Matisyahu's Blog

I woke up today and headed down to the store to pick up the last few things I have been putting off but had on my list of things to do. All the old pots and pans I had I took down to the local Salvation Army because I’d sooner someone in need have them rather than them just sitting in a bag in my cupboard or being thrown out at the tip. After dropping it off at the Salvation Army I then headed down to Farmers that had a 55% off sale and I bought another pot because at the moment I just have the Dutch oven but I’ve bought another dutch oven but slightly smaller since it was cheaper than getting a pot so I’ll be using that in future to boil things that I cannot put in the microwave such as pasta. While I was down there I popped into Bed, Bath and Beyond to pick up a throw rug and an extra pillow for my seat which has made it very comfortable with the throw rug providing protection on the seat. I also bought a Blu-ray player so then I could play back the DVD collection I have although I’ll need to go for a walk tomorrow to purchase an HDMI cable since there wasn’t one included in the box.

I mentioned how there is a big push by Apple towards making the default being on device LLMs and the exception being workloads that cannot be done locally to be done in the cloud. I was having at look at the Windows App SDK on GitHub and I’ve had a look at the latest experimental release for the 2.2.2 branch (link). What stood out to me was the feature “Language Model APIs on GPU [Experimental]” in the list of features added to the build. As Ed Zitron has pointed out, the much of the data centres that have been announced have either been cancelled, not started or severely scaled back as a result I can’t help but get the feeling that with Microsoft adding such functionality to the Windows App SDK that Microsoft are hedging their bets – that maybe the future is edge (running on the end users own computer) rather than it being cloud based except for those work loads that need to be run in the cloud. It makes me wonder whether we’ll see OpenAI offer scaled down versions of their LLMs and charge a subscription that will provided regular updates to their models that developers can download and use locally then jack up the price of cloud based models to cover the fulling cost of running it in the cloud.

On a side note, there were some Wharfdale speakers I was going to purchase from JB Hi-Fi that looked like a great deal until I found out that it required the purchase of an amplifier thus making it a complete overkill for what I needed it for (the original plan was to get it, hook it up to a WiiM Mini Wireless WiFi + Bluetooth Audio Streamer then having wireless streaming from my computer, Apple TV 4K etc). What I’ve decided to do is instead what for the HomePod refresh to be announced (rumour has it that’ll be released this year along with an Apple TV 4K refresh). I’d be looking at getting two HomePod speakers so then I can have stereo sound at home then use Airplay 2 to stream my music so then I’ll have good sound regardless of the device I am using.