Saksham Singhal

Agent Optimization @googlecloud | ex-@Microsoft Turing and Copilot | Ex-Onsite Engineer @OpenAI

[Saksham Singhal @saksham_singhal]

Excited for everyone to try the first frontier model of the Gemini 4 series as it’s getting rolled out.

[Google DeepMind @GoogleDeepMind]

Introducing Gemini 4 Argon – our new frontier model.

It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.

[Scott @scottstts]

My god this is such a good speech that every SWE needs to hear. You know what? Every person should hear it

Keep the happy memories, eyes on the reality, be excited about the future. That’s the best that anyone can do

[Saksham Singhal @saksham_singhal]

Opus 5 feels like a strong regression from 4.8. It seems to overthink and verify everything and thats a pattern which often leads to a fault.

Have noticed better results mostly with shorter prompts than context-heavy ones. Time to reset old skills and memories and start fresh.

[Saksham Singhal @saksham_singhal]

@doomerzoomer @grok any ETFs that cover these?

[Saksham Singhal @saksham_singhal]

I am attending #GTC2024 in San Jose. Feel free to reach out if you want to meet or just grab a coffee. 😊

[Saksham Singhal @saksham_singhal]

Finally rolling out a new model version (Deucalion) for Microsoft Copilot in Balanced mode. Some people like it better than Creative mode. Try it, it might surprise you. 😉

[Vitor de Lucca 🇧🇷 🏳️‍🌈 @vitor_dlucca]

@MParakhin was there improvements that shipped to Balanced mode?
It seems the answers are now better and bigger...
Ironically for this one was even better than Creative mode.

[Aman Sanger @amanrsanger]

Llama and many recent open-source models have a significant architectural limitation

They use multi-head attention instead of multi-query attention (which is used by PaLM and probs Claude 100K)

This can result in slowdowns of up to 30x

Heres the math behind why (1/n)

[Saksham Singhal @saksham_singhal]

I'm proud to be part of the team that is leading the way in search innovation with the new Microsoft Bing. It's an exciting time and I can't wait for everyone to experience it firsthand.

Sign up for our preview today at https://t.co/d3sVYj0N0d.

[Satya Nadella @satyanadella]

Bing and Edge + AI: a new way to search starts today https://t.co/0y8sw7waNb

[Saksham Singhal @saksham_singhal]

@adcock_brett @SaveToNotion

[Saksham Singhal @saksham_singhal]

@george__mack @SaveToNotion as #Tread #HighAgency

[Saksham Singhal @saksham_singhal]

@lilianweng State Space Models

[Saksham Singhal @saksham_singhal]

@bbourque @SaveToNotion

[Saksham Singhal @saksham_singhal]

T-ULRv6 is largely based on "XY-LENT" (https://t.co/XetaxKLkjc) which leverages X-Y bitexts as well as incorporates the key innovations of T-ULRv5, such as the XLM-E architecture, the novel pretraining tasks of MRTD and TRTD, the improved training data and vocabulary.

[Jordi Ribas @JordiRib1]

I'm excited to share that the Turing Universal Language Representation Model (T-ULRv6) is the first multilingual language understanding model to achieve 1st position on both the Google XTREME and GLUE leaderboards. Congrats to our @Microsoft Turing team on this accomplishment!

[Jordi Ribas @JordiRib1]

Today Image Creator from Microsoft @Bing begins rolling out in a limited preview to select markets. Now if you can't find the image you're searching for on Bing, you can create it. Learn more about how it works and the steps we're taking to ensure it remains fun and inclusive.

[Saksham Singhal @saksham_singhal]

We not only work on innovative research but also ensure productionizing it into various scenarios across Office, Azure, and Bing. Learn more about our team at https://t.co/ccqaANMdsG

Feel free to reach out to me on LinkedIn or send me an email at turingjobs[at]microsoft[dot]com.

[Saksham Singhal @saksham_singhal]

We are looking for talented and ambitious students to join our #MicrosoftTuring team as AS/SWE/Research interns for Summer '23! If you are passionate about working together towards our mission of "making machines literate" to benefit all of humanity, then we want you! 🧐

[Saksham Singhal @saksham_singhal]

As an intern on the #MicrosoftTuring team, you will have the opportunity to work on state-of-the-art multilingual and multimodal Large-Scale models focusing on language representations, language generation, image representations, and reinforcement learning.

[Li Dong @donglixp]

call for the development of Foundation Transformer for true general-purpose modeling, which serves as a go-to architecture for various tasks and modalities with guaranteed training stability #Megneto

[Vishrav Chaudhary @vishrav]

Excited to share our work titled "Foundation Transformers" and presenting MAGNETO, an implementation of Foundation Transformers towards a true general-purpose architecture across various tasks and modalities.

Paper: https://t.co/3QMQhyUMaS

[Saksham Singhal @saksham_singhal]

BEiT-3 = Masked Data Modeling + Multiway Transformer + Scaling Up, a significant milestone of The Big Convergence towards large-scale language, vision and multimodal pre-training.

https://t.co/zWAEuPybKv