alex_duf

I don't think a big ai lab like Deep Mind would accept leaking their IP to their competitors so that their engineers can use code completion and generation.

are we that much in a rush that we need to release checkpoints this close? or maybe argon was a little rushed?

Yeah that was my first reaction. In survival mode you want low tech not high tech.

Books are pretty damn good and don't need 500W of GPU

Look at it the other way around, there are 8 months in the year where you can turn off or reduce fossil fuel consumption. It creates financing challenges (keep generation around for a few months in the year means more expensive power), but it also provides independence against foreign fossil fuel imports.

We need it all. All the low carbon sources: The cheap intermittent renewables, the expensive and stable nuclear, the biogas from organic waste etc.

Putting one clean tech against the other is a waste of time we don't have right now.

Jenny Chase is a renewable forecaster for BNEF (Bloomber new energy finance), and the way she works around that is by looking at China's solar panel export rather than what's reported in each country.

If you're not following her, she's often sharing really interesting stuff.

There's a significant engineering cost to having a custom classifier.

An off-the-shelf API has way less friction, if the results are good enough why bother?

are you saying that we may have a situation where 10:00:10 UTC might be 11:00:11 BST ?

If so I'm failing to see how this is less likely to cause bugs

I don't understand the link between attention and kernel smoothing in your analogy, what am I missing?

I thought both were useful in their own way, it's not like one replaced the other right?

You might be missing three 0s on the parameter count, or what am I missing?

I'm assuming you need somewhere 0.5 to 1TB of RAM for the weights only

I guess how bad the situation is depends how quickly that 100% chance of dying pans out.

The quicker it is, the less it spreads.

Has this got a name?

Not OP, but I'm going to guess:

- Nissan's petrol engine is the crossbow, a well optimised generator for onboard battery

- Every other brand's EV is the machine gun. About 4 times as energy efficient as a petrol car when you take into account the lifecycle of the vehicle, so probably more when you talk about a hybrid car.

Isn't it the whole point?

Throwing money at a problem was expensive, it's a lot less expensive now

I don't think there are many fabs capable of producing HBM

My hope is on Huawei, which has recently developed a stacked chip architecture that reminds me of HBM, and I'm wondering if that's step 1 of China entering HBM production.

I don't know enough about hardware, so I might be saying absolute nonsense, happy to be corrected so I can learn where I'm wrong