Adi Mukherjee

Adi Mukherjee

@adim.in

Software Engineer, currently on sabbatical in Japan. Prev: Apple SRE. Working on something new.

41 followers 53 following10 posts

Adi Mukherjee reposted

@jonhendren.com

fuckin cool

Bryan Boatman@bryanboatman.bsky.social

Direct hit on the Sunset Fire in Hollywood. 🎯 Nice work, @lafd.bsky.social #LAFD c/o @nbcla.bsky.social 👨🏼‍🚒

Really informative episode with SemiAnalysis’ Dylan Patel: share.snipd.com/episode/add3...

share.snipd.comAI Semiconductor Landscape feat. Dylan Patel | BG2 w/ Bill Gurley & Brad GerstnerAI Semiconductor Landscape feat. Dylan Patel | BG2 w/ Bill Gurley & Brad Gerstner

Interesting video about building isochromic maps: youtu.be/rC2VQ-oyDG0?...

youtu.beI made maps that show time instead of spaceYouTube video by Václav Volhejn

Adi Mukherjee reposted

@moultano.bsky.social

All of Randall Munroe's books are GOAT for kids' non-fiction.

Great blog covering the progress this year. “Asking o1 to complete proofs in creative ways is effectively asking it to be a research colleague. The model doesn't have to get proofs right to be useful, it just has to help us be better researchers.” Good example of utility that evals fail to capture.

Ethan Mollick@emollick.bsky.social

Bluesky can be a fraught place to post about AI but it is worth noting that the buzz over o1 (& now o3) is not “hype.” We know o1 can actually do some very hard tasks (see my post) & o3 appears to represent a big further leap. They aren’t AGI, but will matter. www.oneusefulthing.org/p/what-just-...

oneusefulthing.orgWhat just happenedA transformative month rewrites the capabilities of AI

Adi Mukherjee reposted

@emollick.bsky.social

Benchmarks are flawed but a way to trace AI over the last year is GPQA Diamond. This is a Google-proof question set that experts get 81% right in their fields & highly skilled non-experts with 30 minutes per question and Google use get 22% GPT-4 got 37% at the start of 2024. o1 got 78%. o3 is 87.7%

Adi Mukherjee reposted

@justincormack.bsky.social

Tools for your LLM in containers? Yes please! www.docker.com/blog/the-mod...

docker.comThe Model Context Protocol: Simplifying Building AI apps with Anthropic Claude Desktop and DockerDiscover how the Model Context Protocol (MCP) simplifies building AI applications by seamlessly integrating Anthropic Claude with Docker Desktop, enhancing developer productivity and workflow efficien...

Adi Mukherjee reposted

@simonwillison.net

I wish people would post more links to interesting things I feel like Twitter and LinkedIn and Instagram and TikTok have pushed a lot of people out of the habit of doing that, by penalizing shared links in the various "algorithms" Bluesky doesn't have that misfeature, thankfully!

Comparing NotebookLM audio overviews to @elevenlabsio.bsky.social’s GenFM podcasts: I’m still blown away by the naturalness of NotebookLM’s conversation, but prefer GenFM’s level of detail, even though it’s a more stilted conversation

OpenAI released its 2nd gen reasoning model, o3 (yeah, even they admitted they suck at names). The evals are perhaps the final nail in the coffin for the scaling wall hypothesis, showing that AI models aren’t hitting a plateau in capabilities. arcprize.org/blog/oai-o3-...

arcprize.orgOpenAI o3 Breakthrough High Score on ARC-AGI-PubOpenAI o3 scores 75.7% on ARC-AGI public leaderboard.

View full thread

In the ARC AGI eval (linked article in the first post), the ‘high compute’ mode results came from spending ~$350K in total on inference, giving the model more compute to search the solution tree.

This shift from training to inference compute is good news for hyperscalers and Nvidia.

OpenAI released its 2nd gen reasoning model, o3 (yeah, even they admitted they suck at names). The evals are perhaps the final nail in the coffin for the scaling wall hypothesis, showing that AI models aren’t hitting a plateau in capabilities. arcprize.org/blog/oai-o3-...

arcprize.orgOpenAI o3 Breakthrough High Score on ARC-AGI-PubOpenAI o3 scores 75.7% on ARC-AGI public leaderboard.

These models excel at reasoning-heavy tasks like coding, summarisation, and can work through PhD-level problems with sufficient test time compute. Unlike their predecessors (4o/3.5-sonnet), these reasoning models get ‘smarter’ with inference compute.

Lots of apps have had text-to-speech for years, but ElevenLabs voices really stand out to me for naturalness of enunciation. I use it a lot for listening to articles. elevenlabs.io/blog/introdu...

elevenlabs.ioElevenLabs — Introducing the ElevenLabs Reader App | ElevenLabsThe ElevenLabs Reader App lets you listen to any text content, with ElevenLabs voices, on the go

You are up to date.