Renu Raman

[Renu Raman @renuraman]

Gokul - spot on..... But the winners for infrastructure (today Nvidia) was not obvious then (Sun -> Intel). The game started in 1995 and was finished by 2005. So right. I remember Andy Grove rebranding PC as 'Internet connected PC' when we at Sun thought we nailed Intel on the chip side (SPARC). Guess what - by 2005, that was lost too...Here's a visual for that

[Renu Raman @renuraman]

Separation of church and state was a big thing for abrahamanic religions culminating with the US constitution.

How do we formalize the same for church and money. All church contributions have to go to a common pool and equally divided by based on population of religion ? @ShashiTharoor seems this one topic you and Shashi see the same way.

Perhaps there is a legislative topic. While difficult to find an answer and impossible to enforce, one needs to make an attempt.

The ‘church’ does need money but not to be misused!!

[Renu Raman @renuraman]

@GavinSBaker @vercel @rauchg Gavin, If you plot 1995 to 2003 - the rise of open source and white box over proprietary boxes , where are we on that timeline. I suspect 1998/1999. @gokulr - gokul had a thesis around 10 year life of Amazon ( early )

[Renu Raman @renuraman]

@IanCutress @intel @Arm @risc_v Ian - Intel should funding printing a full shirt. x86 on the body (like below). The 2 ARMs and extendng to the palm with 5 fingers. x86 + ARM + RISC-V. Funny and might sell a lot.

[Renu Raman @renuraman]

Raja, (Warning - Rant) I disagree. Two points. Culture and SW/HW design trade-offs.

By the time Sapphire Rapids came around, Intel has lost all its marbles (process technology, Marketing and Product Management). Who in their freaking mind would put all the wish list of customers in 1 chip? It was just not what is measured by, there was nobody thinking. I was out of chip business for a decade (ex SPARC guy) and when saw Sapphire Rapids with its 150+ different modes - was easy to call out. PMs could nto decide which features and Engineering heads were obedient and Management (your peers) were clueless and thinking their monopoly in chip is there forever. Failure of PM + Architect + Head of Engineering (your peers). None challenging each other (anti-intel culture). I am skeptical if Intel can , has and will recover from that culture degradation despite changes in management.

Back to main topic - much like we have a spinal cord (which does some computation) and a brain with specialized section, computing is also a combination of interconnected memory coherent (within 1 address space) elements trading of Flops/Ops/Memory capacity/Memory BW/Memory Latency. For 50 years, Operating systems had a simple set of homogenous components to deal with. That era has gone. We can build the hardware, we now have tools to build the software to use these heterogenous components

[Renu Raman @renuraman]

@ericvishria @brettberson @benchmark It’s interesting to note the transition from
It’s a pain in the ass to ‘ I am in the venture business to work with founders’ ( recent interview ) . Relate to that , though

[Renu Raman @renuraman]

@RajanAnandan Prefer a dot for .AI. And red color
Like a bhindi

[Renu Raman @renuraman]

@PatrickMoorhead @Signal_65 If true and I believe in @clattner_llvm AMD should have bought and open sourced

[Renu Raman @renuraman]

My mental model has one common and one different. The common is the cost of search space exploration for a sub branch in a thesis is basically free and zero time. ( pro / con - thinking fast and slow )

The great scientists / CEOs / product managers - Elon for example / political leaders ask the right questions. The answers are there and everywhere esp with LLM

So it’s like jeopardy.

If all the answers are there what is the right question.

So I believe it’s training me
To ask the right question faster and better

For the right question , the answer is obvious

[Renu Raman @renuraman]

@vkhosla Recalling Ed Zander’s quote
“Life is all about refining your gut”. More
So for entrepreneurs and CEOs . Recursive self learning loop.

Now we have AI to scorecard your gut calls over your life.

[Renu Raman @renuraman]

@NaveenGRao @amasad @unconvAI Knowledge has been and will be free ( OSS today ) and

intelligence will be distributed and mobile ( power is the enabler )

Wisdom will be personal and contextual.

[Renu Raman @renuraman]

@luisceze Yes.

[Renu Raman @renuraman]

@PatrickMoorhead AMD can be a manufacturing partner

[Renu Raman @renuraman]

@tqchenml Agreed. Can one AI agent /framework with more
Than 1 compiler geek like @vinodg will it be 1:1 or M:N ?

[Renu Raman @renuraman]

What is the min and max ( for the visual cortex ) needed to recognize an artifact and. Max is saturation point beyond which the visual cortex can’t get any new information. Translate that to say diffusion models for both creation of content and recognition of object. Anybody has studied?

[Renu Raman @renuraman]

Vinod

Long time. Please expand.

Was having a big debate with @narunraman about this exactly. Should we go direct from NLP to machine code because you can or abstractions are useful and necessary. Seems like machines are going to find abstractions that are useful
For ‘then’ as well as layering

I am eagerly expecting ‘FPGA’ to be trivially programmable. ( I say FPGA in quotes to refer to reimagined FPGA in the age of AI )

[Renu Raman @renuraman]

@tejeshwi_sharma @bitspilaniindia @BITSPilaniGoa Yes. My best engineers here in US have been from Pilani. It’s a unique combination that I am trying to figure out why and how ?

[Renu Raman @renuraman]

@sundeep Should add an AMD Ryzen for fun into the mix.

[Renu Raman @renuraman]

@realSharonZhou Totally. Well said.

Code ( any ) is like assembly.

[Renu Raman @renuraman]

@coltmcnealy I think code is becoming like assembly. Almost nobody programs today. So perhaps if I may suggest business as a model. ( BaaM).

[Renu Raman @renuraman]

@wmf @NaveenGRao 5% allocation for. #2 and #3 player is good in case there is a hiccup with #1 player.