mishraka

Founder & CEO @trisetra

Your dream home, realized with Trisetra,

[mishraka @mishraka]

@mattpocockuk Are you suggesting to use these phrases in the prompt to get better results?

[mishraka @mishraka]

@AjdDavison Also, the whole thing ends up taking much longer

[mishraka @mishraka]

@AjdDavison I struggle to find the right abstraction level for the LLM to express its ideas effectively, ensuring a productive back-and-forth conversation.

SAD, UML diagrams, and sequence diagrams seem promising abstractions.

[mishraka @mishraka]

@jasonlk @danielrachlin Curiosity is hard to fake

[mishraka @mishraka]

@mrdoob @fedemarino31 How fast?

[mishraka @mishraka]

@RayDalio @grok who is going to rise in the yvle based on Ray’s prognostications

[mishraka @mishraka]

@jasonlk Except that it takes significantly longer to do the same thing

[Kartik Hosanagar @KHosanagar]

Hey @AirFrance, @fapelous, @AFnewsroom: I was charged $17K for biz class tickets for attempted booking that gave errors on the website. You only refunded taxe, even though support confirmed a full refund. I’m stuck chasing $17K for over a month. Expected better.
#AirFranceFail

[mishraka @mishraka]

@jaltma @pmarca Marc’s ability to frame the large market changes in simple digestible form is so impressive.

While he may not right about all his trend predictions, he is always clear about how he thinks of them.

[mishraka @mishraka]

@hubermanlab Walking for 30-40 mins every morning does feel like greasing the wheels of the mind before putting it to work.

[mishraka @mishraka]

@rao2z LLMs can't reason because they don't really understand anything.

[mishraka @mishraka]

@Xfinity even a basic LLM can help avoid these fundamental gaps in customer care.

[mishraka @mishraka]

"$230 million" how is this okay?

U.S. Pier for Gaza Aid Is Failing, and Could Be Dismantled Early via @_Glasp
https://t.co/koWJ0CAQHl

[mishraka @mishraka]

@DrZeeshanZia When I tried to create similar quality output, the journey of tweaking and adjusting the prompts 100s of times and yet not able to remove artifacts in the generated image made me feel these genAI systems are close to the "stochastic parrots" label than not.

[mishraka @mishraka]

@dharmesh Image embeddings are not invariant to significant geometrical variations and perspective change you would encounter in different views of a room.

Multiview reconstruction is still a hard problem to solve robustly.

[mishraka @mishraka]

@ylecun Like storm chasers, there are tech hype chasers who will simply exaggerate any reported development into a world changing event.

These hype chasers are responsible for stirring confusing in the masses.

[mishraka @mishraka]

@OpenAI It would be helpful if you also speak to the difficulty in refining the prompt to get the right output, and how that effort compares to someone firing up an Adobe After Effects or similar softwares to create the shot.

[mishraka @mishraka]

I tried Amazon Q today, and the responses left me puzzled about the intended use of this feature.

It behaves like an old school chatbot which you could only ask one of 10 questions and that too the same 10 questions everytime?

@awscloud @amazonqerc20

[mishraka @mishraka]

@AjdDavison I can't imagine us going back to the days of CV optimizations purely based on pixels and regions.

The priors collected in foundational models will remain critical for robust inferences.

The current approach of running large DL models in fast loop will likely go away too.

[mishraka @mishraka]

@omarsar0 DL has turned applied scientists into alchemists.

This is what happens when you discard the need for understanding how our systems work for too long and spend all our resources on mimicking x <> y.