# AI Safety Rhetoric vs. Competitive Acceleration

**Podcast:** The AI Engineer Podcast
**Published:** 2026-09-15

## Transcript

Welcome to the podcast.
Lately, if you go on X or LinkedIn or basically any social media platform, you hear a hundred people talking about perhaps the ways that AI is going to end all of humanity and, you know, whether it is or isn't.
And there's this massive argument going on right now.
A lot of this came from Dario Amadei.
who in response to a quote unquote whistleblower, which I've covered on the podcast in the past, came out with a piece called We Must Paste the Frontier, basically saying that he wanted to slow down the development of artificial intelligence so that we could make it safe.
Now, there's pros and cons to a lot of the different arguments he covers in here.
I'll talk about a couple of them.
But before we get into that, I think today on the show, I just want to show all sides of this argument and why I don't believe slowing down the development of AI and the pace of development is the correct answer.
And it's really interesting and really kind of a weird time because I feel like there's sort of this weird political divide on this argument.
But at the same time, a lot of people from different sides of the political spectrum are lying on both sides of this argument.
So it's just like a really weird, interesting time.
I'll get through all of the tweets.
There is a ton.
Basically, everyone is digging into this and I'll get through some stories of.
some people and some things that I'm working on and some other stuff like that.
Before I do, though, I wanted to mention if you want the ability to test out the top 90 different AI models in one place without having to get subscriptions to all those different platforms, I'd love for you to check out my own startup, which is called AIbox.ai.
We let you have a single chat interface where you can get all of the image, audio, video, text, and even music models all in one place so you can generate all of that type of...
AI content.
You can test out all of the different models in one spot.
So you can switch from talking to Claude to ChatGPT to Gemini to Grok, all in the same conversation.
You can have OpenAI generate images inside of that.
Then you can have Google VO3 generate video.
You can have 11 labs generate audio all inside of the same chat thread.
It's incredibly useful.
If you want to try it out, it's only $8.99 a month to get started.
It is AIbox.ai.
I'll leave a link to it in the description.
We're super excited with the progress we've been seeing here.
Okay, let's talk about this.
First of all, the blog post that kind of kicked off, I think the big argument here, because it started out with someone who went and worked at Anthropic for I think like six weeks or something like that.
He wasn't like an executive or anything.
He was one who worked at Anthropic for six weeks.
He made a tweet that said, you know, Anthropic and OpenAI and everyone else was gambling with our lives because self-improving AI was getting too good that it could make bioweapons and kind of wipe out all of humanity.
This tweet went mega viral on X with over 700,000.
Last I checked it, probably close to a million by now because it was a few days ago.
Likes on the tweet.
And so a lot of people are talking about this.
He then went on a press tour on every kind of mainstream news site, CNN and all the rest of it.
telling people that AI was a massive danger.
This is going right in at the same time that I believe next week we have Bernie Sanders coming out with a bill to quote unquote ban super intelligent AI and impose a 20 year prison sentence to any developer that is caught developing this, you know, advanced AI systems.
So there's a lot at stake and there's a lot politically that's moving in this.
But what's interesting is I don't really find this is something that's kind of like Republican versus Democrat, although it seems like there's some.
There are some political, you know, obviously Bernie Sanders is a Democrat, so there are some politics to it.
But we do have Democrats coming over to the acceleration side and we have Republicans going over to the deceleration side.
So I would say it's not super clean and clear cut in that regard.
And there's a much more nuanced conversation that has to be had about all this.
But before we get into all of the exciting drama of that, let's talk about some of the tweets.
And the very first thing that I wanted to mention is after Dario came out with this, we had Elon Musk say, yeah, let's slow it down.
We had Sam Altman say, yes, let's slow down.
We had the head of Google say, OK, let's slow it down as well.
So all the big labs, for all intents and purposes, publicly, whether they're actually going to do this or not, say, yep, let's slow down AI development.
Now, is this kind of a PR stunt?
Like, in my personal opinion, I don't actually believe any of them are going to slow down development.
Maybe they'll slow down when they roll out models or they'll slow down how they talk about models.
But like, none of them want to get left behind by the other.
companies.
So I'm sure they all just like pile on to, in my opinion, they're all piling on to this, let's slow down because it's, you know, good PR and makes them seem responsible.
And they're just going to keep developing stuff probably as fast as they can to stay competitive.
Okay, who is not slowing down?
Well, publicly, China's newspaper has called this whole push from Anthropik and everyone else in America.
a Cold War tactic.
Beijing's official press is framing anthropic safety advocacy as a US strategy to hobble Chinese AI development.
And they say, look, we're not going to be involved in anything to do with slowing down AI because we think you guys are just trying to slow us down.
So I'm not listening.
So that's kind of the first big thing that's come out.
China is not slowing down.
So if the US firms do slow down and China becomes the leader, they've essentially handed that over.
Now, Some would argue this is a major economic disadvantage.
That's certainly the argument that's being made from the top political figures in the United States.
Trump recently came out and tweeted, and I'll give you the first and bottom half of his tweet, and I'll skip out some of the funny bits in the middle.
But he says, concerning AI, when in history of business did anyone see the leaders of an industry call for regulation that...
if strongly implemented, will drive them into oblivion and bankruptcy.
AI taking over the world, destroying humanity, and all the other bad things is a hoax.
He then says all of these hoaxes that Americans were forced to endure through deconstructionists and deviants, foul play, and illegal conduct.
President Xi of China just announced that China will be doing absolutely nothing to stand in the way of AI or its future.
Google has recently stated that they plan to build a massive plant in Finland.
all because they are finding permitting too difficult in the United States.
I'm not happy about that and want them to change their mind or their thinking.
AI and data centers will be the greatest economic deployment engine in history, bigger than oil, gold diamonds, or even the internet.
It will not be stopped by brilliantly run destructive forces during the term of President Donald J.
Trump.
Okay, so Trump is coming out and saying, you know, basically he's not going to, because the way that a lot of this is going and I...
We've seen some interesting tweets from the former AI czar, David Sachs.
He said he recently tweeted out and said, he's like, if OpenAI and Anthropic say they want to slow down AI development, he's like, my take is maybe you think it's controversial, but I say go for it.
Slow it down if you want.
No one is stopping you.
But what I don't want you to do is request regulatory force so that every other AI company is going to have to slow down the development of their models.
So, you know, by all means, if they want to slow down.
go for it.
But the problem is that the lobbying from them and what they're advocating in their posts and stuff is that they would like everyone to be forcibly slowed down inside of the United States.
And obviously, China's not going to do that.
So a lot of people economically see this as a big issue.
And then we also recently just had Zuckerberg come out and say that AI development needs to move faster.
So Zuckerberg, traditionally very...
on the you know has donated heavily to democratic candidates and has kind of been in that camp has moved over into all of a sudden this you know he's allying with trump or it's really not a conservative versus you know liberal thing at this point in my opinion but he's kind of moving over to that that camp, which is really interesting because we have Elon and everyone else saying to slow it down.
Elon also obviously is Republican, you know, campaigned for Trump and Trump is saying don't slow down.
Elon saying slow it down.
Mark Zuckerberg, who donated to Joe Biden, is saying, right.
So it's like these political forces are all switching sides, which is just so fascinating.
At the same time, we have Yanlican.
who literally left X because he doesn't like Elon because he's conservative and Yann LeCun is very liberal and there's all the drama in the backstory of like that.
Yann LeCun says, right, Dario was already claiming that GPT-2 was too dangerous to open source back in 2019.
I made fun of them then.
Everyone should be making fun of them now.
We also have many other, I think, prominent conservatives that are going over and saying, They're going over with Bernie Sanders and announcing that they would like to slow down development.
So it feels like it's not super political, although we do have Trump saying that doesn't want to slow down.
But I think from probably just an economic perspective, he doesn't want the stock market to crash, which basically, if we look at the last few years, the stock market has been held up by a lot of the biggest AI firms outperforming NVIDIA, etc.
And so obviously, Trump doesn't want that to slow down.
It's getting better for the economy.
We're seeing a lot of switching of sides here on a lot of different people.
And Jan LeCun is saying, do not slow down, make fun of Dario and everyone else.
Now, Elon publicly said, slow it down, which is interesting.
And a lot of people then, you know, are asking, okay, well, what can we even expect XAI to be building?
Someone said, apart from Grok 4.7, what are we expecting from SpaceX in September?
And Elon said, Grok 4.8, which is a 2.5 terabyte model trained with our new C++ software stack, will finish training this week and start reinforcement learning.
Someone commented on that and said, sounds like slowing down is going well.
And this is kind of what I'm getting at.
Elon said slow it down, but they're releasing, they're getting ready to release their brand new model.
It seems like they're going to break next speeds.
I don't believe any of these people are actually slowing down, even if they say that they're slowing down.
And it kind of reminds me of Elon, if you guys will all remember, you know, not to throw like shade or anything, but like when OpenAI was crushing it with, I think, GPT 4.5, Elon signed a letter with like 100 other AI experts that was like, you know, GPT 4.5 is too powerful.
We need to slow down the front-tier models so they don't get much better because it's dangerous for society, blah, blah, blah.
And at the same time as he was doing that, he was going and starting XAI and getting it all off the ground.
So it just feels like he was like, look, that guy's ahead.
Slow him down so we can catch up.
It's kind of the same thing right now.
Grok isn't the number one top-tier model right now that belongs to OpenAI and then probably Anthropoc is number two.
And so it feels like Grok is, you know, he's just throwing it in the ring to, okay, fine, everyone slow down a little bit.
We can catch up.
That's my vibe on all of that.
Here's something that's interesting, though.
We're actually seeing some interesting statistics.
I saw a great post by Cremio over on X, and he said the electrical prices just keep getting better and better for Georgians.
Georgia Power announced that new large load contracts, overwhelmingly data centers, will save their customers nearly a billion dollars, erasing about $180 per year from their bills.
So a lot of people have complained about prices of electricity going up due to data centers, and I think that that was fair.
However, I think because of that, there's been a lot of pushback and a lot of different industries and people are now requiring data centers, anyone doing these new build outs instead of getting subsidies to pay extra and to pay more.
And they're paying for these large load contracts.
They have to pay extra for it.
And because of that, the people around the data centers are seeing their bills actually drop.
We're actually seeing this in a bunch of different places, along with many people being encouraged to go and post, you know, go and create their own energy.
production plants next to the data centers to power them so they're not pulling it off of the grid.
This is also happening in a lot of places.
So I recently had someone ask me and I posted a screenshot of this over on LinkedIn, but I had someone recently asked me, you know, they said, how do you not have a ton of AI crash outs?
And they're like talking about the future.
They said that their TikTok is very doomsday.
What I replied was that I think this is a very strong psyop.
It's technology that can find a cure to cancer and most diseases.
We had Dario say that in five to 10 years, he thinks most diseases and cancer could be cured with the technology.
And so obviously China isn't going to slow down.
I believe America needs to accelerate, if anything, and they need to do good with what they're able to create.
Yes, we need to have perhaps good guardrails on it.
Yes, we need to do some of these great systems where data centers are forced to do closed.
water systems so there's not all the steam and water waste getting evaporated if you keep it closed and it essentially eliminates water waste issues and we need to have it so that they pay extra for their electricity sure if that's the way we want to do it or we can have it so they generate their own electricity right there are these like genuine complaints about maybe the industry or or the infrastructure so fix those complaints don't just say we're going to stop ai like they get it the good thing that we're all getting out of this and they they just like go to the base and try to like cobble all the data centers and cobble all the ai providers make people do it more responsibly, make them spend more money.
I recently had a great comment on another post where they were essentially saying that instead of just deleting all of these different data centers and cobbling them and slowing them down, just make them or all of these AI model providers.
make them spend more money on alignment, right?
And I think we've seen over the last few years, OpenAI and Anthropik, a lot of them used to spend more time and more attention on alignment before putting out a model.
And eventually it just got so cutthroat, they're trying to release their models faster and faster.
And because of that, then all of a sudden that kind of became a byproduct, right?
And now we have a lot of people complaining about it, but the biggest complaint isn't even coming from people so much as like OpenAI and Anthropik are releasing these things saying that they're super dangerous.
And so it feels like it's coming from inside of their companies.
Now, if this is true, there's a great solution.
They could just spend more time red teaming, more time doing alignment and more time fixing the models internally.
And if they want to slow down, they can slow down.
But I don't think we should have them forcing people.
Okay, this gets to part two of this.
Why is this going on?
Why are they saying this?
And I think I've probably talked about this in podcasts before, but there's been some interesting new tweets that I've seen recently.
which is essentially getting to the point where, and the accusation I think that's leveled, and I feel like I've got some good confirmation of this, is that at the end of the day, they would like to ban open source models.
That's, I mean, whether they're saying it out loud or not, this is the goal, right?
They're like, look.
we need all of this to be regulated.
You can't regulate open source models because it's not coming through an API.
So people could take it and do anything with it.
And we even had Sam Altman.
He has like an alt account on X that he posts from.
And on his alt account, he, you know, he kind of says things that he can't say from his main account, but he was responding to people and saying he believes that open source models will be banned.
And he said that that's great that those models should be put on APIs where they belong.
So Sam Altman is saying the quiet part out loud.
He believes that going along with this, which he's going along with it, and he's endorsing this path is going to lead to open source models getting banned, which obviously would be terrible for the bottom line of what the banning would be great for him and for Dario, but it'd be terrible for the bottom line of Anthropic and OpenAI if all the biggest firms are just using open source models.
There was a report that one of the biggest law firms in America, you know, making billions of dollars a year.
has decided that they're going to run an open source model.
They bought a massive shipment of GPUs from NVIDIA to run their own open source model, reinforcement learning done on their own data that OpenAI and Anthropic don't have access to with all of the cases, all of the contracts, everything they've done in their past.
So they can run their own open source model, essentially, on-prem, and they don't need to pay Anthropic and OpenAI a dime in order to do that.
That's obviously a massive threat to...
this entire industry and this entire business model of Dario and Anthropic and OpenAI.
And so this is what I believe they're attempting to ban and to stall and to get out of circulation is all these open source models, which I mean, it's interesting because OpenAI has released some really impressive, great open source models.
They've done some good to the whole community by doing that.
It's kind of like they bought a lot of goodwill in doing that.
But at the same time, you can see why they move from an open source nonprofit to a for-profit closed source, right?
It's obviously much more profitable to do it this way.
In all of the follow-up from this, Sam Altman has recently went on and said that he is not planning an IPO for OpenAI in 2026.
He's going to be waiting probably till 2027.
And yet, Anthropics has picked the NASDAQ as the place where they're going to IPO.
So it doesn't feel like they are pacing the frontier when it comes to their IPO and they're going to go public.
sooner than later, I think.
So it's very interesting to hear a lot of these companies talk the talk and maybe not walk the walk, but we'll see what happens with all of this.
Guys, thank you so much for tuning into the podcast today.
I hope this was interesting.
If you watched over on video, you saw all of the screenshots I was sharing along the way of different tweets and different news articles that I'm sharing in all of this.
If you're listening on Apple, you can watch this all on video now.
I publish this all on video on Apple.
Thank you so much for tuning in and I will catch you guys all in the next episode.
Make sure as always, you go check out AI box.ai.
If you want to get access to the 90 top AI models all in one place for $9.
There's so many incredible things you can do.
I'll see you guys all in the next episode.