not fail, right?

And I think that is literally what we're arguing for, that crystallizing anything at the global level right now will cause later failure because we're gonna be maladapted to reality that we failed to predict accurately.

I feel like that's a pretty good overview of what they believe.

Obviously I can't capture all the detail and nuances.

There's so much more.

You can watch the full interview on Machine Learning Street Talk.

They recently also released like a bonus section where they go deeper into some of these ideas.

I'm curious where you stand on this.

Do you think that over-regulating now is the potentially more dangerous way to go?

Or do you think the danger lies in not having a plan, being flexible, not putting in place these compute caps and regulations?

What path would you choose given the choice?

Who's more correct here, Connor or Beth?

Let me know in the comments.