sumukh

founder of kubo | building intent-aware computers

[sumukh @sumukhbett]

we recently got stuck with a stupidly simple question: what if an AI could learn your preferences implicitly?

[Natural Language Processing Papers @HEI]

Riding Brainwaves in LLM Space: Understanding Activation Patterns Using Individual Neural Signatures

Ajan Subramanian, Sumukh Bettadapura, Rohan Sathish
https://t.co/SknmXZD8o7 [𝚌𝚜.𝙲𝙻]

[sumukh @sumukhbett]

@EMostaque unbelievable

[sumukh @sumukhbett]

The fact that this is still happening is an unbelievably bearish signal

[Richard Sutton @RichardSSutton]

I can’t say enough good things about John Carmack @ID_AA_Carmack and his Keen Technologies. But now Khurram Javed @kjaved_ and I have broken away to start our own startup and pursue a slightly different path toward understanding intelligence. Like Keen (and like Ineffable) we at Oak Lab @oaklab_ai believe in reinforcement learning and that intelligence is created and maintained from run-time experience. But we think current deep learning methods are weak and inefficient, and need not more tweaks, but fundamentally new ideas and a thorough reworking before they can provide a solid foundation for achieving the more ambitious goals of AI.

[sumukh @sumukhbett]

@khoomeik What does this really mean? Could it be because a disproportionate amount of data of these languages are probably books and media and not from the internet?

Also, I’m wondering if clear formal and informal use of pronouns (which English doesn’t have) causes this skew?

[sumukh @sumukhbett]

@kjaved_ @RichardSSutton This is awesome!

[sumukh @sumukhbett]

@alokbishoyi97 hi Alok, I’m the founder of a small intent-aware computing startup

would love to be a part of the group

[sumukh @sumukhbett]

@steveruizok @juvoni More software behind the scenes.

[sumukh @sumukhbett]

@steveruizok When we were working on Text-UI in 2022, we kept collapsing into defining ontological structures for UI. That would either constrain the user or constrain the LLM, both of which are bad. This is probably the nature of all DSL-like/based interfaces.

[sumukh @sumukhbett]

@zoink what does it mean to make a novel discovery in a non-verifiable domain?

[sumukh @sumukhbett]

@DavidSHolz we are processing a lot more information, i think taking moments to slow down and actually touch grass is what makes it easier

[sumukh @sumukhbett]

@ArthurMacwaters I’m sorry, I thought we were asking it to count the number of r’s in strawberry

[sumukh @sumukhbett]

read more here: https://t.co/w9Vla91Epy gg

@xyliscx
@rohansathish21

[sumukh @sumukhbett]

we recently got stuck with a stupidly simple question: what if an AI could learn your preferences implicitly?

[Natural Language Processing Papers @HEI]

Riding Brainwaves in LLM Space: Understanding Activation Patterns Using Individual Neural Signatures

Ajan Subramanian, Sumukh Bettadapura, Rohan Sathish
https://t.co/SknmXZD8o7 [𝚌𝚜.𝙲𝙻]

[sumukh @sumukhbett]

in a nutshell, your brain reacts differently to words compared to other people. if we use an LLM like an MRI to decode each person's reading patterns, we can actually see different signatures.

[sumukh @sumukhbett]

@PrasadInchara This is awesome! All the best 😁

[Inchara Prasad @PrasadInchara]

My first-ever internship was at a startup,with the most cliché setup you can imagine - a literal garage office.

The CEO and interns shared one table.
The culture was incredible(if you were a hustler. ) Everyone there was insanely good at what they did.

Its funny how we have unconsciously created the same mechanics at Chiseled 😅

- we never start a brainstorm without a whiteboard.
- ideation always happens at the end of the day (tired brains = wild ideas xD).
- no one’s tied to a fixed desk.
- and yes, we turned a beautiful home into an office too.

@sumukhbett thanks to you, I don’t just know what first-principles thinking is. I know how to live it.