Shamit Shrivastava
@Shamits
Building sensory computers from thermodynamic principles @BUCollegeofENG | @IITGuwahati | @oxengsci | @rosfrankinst | We are hiring!
We have shown that biphasic pulse shape, all or none, refractory period, annihilation upon collision, all typical AP characteristics are shown by sound waves propagating in pure lipid membranes, which raises the interesting question "what's the exact role of ion channels?" @sciam
Brain cells communicate with mechanical pulses, not electric signals bit.ly/2DIM8ls 🔒
Applies to atoms as to bits and the pain is even worse in debugging atoms in the real world. Sequence and structures are just the knobs, what one designs for or targets is material behavior. AI is automating the sequence/structure generation part but you are left with more pain
All coding projects have two parts: 1. The fun part: where you get to "create" 2. The pain part: where you have to debug Code LLMs are "automating" the fun parts while introducing bugs and not helping much with debugging. As a developer, you’re left with more pain to deal with.
What if matter could think? We’re building the Liquid Brain to sense and design functional materials inspired by biomimicry. Looking for a curious material scientist to join the journey at apoha.com Job Post: linkedin.com/jobs/view/4262… #MaterialScience #Biomimicry…
An interesting argument from Oxford biologist Denis Noble: AI needs to be built out of water to become truly intelligent
Just released our first scientific preprint from @ApohaAI At the surface: a new assay for antibody "druglikeness". Under the hood: a physical computing substrate generating a new class of data for protein behavior. We call it Liquid Brain. Instead of predicting from structure…
Transitioning from scientist to founder is everything but easy. Performing as a founder is only for a few. BSF is designed to help you win the founder mindset shift, and to equip you with the tools to make fast and effective decisions when it matters most. wilbe.com/bsf
Special evening coming up on 09/26 at our London HQ If you are a scientist navigating what life looks like as a scientist founder, startup operator, an entrepreneurial PI, come chat w/ @tombergk @Shamits @AnnaVujicic - some of the best humans around!
We started the Become a Science Founder Fellowship to help scientists get practical with what it takes to build a company, learn the trade to avoid expensive mistakes + to lay the best foundations possible. 3 years in & the course keeps getting better! wilbe.com/bsf👇
1/ Here are some REALLY interesting findings from a controversial empirical paper asking: *WHY do some people become ENTREPRENEURS?* Of the many findings + hypotheses, a few stand out -SIGNALING theory -REJECTING a system that REJECTS them...
Data is all you need :D
At the sake of sounding repetitive, while it wasn’t clear at the outset, it’s pretty clear now. This is a race to own data that is valuable in fine tuning and RLHF. You should expect foundational models to be largely commoditized and largely free, largely soon. Non obvious…
As always highly recommended! If there is an idea that you think deserves to impact the world and will get smothered if it remains in the confine academia, join the course and learn how to be the one in control.
BSF time🤠 If you are a scientist exploring entrepreneurship + what it means to jump from academia or industry to starting a company: get PRACTICAL with our 4 week Become a Science Founder Fellowship and join a Community of 200 entrepreneurial scientists wilbe.com/bsf
Reminded of these lines from a popular anthem "This happens only in India" :)
Law and order vacuum is going to cause more violence as citizens are left to fend for themselves. In the end everyone get hurt and too much leniency is not going to do anyone any good.
Guy is robbing the store in the US, as usual the store guys are helpless, but then Indians arrive at the picture. 🤣https://t.co/4YAUFugV7K
While everyone's waiting for more #LK-99 news, thought I'd share my own anecdote about making superconductors and how hard it can be to reproduce even a "simple" material...
Atomic-scale simulation is basically impossible with classic computing The only credible paths are: 1) Some sort of magical AI 2) Some sort of magical quantum computer 3) Robotic labs to do testing at massive scale If I was starting a new company it would combine #1 and #3
Jesus 100 million years of compute on an exascale cluster to find the superconductor a human trial and errored into in 20 years. Even with Jensen’s promised 1 million fold step up in compute by 2030, it would still take a 100 years. Can AI reduce this further soon?
This should be a rule: If you're a pre-seed VC, you don't get to call a startup "too early" for you.
Some thoughts on why we need to rethink sensors for better AI - From Linear to Dimensional: Reimagining Sensory Perception in Machines link.medium.com/vQCRMuzKHBb
Opening speech @ALifeConf by @Hiro_Iizuka. #AI -> intelligence without the body, only algorithmically. #ALIFE -> self-organization & embodiment, physical computing, open endedness ❤️
Indeed, thermodynamics requires deduction from observable quantities, whenever I see bottom up computations or simulations described as "thermodynamics", energetics or mechanics is IMHO better description of these models - just like kinetic theory of gases is not thermodynamics.
Thermodynamics makes sense only if one empathizes with the experimentalist.
Thermodynamics makes sense only if one empathizes with the experimentalist.
Your qualifications only matter when you have a losing record. The corollary to this is that nobody cares where a winner went to school or who they worked for. I was reminded of this as I read the saga about the President of Stanford resigning over some research papers. Had…