dharma1

I think both can be true - ie. We can harness energy from the sun at much bigger scale than nature has so far, yet we can still learn a lot from how nature has spent massive parallel search over millions of years on coming up with incredible nano engineering we still have no clue how to do

I think Jeff Dean is right in that we will see much more specialised silicon in the future.

If something more bio inspired ie. predictive coding and in-memory compute fundamentally makes continual learning and much lower energy consumption possible there will be specialised hardware for it at some point

FWIW I think the brain has multiple “learning rules” and operates at multiple timescales

Which era Trane are you talking about? The man went through at least 4 major shifts. I’m guessing you are talking about Giant Steps and the later avant garde years?

He did a lot of stuff too which is not dense at all and very approachable - Naima, A Love Supreme, My Favorite Things, Kind of Blue with Miles, a lot more, enjoyable, classic albums.

Dig deeper

Plus you get the license for free when you buy a camera from them!

I’ve been using DaVinci Resolve Studio for 12 years now with frequent updates, with the license I got with the original Pocket Cinema Camera for £800 (still a fantastic pocket size cinema camera, shoots raw cinemadng and ProRes with a beautiful sensor).

One of the best founders and companies around.

yes GPU is significantly faster, but cpu only lets you do it anywhere - wasm in the browser, any server etc.

NPUs - definitely a good use case for at least part of it, there are ports of whisper that use coreML/ANE with less power and 3x speed of CPU only

Give it a shot, 3.1 live one in AI studio/API and max out reasoning - not the one in Gemini app it’s an older model.

Another option is to use pipecat with their VAD and separate STT and TTS and any (fast) LLM of your choice - but it’s more plumbing and not a true speech to speech model

Yes the voice part of OpenAI realtime/voice mode is great but it’s pretty dumb compared to newer models and often gets stuck repeating itself.

Google’s Gemini flash live 3.1 is better, especially used via the API - it can do tool calling (including to other, even smarter LLMs if you set it up yourself), you can set the reasoning level (even high is still close enough to realtime) and it can ground answers in google search. I love bidirectional voice and right now it’s probably the best option. You can try it in AI studio

this kind of thing is super cool to close the loop.

waiting for FPAA to get better so we can vibecode analog circuits

https://www.eetimes.com/podcasts/making-analog-chip-designs-...

“Eric Schmidt, who founded Google” no he didn’t

Easy install, discord/whatsapp/tg out of the box. And some agent orchestration out of the box where the main LLM can farm out tasks to different models/agents - yes Claude code has some of this too but I think this has more

Always take d3 with k2

put the APP3 through a washing machine recently by accident and they had the high pitched feedback if NC was used until they were fully dry after a couple of days. Have also put APP2 through washing machine before but never had this

Give them enough time and they will. EUV will hit limits anyway in a decade.

For china it's DUV+packaging for now, NIL/DSA mid-term, and MoS₂/2D chips long term. But wafer scale, defect free 2D logic is 20–30 yrs out, so no EUV shortcut anytime soon

Mid 90s efnet #philosophy was great. And undernet too

Can be a couple of hundred years with trees and wood used for housing. Long enough to figure things out

Script fonts don’t really look like handwriting - too regular.

But one of the early deep learning papers from Alex Graves does this really well with LSTMs - https://arxiv.org/abs/1308.0850

Implementation - https://www.calligrapher.ai/

They have a bit different characteristics to dynamic/condenser mics - usually less high frequency content, and pronounced proximity effect.

Often used for horns, violin, guitar cabs - sources where you want to reduce "shrillness" but you can use them anywhere.

It’s mostly useful as a hands free shortcut to get replies from chatgpt. Beyond that it’s useless

All the promises of integrating deeply into the OS and have APIs directly into apps so you can use natural language to get any app to do stuff for you is vaporware

So much potential but none of it delivered yet, I hope it will change soon

looking at them again, they're definitely all 3d renders with a consistent toon/outline material

beautiful.

love the illustrations! all custom made?

Not a big proponent of saturated fats but dietary LDL has only a modest impact on LDL-c - 5-10%. Other things that have similar or larger impact are exercise, reducing sugar intake, not being overweight, and consuming soluble fibre. Plant sterols/stanols also help

probably but we don't know how to build that type of intuition. in humans or in machines.

AlphaProof does do some kind of neural network guided search and automated theorem proving to validate it https://deepmind.google/discover/blog/ai-solves-imo-problems...

But it's still fairly brute force and inefficient

thanks, will be checking the book!

My intuition is that some people are able to develop abstractions on top of abstractions (compression is a key part of intelligence) that allow them to traverse the search space much, much faster than without these, and with enough repetition this becomes routine, even faster in the brain.

I don't think we have a good theory of how this works, or at least I haven't come across it.

Neural network guided search is somewhat similar but I think we are missing several key pieces

I often think about Ramanujan in the context of AI, if reasoning is type 2 thinking how did he (apparently) come across a lot of his ideas as intuitions

it was very early. probably one of their all time best acquisitions in addition to YouTube.

Re:using RL and other types AI assistance for chip design, Nvidia and others are doing this too

are these better than qwen at codegen?

Caffeine metabolism becomes less effective as you age

There’s also Guayusa which is pretty much same routine as tea

What’s the best local codegen model that gets close to Claude 3.5? Deepseek coder?

Don’t want to pay for Claude API and Cursor sub gets past the limits quick. Gonna get a M4 Mac with maxed out ram when they come out next month

thanks, was looking for confirmation of this. shame. going to skip this years model and hope they upgrade the camera sensor in the iPhone 17 pro. it's the only reason to switch to a new phone these days