Erik Herou

Principal Systems Architect. Building sovereign, local AI infrastructure. Dropping pragmatic AI tooling from the lab at @prefrontallabs. Stockholm.

[Erik Herou @erikherou]

Scales from a quick n=3 smoke test to deep offline judging via local ensembles.

I built it to quantify what is viable cross different hardware: std 32GB PC laptop up to an M4 MacBook 48GB or GPU server.

Code w @prefrontallabs lab: https://t.co/kv414EB8aG

#LocalLLM @ollama

[Erik Herou @erikherou]

Running local LLMs usually means a lot of trial, error, and out-of-memory crashes.
I got tired of guessing what works and why, so I built a pragmatic benchmarking tool for Ollama setups. Best feature: swap-kill protection if VRAM cap hit. @prefrontallabs

[Erik Herou @erikherou]

@GustavS - this, you have to do this in house. And I want to be there when you do. There is so much to do here. https://t.co/UinjqIgaVj

[Erik Herou @erikherou]

@emmlof Jag känner igen udden i bakgrunden!

[Erik Herou @erikherou]

This should probably get additional attention. The initial reasoning resembles Hofstader's MIU recursive complexity ideas, very intuitive illustration of simpleness creating beautiful complexity. If not true for the universe, certainly true for emergent intelligence.

[Gustav Söderström @GustavS]

“Finally We May Have a Path to the Fundamental Theory of Physics…
and It’s Beautiful” How fitting that this would be presented on the day John Conway (of cellular automata fame) passed! https://t.co/AwitbgUqhF

[astrocyte* @_astrocyte]

Here's my takeaways from #NeurotechGaming2020 day 2, full of wonderful talks by @reynoldsphobia, @molnar_ad, and @ErikKLloyd. It covers why #neurotech is not big yet, how it relates to #gaming and other themes. 👏👏👏 @_yroy_ for organising and moderating. https://t.co/jMeJU8BGEH

[Erik Herou @erikherou]

@emmlof Jag fattar hans feeling. Jag vill gärna ställa följdfrågorna han inte fick. Kanske skicka några länkar. Ett par egna bilder. Och en film på insidan av våra förra avloppsstammar? Eller var det över gränsen?

[Matt Shumer @mattshumer_]

AI INCEPTION!

I just used GPT-3 to generate code for a machine learning model, just by describing the dataset and required output.

This is the start of no-code AI.

[Erik Herou @erikherou]

@4DlasttimeitsIK @elonmusk Best next place to look perhaps: https://t.co/f7uEUAvux8

[Pieter Roelfsema @Pieters_Tweet]

This is a highly impressive paper on how nanoparticles in the retina can restore vision in a rat model of blindness. https://t.co/LhZQoQbSZN

[Demis Hassabis @demishassabis]

Awesome and beautiful! excited to watch the launch tomorrow, good luck and godspeed @elonmusk, @SpaceX and @NASA

[Erik Herou @erikherou]

@ilyasut 2/ so should it be that or conscious part of it all is just a spectator, it could certainly feel like something free and conscious. Even if it was just a deterministic computation result.

[Erik Herou @erikherou]

@ilyasut There is for sure an ocean of things going on in the unconscious. 83 billion neurons, multiplied by thousands of connectomes, add chemistry and micro electromagnetism, add quantum uncertainty and finally the fact that signalling is analog. Brain is The great mystery.

[IEEE Spectrum @IEEESpectrum]

An AI-powered rat could be the neuroscience equivalent of a wind tunnel, allowing researchers to test different types of biologically inspired neural nets to see how well they perform. https://t.co/3NpOQT2hoc

[Elon Musk @elonmusk]

To those who quietly help advance the causes we mutually believe in, knowing advancing the cause is the only reward: thank you

[Erik Herou @erikherou]

@emmlof Ni kan få kolla på filmen av vårat gamla avlopp istället helt gratis! Jag har den på USB! Finns även stillbilder på rör i genomskärning, hålen under garaget och hur rännan bilades ur längs hela källaren.

[Demis Hassabis @demishassabis]

Some of @DeepMindAI’s most fascinating work happens at the intersection of AI and neuroscience. Our paper published in @Nature today is a good example. It takes inspiration from our recent work in Reinforcement Learning to shed new light on how dopamine works in the brain.

[Google DeepMind @GoogleDeepMind]

More exciting @Nature news today: an example of how AI and neuroscience continue to propel each other forward. (1/2)

[Erik Herou @erikherou]

@demishassabis @DeepMindAI I totally agree with this. I cannot understand either how everyone does not think and talk about these things all the time. The great mysteries. We are beginning to get some hooks into them!

[Erik Herou @erikherou]

@demishassabis @DeepMindAI Quote @demishassabis; "this deep desire to understand [...] the mysteries going on all around us. I don't really understand why more people don't think about that all the time. I can barely sleep [...] the meaning of life, how did the universe start, what is consciousness, ..."