AI Ethics and Technology Giants: Where Do We Go from Here? - NC2C020

I'm your host this week , tim McConaughey , at JuanGolbez on Twitter , and with me , as always , is my co-host , chris Miles at BGPMain on Twitter , and with me , as always , is my co-host , chris Miles at BGP Main on Twitter , and conspicuously absent is some guy we don't remember anymore . So dead to us , dead to us . So we'll just jump right into the news .

This week , there wasn't a huge amount of specific cloud networking news . There were a couple , a lot of AI stuff . It seems like this AI thing just keeps heating up more and more .

So top of the list tonight is something that really should be a surprise to absolutely no one . That is , that OpenAI has removed the nonprofit control of itself essentially and reincorporated as a B Corp , a benefits corporation , which is a for-profit still a for-profit vehicle , by the way and it's given Sam Altman ungodly amounts of equity .

So this is not a surprise to anyone who's been following this right .

Like last year , there was all the drama about Sam being ousted because of essentially what's probably what's happening now , which is , you know , he was ousted by the board for , I mean , we don't know exactly why , but a lot of the safety and trust and ethics and for profit tendencies that Sam Oatman , you know kind of champions kind of come home to roost now ,

that Sam Altman kind of champions kind of come home to roost now . So OpenAI is now incorporated as a for-profit corporation and Sam Altman is even more ungodly rich , so anything to add there .

But I think the way they've been operating over the last six to eight months , they've really been operating just as we've predicted . That the for-profit arm is going to take over , it's going to get the lion's share of the support and all the previous board members have now been either forced out , like you said .

We don't know the details , but pretty much everyone from the original board is gone . It's been replaced with a new suite of people and now the nonprofit doesn't have a majority stake anymore .

So I don't know if dark days are ahead or or if we are going to actually reach this level of of AGI , of artificial general intelligence , as as we're being sold that it is . But yeah , I mean , I mean it pointed out in the article , obviously the opening .

I was just going to start operating more like , you know , anthropic or any other startup at this point . So it's it's not not surprising , but uh , yeah , not , um not not super excited about it .

It's not so much that , like , it's huge news that an ai company is trying to make money , right , but that , uh , this kind of roundabout way that it started as a specifically as a non-profit betterment of humanity type of of play , if you will , uh , and you know , basically was like , all right , well , everybody else took their , you know , everybody else has

their gloves off , so we'll go ahead and take our gloves off too , and and just , you know , uh , unmask this old man with a spoon at the haunted music part , uh , haunted amusement park , and just , just , you know , uh , unmask this old man witherspoon at the haunted music part , uh , haunted amusement park , and just let everybody know that , yeah , actually we're in

it for the money too . So , um , yeah , not not surprising , a little disappointing I , I still think that , uh , we're gonna need to get to a point where somebody I don't know who , but somebody's gonna have to figure out how to get ethics , privacy controls , copyright controls .

I mean , sam Altman's gone on record before as saying basically , like , without ingesting copyrighted materials that AI can't advance , which I'm like , oh gee , that sounds terrible , like I'm just kind of shrugging my shoulders here like the Kevin James mean , you know , like but but OK , so , anyway .

So this is not surprising , but it does lead , uh it does kind of segue , into the next story , which is interesting . Uh , so the next story is about california and , uh , gavin newsom , the governor there , has just vetoed a bill that was sent to his desk that was supposed to kind of compel AI companies to build some kind of .

They're calling it a kill switch in the article . Oh , let me this one's from actually hang on a second . So the first I should have credited . The first article about OpenAI is from Reuters and the second one , one second . Yeah , the second one's from BBC , that's right . So the BBC is reporting that .

So again , newsom blocks what they're calling a landmark AI safety bill . Again , a lot of the safety that's shown in this article . Talking about what was in the bill seems a little bit far-fetched for anybody who understands AI technology at all .

For example , there was a part of a bill that was going to compel ai companies to build a quote-unquote kill switch into the system , like I don't even know .

I mean it sounds great , right , like you can imagine , you get like the , the , the movie idea of somebody running to the server room , like you know , like trying to to pull the , the big lever that shuts down the ai or something , but I'm not sure how they envision that they've got got the axe out of the thing on the wall .

Right and they're just chopping the data cables or something , but anyway , yeah . So yeah , I think what I've heard is that this was blocked because it was dumb , basically it didn't know what to target . I don't know , chris , what .

Uh , point in case , being Elon Musk , like like . People were like , uh , yeah , elon doesn't like this bill , but sadly , he's right , it's , it's , it's shitty , um , so that's , that's at least my , my , uh , my purview of it . But yeah , so hopefully they'll be able to reinstate this with something that's better , um , but , to be honest , probably won't happen .

So we'll see .

So there was definitely a self-interest motive there as well to the whole thing . What a surprise and right exactly . So my question is then okay , so what about when ?

A , when a decent bill that actually might have some some real enforceable teeth and actually tart , you know , be useful , shows up on his desk , it doesn't get vetoed for the same reasons ultimately Like and is that a good thing , right ? So , yeah , to be , be , to be continued , I suppose , all right . So getting away from ai for a little ?

Well , no , not getting away from it all , actually . Uh , so we have a another article from yahoo finance here , uh , saying that meta , uh , it has now partnered , or chosen , if you will , arista networks to um be the , the network infrastructure for its new AI . I don't remember if it's a data center or yeah , they're .

They've built like a , is it a thousand GPUs they've got here , that they're they're going to be running to to train Lama four , and they've chosen Arista as the underpinning network infrastructure to to power that , that , that whole thing . So it's interesting , because I don't know enough , like we had .

I mean , it's been now well over a year since we had Peter Jones on here just pontificating about what an AI enabled network data center might even look .

Like God , it feels like it's been so long , even though it just that's how fast this thing is moving right , where we were just thinking about like , okay , well , let's just make shit up about what that might look like , and here we are actually now building .

These things are being built Right , so we need to get Peter back on here and be like all right , well , what do you think ?

The gpus themselves are an astronomical amount of money , right . So I guess we'll see if Lama 4 ends up being a better model , and also we'll see if whatever city is next to this data center ends up having frequent brownouts .

You know , if they're going to be building these crazy AI workloads , then Arista can definitely play a role there .

One thing that's crazy to me is this isn't really related to the article , but it's just crazy how Zuckerberg has like come back into the forefront as , like you know , obviously he's like this , you know , swagged out cool guy now or whatever he is , but it's like I don't dislike what he's saying now , which is crazy compared to , you know , just a few years ago

. He's talking about the fucking metaverse , which I could not care less about . Um , just just what a 180 dude like . Well , not , not a 180 , I still don't . I'm still not totally in his bag or anything , but like it's , it's just nuts yeah , it's interesting , right ?

He just seems to have picked the wrong tech and now he's trying to catch up with llama .

, absolutely okay . So moving along , uh , so we have a more kind of a technical post here .

So , um , usually on the news we don't get this kind of granular with with uh technology blogs and things like that , but we have an update from aws on the on their official blog blog that you can now do security group referencing across AWS Transit Gateway , which this is .

I don't know if , for anyone that's been working with AWS for a while now , this has been a requested feature for , like I want to say , like six , seven years at this point .

So , just to kind of break it down into the nuts and bolts , so you know , when securing traffic between your workloads within a VPC or even across a VPC peering , what you can do is , you know , as traffic goes through one security group , you can reference that by a separate security group so that you can , you know , kind of do this security chaining .

So to say , right , you know , if traffic is going through exactly . So if traffic has gone through this security group , then I'll allow it into a different security group . So it's it's . It's . It's really cool and useful feature for building , you know , proper security for your , you know inbound and outbound data to your workloads .

But if you use the most popular transit networking solution that they have Transit Gateway , it all breaks right , it's all gone . So they have finally launched this feature , still only supported in a single region , which that was Well , it's TGW .

I think there are some slight limitations that I can't bring to the top of my mind right now , but it's yeah , this has been long time coming . I can't believe it took this long . So , yeah , what about you , tim ?

It was absolutely , absolutely Nobody had it right . No other CSP had anything like it . They created it . They were like all right , everything connects to it . We have all of our primitives work with it . It was perfect out the gate . And then they were like all right , that's it .

They moved on to other services and Transit Gateway has just kind of been it since then . So I think all things that AWS is coin operated . So , even though customers are asking for it , there has to probably be money .

Customers have to be leaving or saying I need this or I can't do business in your cloud , or something like that , to the point where they're going to be like okay , there's revenue , there's money on the table , now we'll take care of it . Because it can't be technological .

It could not have been a technological problem to solve , right , because , all , like you said , it already existed for vpc . Peering like tgw is not that technologically different , like the way , the way that it works under the hood is my understanding you know .

So anyway , yeah , I think I think it just had to do with less about customer obsession and more about , like revenue impact 100% , I guarantee you're right .

But there's been so many announcements in the last six years where I've seen I was like who the hell asked for this ?

It's been kind of an afterthought , like as long as we provide point a to point b , like who cares about any of the other crap , right , like it's very , very basic .

constructs that they offer . So it's just , yeah , I think you're right . All right , next one so we have an article here from the EETimeseu .

So this is we wanted to bring this up because it looks like this is this is a topic we've discussed a few times on the podcast , and now this is expanding into the , the EMEir region as well , talking about repatriation , um , so in the article here we have um , a comment that 83 percent of cios plan for workload shifts in 2024 being , um , you know , which is

kind of a , a vague way they put it in this article . It's like that many CEOs have said some workloads are going to move back on prem right , which , I think you know , we've been preaching hybrid as the approach for for over a year now . There's this focus on , you know , the obviously the cloud costs . We don't need to get into that . We know that's .

That's a rising concern for anyone doing business there .

And then there's this kind of focus on private cloud being able to achieve the economies of scale from a statistical and an operational perspective , which I can kind of understand , but also this emphasis that you can adopt best practices , open standards using open source solutions , and I think anyone saying that on you know , just on its merit alone , has not ever

done it , because that is a very , very difficult thing to do . I'm sure if you were one of these companies that were born in the cloud or you're doing majority of your business in the cloud and you start moving into on-prem , you want that cloud-like experience right .

Business in the cloud and you start moving into on-prem , you want that cloud-like experience right . And we've said on here many times that is a hard , hard thing to do just because of the technical debt that you need to incur in order to get to that kind of playing field right .

But you know it's not surprising at all that these CIOs are going to move some of their stuff back on prem right , like some is such a you might as well just say it just big question marks . They're going to move at least one , but less than 5,000 or something like that is back on prem . It doesn't mean anything , it's a , it's a meaningless number .

We've already we've been preaching the futures hybrid CIOs are figuring out that via costs and also , just , well , mostly cost . It's really cost that's driving this . But part of what's driving the cost is the fact that so many workloads don't belong in a consumption model . They're always on , they're 24-7 , they're always doing something .

So consumption is just through the roof because all you did was put a middleman between you and the compute and the the compute right . So so those and we've said it so many times those workloads don't belong in that cloud , right , if it's static and it's not going to change .

You don't even need that level of agility and all of that that you , you know , you , you needed the cloud for in the first place . So you , you in the cloud and then maybe that one you can move on prem . It's the agility and the elasticity that keeps you in the cloud .

So the stuff that's not going to benefit from that doesn't belong in the cloud if it's going to be on all the time . So , yeah , no surprise there . 84% or 86% of what was in the thing you know , moving back on-prem , absolutely , I 100% believe it .

I would love to know what percentage of those workloads are moving back on-prem , and I think , of course , that's going to be a completely different number based on the applications , based on the infrastructure required to run those applications .

He was just like made a joke and he's like you know , um , across the entire world there's over a hundred songs written about love and that's what I feel like this , like some workloads will move , uh you know , that's what I feel like . Like technically , yes , you are correct , but you're like you've .

You've given me no kind of expectation around what that's actually going to look like , right , but ?

But you know , they point out in this article that over the past decade the amount of international subsea capacity used by these four hyperscalers has increased from about 10% up to 71% . That's , according to the Australian Strategic Policy Institute . Not surprising at all .

Obviously , azure , aws , google are all very vocal about their use of the subsea cable system to build this resilient global network . So I was surprised to see that it grew that much . To be honest with you , I figured well , I guess over the past decade . No , that would make sense . I'm sure that that has slowed in some capacity .

But yeah , so talk about getting into areas like the Solomon Islands and things like that . They say that there's hazards involved with this , about the cable dominance being purely for the hyperscalers . So I don't know . I don't know if you agree with that , tim , but that was kind of one of the sentiments that I pulled out here

as well .

I mean , they're technically infinite , but you would have to , like you know , terraform essentially areas to be able to bring in these subsea cables , you know , and to be able to lay them in the right spots .

There's a reason why , when you look at the map of subsea cables and it's gotten insane , right you look at it 10 years ago and then you look at it now . There's a reason why they generally follow the same pattern , right , like , if you think about it , why wouldn't you just take the quickest between point A and point B ?

And it's because , of course , the seabottom , the seafloor , is not flat flat , you know , and things like that . Uh , you know there's fault lines , tech talk , fault lines and stuff , and you know ocean spreading or subducting and all that . So there's a lot going on there , man .

But yeah , the big one isn't so much about the sea itself , although you'd be amazed by how often undersea cables are both damaged and repaired . And then you know , go find it on youtube sometime . Like repairing an undersea cable , it's crazy . It's those giant ships out there and it's crazy what they do , right , um , but really it's the .

It's the bringing it on to land in . Because think about wherever it comes in , you can't just like there's not going to be a big plug for you to plug it into , right , like that . You have to have infrastructure to take that undersea cable and then essentially provide connectivity to whatever's behind it . There's not an infinite number of that , right ?

So that's really the bottleneck , and I think the the article is trying to say essentially that that bottleneck is getting eaten up by hyperscalers who are laying more and more cable . And there , you know , I always thought , like more cables is good thing and and whatnot .

But if you think about it , like , think about it , what if you're like the australian government and you want to do like the internet thing , where you're going to subsidize subsea cables so that you know the citizens of australia have faster internet or more resilient internet or whatever ? But you can't because there's nowhere to bring in a cable Like .

I mean , there's an actual concern , a valid concern there , and I don't know what the answer is , because it when you , when that happens are are governments going to then build the new facilities to bring in more undersea cables and make the taxpayers pay for it , or is it ?

Are the the hyperscalers going to be on the hook to build their own facilities at that point ? Right , so it's , it's . It's a question to answer yeah , it's .

But I mean , I also assume I mean people , people do this shit all the time where they buy up you know the , the majority of them , and if they're not using it , they resell it to someone else .

So maybe we enter into this weird scenario where you're , you know , leasing things from AWS or Azure , but I don't know , I mean the cloud could keep growing and then they're , they're , you know , they just keep buying it and there is nothing to lease out , right ? They've all run into capacity problems here and there over the last decade as well .

If you found any particular article interesting , of course they're in the Fort Lightning news , I guess , document that we release along along with the the news . So all of our articles that we've used and some of the ones that we didn't but still thought were interesting , are in that document . I urge you to check them out and read them on your own time .

Come up with your own opinion . I'd love to hear if you have an opinion .

You know we're going to be there on behalf of our company , so we'll be doing booth duty and Tim's presenting and things like that . So we will be busy with work activities , but we will definitely find time to hang out and chat with you . So if you're going to be there , please hit us up .

Please subscribe , like subscribe , you know , get a , get our name , get Chris's name tattooed on your chest . He'll sign it at re-event .

You can also visit our website for all of the show notes at cables to clouds dot com . Thanks again for listening and see you next time .