Vishwas Lele

Developer | CTO Applied Information Sciences | #Azure MVP & RD l Fleeting thoughts on #Cloud, #DevOps, #ML, #O365& stuff. Personal ( not AIS) Twitter account.

[Richard Campbell @richcampbell]

Responding to government RFPs is complex and time-consuming - could large language models simplify the process?

@vlele talks on .NET Rocks at https://t.co/mavPODq4ru about his new startup company building a copilot for this challenging task - and the potential it represents!

[Vishwas Lele @vlele]

@SarahChieng @AndrewYNg @isafulf @metaphorsystems Thanks @SarahChieng your summary is very helpful.

Job and task guidance makes a lot of sense. Using reasoning engine capabilities to automate a few tasks within a "job" is powerful.

I created this 10 minute video example of automating one such task

https://t.co/pYYvT4DWuc

[Vishwas Lele @vlele]

@AndrewYNg @isafulf Is there a recording of the session available please? Thanks!

[.NET @dotnet]

How do we put large language models to work?

#dotNETRocks and @vlele give us cautionary tales and ideas on using LLMs focusing on specific data sets for building LLMs and how size matters. 🦻 https://t.co/oOcHvaUh6m

[Richard Campbell @richcampbell]

The challenge of building apps utilizing large language models is the size of the data set.

@vlele talks on DotNetRocks at https://t.co/kmMRzdtOQc about how complexity grows exponentially as your data gets bigger!

[Vishwas Lele @vlele]

#live360 Join me at 1:00 PM today for not just a presentation but a crucial discussion on a nascent topic concerning many developers. #GenAI won't make #Azure developers obsolete. However, those who harness #GenAI will gain a significant advantage over those who don't! 💡

[Vishwas Lele @vlele]

@CsharpCorner @noopman @CSharpDotTV Thanks @noopman @CsharpCorner for the invitation to the show. I enjoyed the conversation with Magnus!

[Vishwas Lele @vlele]

@AzurePodcast episode #466
Great show.
BTW, Semantic search in cognitive search is about using the LLMs to promote the most semantically relevant results ( different from vector / cosine similarity search). #azure

[AIS @aisteam]

New LinkedIn Learning Course on Managing and Governing Hybrid Environments with Azure Arc.
In this course, @vlele covers the fundamentals of hybrid computing, Azure Arc, and resource types hosted outside of Azure that can be managed with Azure Arc: https://t.co/XvrFcTflR6

[AIS @aisteam]

Repatriation refers to shifting workloads from the cloud back to on-premises operations; however, there's no substantive evidence to suggest that repatriation is happening on a large scale. Read the full article from @vlele: https://t.co/lAfA9HQU0u

[Waymo @Waymo]

We’ve been waiting for this moment and now we can share it with you: Today, we received our paid driverless permit from the CPUC! In the coming weeks, we’ll begin charging fares for rider-only trips and welcoming new riders into Waymo One in San Francisco. https://t.co/xxNiKuLevG

[Vishwas Lele @vlele]

@FareedZakaria @michiokaku Thanks for GPS, Fareed. I enjoy it every week.

While I agree with your guest about Quantum's potential, I am not sure about the characterization that "LLMs are like tape recorders". LLM are all rage because they can reason ( through chain of thought).

[Itamar Golan 🤓 @ItakGol]

I've been waiting for this 🤯

Open Source LLM Models Surpass GPT-3.5 🎉

In a groundbreaking development, a remarkable set of open-source LLM models has outperformed the capabilities of GPT-3.5.

What truly amazed me is not only the exceptional performance of these models but also the astonishing dataset they were trained on. Surprisingly, this dataset consisted of just 6,000 GPT4 conversations. 🤯

It's clear that high-quality data is working wonders. ✨

A few more details:

- OpenChat: Built upon LLAMA-13B, with a context length of 2048.

- Achieves an impressive 105.7% of the ChatGPT score on the Vicuna GPT-4 evaluation. 🏆

- Achieves a remarkable 80.9% win-rate on AlpacaEval. 🥇

- OpenChat-8192: Based on LLaMA-13B, with an extended context length of 8192.

- Achieves an outstanding 106.6% of the ChatGPT score on the Vicuna GPT-4 evaluation. 🌟

- Achieves an impressive 79.5% win-rate on AlpacaEval. 🥇

📚Links:
OpenChat on HF: https://t.co/ByuXE5zAwg
OpenChat 8192 on HF: https://t.co/6yRYpwsjYk
OpenCoderPlus: https://t.co/hiwF2gwoPC
Dataset: https://t.co/D43Kmn8seB
Code: https://t.co/REmURzPxku

[The Gradient @gradientpub]

Astounding progress in AI has led to speculation AI will cause explosive economic growth. @arjun_ramani3 and @zhengdongwang argue that such “transformative economic impact” from AI is much harder than at first glance.

https://t.co/C4XGou1vwe

[Amjad Masad @amasad]

Not only have they mapped out the fruit fly brain, but they actually boot it up in a computer and made it “eat” and “groom” 🤯

[Kevlin Henney @KevlinHenney]

Code generated from LLMs is going to need more testing than code written by developers.

This seems self-evident to me, but I suspect a lot of people are going to learn it (or ignore it) the hard way.

[Vishwas Lele @vlele]

@alfcnz @alfcnz I posted my question under the comments section of the video. Thanks!

[Vishwas Lele @vlele]

@alfcnz In the lecture " 09L – Differentiable associative memories, attention, and transformers" [1:35:50], Yann suggested that state of art chatbots cannot answer a question like this, "if you ask a question like what is the country that has a common border with Germany with...

[Vishwas Lele @vlele]

BTW, I have greatly enjoyed your videos. Thank you for the great content!

[Vishwas Lele @vlele]

the largest commercial exchanges with China." See below – ChatGPT seems to have no trouble with this question.