gehsty

It depends on if Anthropic make more money on their internal spend on tokens than selling them externally. I strongly guess they do.

Or they won’t give you a device if your intention is not to share back any exploits? Is the expectation that Apple gifts black hats research devices?

How do we know 100% subscriptions are subsidized, it’s a possibility that they are actually ok (ie some people over use, majority under use and on net profitable) and the API pricing is actually just super expensive. Subscription is a buffet and API is a private chef :)

Failing to get a permit is not an act of god, having to pay even when the site does t have power is poor milestone / payment planning. Also it’s quite hard to request payment if the thing you are buying is not working (appreciate this quite complex here), but if the people who are delivering the data center are responsible for delivering the power, it feels like oracle would be within reason to not pay?

Then it would be taxed under inheritance tax (at the full value of the property - not just the gains).

Edit - I’m wrong, I am uk based - just read up on US, wild that you can inherit a house that’s gone up in value massively then sell it and experience no gains or inheritance tax.

Is it not that the gains are unrealized? If they sold the house they would pay capital gains tax?

I can’t see what the event that has caused a force majeure claim is, though local opposition is not unexpected in any infrastructure buildout - I would be surprised if that is permitted / accepted.

All the reporting is about delaying payments, not a good look, and definitely undermines the FM claim. FM generally speaking needs something truly in expected and outside of control - COVID / maybe a literal war in the region your a working in. Increased cost exposure and push back from local opposition would not clear the FM bar in my experience.

It’s a blog post about work an LLM did though?

The pre print clearly states it’s a well defined problem limited by the man hours required to sift through the data. I think everyone knows it’s not setting the world alight?

Edge based storage is very very viable, and very very cheap, people still use Apple / Google Photos (consumers don’t want any friction).

Coorperate people likely just want “the best” and are happy to pay a lot for it to make their workforce more efficient.

It’s hard for me to guess where the data center buildout will go, eventually there will be an oversupply, but we haven’t hit that yet, and at the moment the companies owning these high capacity data centers appear to be making lots of very real money in leases to hyper scalers, hard to imagine this premium continuing, but we need to see where demand for cloud based LLMs tops out.

It’s a constant tension in computing that has been around since mainframes and clients… Neither is going to disappear. My general feeling is normal people care more about how thin and light something is than their privacy, so if data center powered LLMs will have a strong future.

I think it’s clear that they are not turning back to use them exclusively, but clearly are ‘turning back’ in an additive way.

Eg putting these things on ammonia powered vessel will reduce the amount of fuel needed, which is probably a net benefit.

As a technology demonstration it’s amazing, as a product it isn’t particularly compelling. It’s to expensive, to heavy, and just a pain to put on and take off.

A $500 light weight version that lifts on and off more like glasses than a headset, would be revolutionary. I know a lot of companies are racing towards this but Apple with their custom silicon and almost limitless reserves of cash, feel like the most likely to get there.

I disagree with it feeling aimless - other than the Vision Pro, you can’t really point to a weak point in their lineup?

You can only invent the modern smartphone once then everything else is iterative. The iPhone, iPhone Pro, Air and now Duo is a crazy strong lineup. Same for the watches, same for the AirPods.

Vision Pro is super interesting but just not ready for prime time yet, but as a demonstration of what is technically possible it is fascinating. Immersive entertainment is a huge opportunity, watching a gig or a sports game like you are there. If they can repackage and get it down to $500-750 then they could have another hit category that largely doesn’t exist (like Bluetooth headphones pre AirPods - around but no one had nailed it).

Difference is they made a mini, it sold poorly, every time they make a phone with a bigger screen it sells like hotcakes.

I really don’t know how you can say he wouldn’t do this things.

The Air particularly - the thinnest, lightest feeling, nicest to hold iPhone ever made, I think that might have been his favorite.

The Vision Pro is clearly an interesting category of device, Apple needs to be in, if it would have seen the light of day under Steve Jobs is unknowable.

He was ruthless at Apple because back then he had to be - now they can run Vision Pro and it not impact their ability run the business, if Apple had A product that did not succeed in the 90s it might have sank them.

LLVM / Swift / Webkit & XNU (I don’t really know of anything using Xnu up stream but the others are large open source projects?)

As a product, would developers switch to a meta model/harness? I don’t think so.

Only way I see is if it becomes the new SOTA / frontier, does anyone think Meta will surpass Anthropic or OpenAI?

I still can’t get my head around why language models are an existential threat to Meta - they own the platforms people watch adds on?

We can use candles instead of lightbulbs?

INS systems are expensive accurate ones are export controlled (or were when I used them on subsea surveys!

Does anyone have to run Linux on an M series? I’m fairly confident it’s one of the most complex ways of running Linux…

I still struggle with local llm being a reason to drop insane money on one of these - they are targeting people who want high performance local LLMs, that value privacy over using 3rd party api, but also dont want to use Nvidia? small part of the pie by any reckoning?

This feels like paying a lot for something that is just not very good yet - cloud models are too good vs local models (may change in the future..)

cables can get pretty big, you could run maybe 75-90MW through a 66kV 1200mm2 Cu offshore wind array cable, maybe 500MW through a 275kV 2000mm2 cu export cable and 1.3GW through a 320kV hvdc bipole set of cables (2GW through 525kv HVDC!) - you would just have the minor inconvenience of convertors and or substations to deal with at either end ;)

I don’t think any corporation would do anything to trigger the lawyers from MS / Google / Apple etc all at once.

I’m not certain of that - I feel like the enterprise zero retention policy will be tight?

Terrible site to try read.

I have no idea about smaller cables but HV power cables (grid / offshore renewables etc) are consistently designed and produced using aluminum conductors. If you want to get really fancy I think you can even buy them with aluminum produced from geothermal power in Iceland (could just be an industry story though).

Not saying it’s not a problem but may not be as challenging as thought (and in debt-financed infrastructure land the price stability of aluminum is as much of a benefit as its low cost).

Local vs remote compute is a constant thread in tech history - mainframes and desktops then local and cloud compute (think Google Photos bs Apple photos - one indexes on device the other indexes in cloud). Now we have the next chapter local vs cloud LLM models.

There will always be a market for frontier labs in the cloud based models - these models will always be able to be bigger, and that will likely translate to doing things local models can’t.

Logically also we’ll likely get to a point where RAM drops in price as production ramps up, and local LLM is both capable and cost effective. This feels like it is coming for Siri / Gemini / Alexa personal assistant type use cases.

So I think the local LLM will become a thing in laptops and phones in a year or two, offering PA type use cases. Professional LLM services will likely remain at the frontier (and in the cloud) for the foreseeable.

Ding ding ding…

I’m not sure what I need to cite but logically when you’ve paid 1B for a lease up front, you’ll want to recoup that through the money you get paid for the electricity you generate. The project accounting probably includes a return on that similar to CAPEX (could be as high as 10%), so its impact grows every year as well.

My feeling is that these projects would require offtake costs so high to be viable that they would not be appealing to the states that awarded the leases. The blockade from trump is a very appealing exit ramp for projects that are dead in the water.

You can see in Germany / UK projects awarded with high lease costs failing (Mona already went pop in uk, I imagine any r4 lease not awarded cfd in AR8 will be returned, Total not moving forward with German portfolio…)

I’ve worked on debt financed offshore wind projects for the last decade, the whole project viability hinges on devex/capex balanced against offtake and resource.

We need to be careful in not throwing the babies out with the bath water - the fees paid for the leases for these projects were so high that the projects would not be economically viable in the first place. The developers are probably doing back flips behind closed doors that they have exited these positions without major financial impairments.

This does not mean that renewables or offshore wind should not be built it just means that the regime around how sites are awarded, and how power purchase contracts are made need to be reworked.

A lot of the world is struggling with this (recently UK re-tendered for Mona lease area with much more favourable conditions than r4 initial lease).

Generally speaking I think the best approach would be to have almost zero / nominal lease fee, then have a competitive auction around a two sided CFD - if the market price is below the strike price the state makes up the difference, if the market price is above the strike price the excess goes back to the state.

I use UniFi Nas Pro - it is just a NAS, not a home server. I have it directly connected to my home server with a 10GB SFP+ link and it works super well.

Thing is this is a technical justification for something that was done purely to add an extra add break in - there is no point in having the technical discussion.

The problem I had with this World Cup was the over commercialization of it. The ticket prices, the cost of getting to the stadiums, the fucking trophy being presented in some bullshit LV case (like it mattered).

From my perspective I let it slide thinking it was the US’s take on how they do sports so fair enough. I can see now following the hyper commercial US World Cup that FIFAs has dollar signs spinning in their eyes and want to continue this way of working.

Unfortunately it doesn’t align with the values of football. It smacks of the Euro Super League (which would have disrupted the whole pyramid of football). It should be nipped in the bud, Infantino fired, and FIFA continuing as a not for profit.