Colin Webb

I write software.
Very happy to talk about #AI, #RenewableEnergy, and #RustLang

[Colin Webb @colinjwebb]

One crucial decision when building AI Agents: should you use the naming conventions for AI from Iain M. Banks' Culture series? 🤔

[Colin Webb @colinjwebb]

I'm playing with AI Agents today, using AutoGen.
After some initial hurdles, using Git as a source of communication seems the most promising path at present. The passing of context is reduced.. historical decisions are naturally persisted.. and it fits well with writing software.

[Colin Webb @colinjwebb]

There are two ways to start the day; shock or solace.

Shock - cold shower, hard exercise, ginger hot shots.
Solace - warm shower, gentle movement, coffee or tea.

One kickstarts the day. One warms it up gently.

Which one is best?
Which story do you want to live?

[Colin Webb @colinjwebb]

@bindureddy I love the skyscaper waterfall! 💖

[Colin Webb @colinjwebb]

@dylan522p Wow, looks like a goldmine of useful info!
#getinmybrain

[Anthropic @AnthropicAI]

New Anthropic research paper: Scaling Monosemanticity.

The first ever detailed look inside a leading large language model.

Read the blog post here: https://t.co/6RYwxt6nWI

[Colin Webb @colinjwebb]

@paulg I agree; April and May are the best months in England.
Everything is green and full of life, yet hayfever hasn't ruined it all yet

[Colin Webb @colinjwebb]

@bilawalsidhu Very cool. The meta is mindblowing.
At least they were courteous about their competitor!

[Colin Webb @colinjwebb]

Microsoft's #Phi #AI models are interesting in many ways.

1 - Their training dataset contains a significant amount of synthetic data.

Copyright concerns plague other models, but if training data is generated - does that eliminate a whole host of concerns?

[Colin Webb @colinjwebb]

I've only just realised that M$'s #phi3 model has been named as a Small Language Model (SLM)

It feels like a branding-error.

Perhaps a better name: #ConciseLanguageModel ?

Do we have a name for models only trained on highly-curated and high-quality data?

[Colin Webb @colinjwebb]

@minchoi What are the best local/self-hosted models for this kind of thing, in your opinion? Still stable-diffusion?

[Colin Webb @colinjwebb]

@natfriedman Quiet, gardening robots would be the best.
I'd like one that quietly roams the local area picking up pieces of trash blown away on the wind.

[Colin Webb @colinjwebb]

@timClicks I enjoyed the stream yesterday, thanks!
How long did you code take to run on 1B rows in the end?

[Colin Webb @colinjwebb]

Trying out #rust #rustlang for solving Leetcode questions 🚀🦀

Still not as easy as Python, but its fun seeing the results when submitting.

[Colin Webb @colinjwebb]

@thepatwalls Although, if you got an ice bath, would you get better at tennis? This is the important question! :D

[Colin Webb @colinjwebb]

I was thinking of getting a nice rug, a large TV, and a sofa in my study.

Now I'm thinking I should keep the space ready for a Holotile floor, and an Apple Vision headset.

[Dogan Ural @doganuraldesign]

One step closer to the holodeck!

Introducing: Holotile

AR/VR walking is now possible!

[Jeff Dean @JeffDean]

I’m very excited to share our work on Gemini today! Gemini is a family of multimodal models that demonstrate really strong capabilities across the image, audio, video, and text domains. Our most-capable model, Gemini Ultra, advances the state of the art in 30 of 32 benchmarks, including 10 of 12 popular text and reasoning benchmarks, 9 of 9 image understanding benchmarks, 6 of 6 video understanding benchmarks, and 5 of 5 speech recognition and speech translation benchmarks. Gemini Ultra is the first model to achieve human-expert performance on MMLU across 57 subjects with a score above 90%. It also achieves a new state-of-the-art score of 62.4% on the new MMMU multimodal reasoning benchmark, outperforming the previous best model by more than 5 percentage points.

Gemini was built by an awesome team of people from @GoogleDeepMind, @GoogleResearch, and elsewhere at @Google, and is one of the largest science and engineering efforts we’ve ever undertaken. As one of the two overall technical leads of the Gemini effort, along with my colleague @OriolVinyalsML, I am incredibly proud of the whole team, and we’re so excited to be sharing our work with you today!

There’s quite a lot of different material about Gemini available, starting with:

Main blog post: https://t.co/NzSycJl7aE

60-page technical report authored by th Gemini Team: https://t.co/CEdMRyYSLo

In this thread, I’ll walk you through some of the highlights.

[Colin Webb @colinjwebb]

@thepatwalls "I play tennis every day" must be the best part though! 🎾

[Colin Webb @colinjwebb]

@tobi_emonts .. and move like a ninja around the house! 🥷

[Colin Webb @colinjwebb]

I'm going to need an @Ollama_ai `prune` function soon.
My hard-drive is rapidly filling up with LLMs.