The Promise and Pitfalls of Chaining Large Language Models for Email
Tomasz TunguzVenture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. The Promise and Pitfalls of Chaining Large Language Models for Email
Over the last few weeks I’ve been experimenting with chaining together large language models. I dictate emails & blog posts often. Recently, I started using Whisper for drafting emails and documents. (Initially there were some issues with memory management, but I’ve since found a compiled version that works well on my Mac called whisper.cpp) After tying Google’s Duet I wondered if I could replicate something similar. I’ve been chaining the Whisper dictation model together with LLaMA 2 model from Facebook. When drafting an email, I can dictate a response to LLaMA 2, which will then generate a reply using the context from my original email. So far it works sometimes, but there are some clear limitations: First, the default tone of the generated emails is far too formal. [Link to example] Second, if I prompt LLaMA 2 to use a more casual tone, it often goes too far in the other direction. The problem is a lack of nuanced context - the appropriate level of familiarity varies greatly between emails to close colleagues versus board communications or potential investors. Without that nuance labeled and incorporated into the training data, it’s hard for the model to strike the right tone. Third, in multi-party email threads things can get confusing. If Lauren introduces Rafa to me, then Rafa bccs Lauren on the email, LlaMA 2 often replies as Lauren. Fourth, figuring out exactly the right settings for the model can be tough. Sometimes I dictate long emails, in which case the context windows (how much the computer listens to before transcribing) should be very long so the system can remember what I’ve said previously. Other times I’m just returning a very fast email. A quick see you soon or thank you very much. In which case a long context window doesn’t make sense and I’m left waiting for the system to process. I’m wondering whether small errors in the first model compound in the second model. Bad data from the transcription -> inaccurate prompt to the LLM -> incorrect output. I’m loo Overall the potential is exciting, but there are still challenges around tone, context, and multi-party interactions that need to be addressed before this can become a seamless productivity tool. Tn machine learning systems, achieving an 80% solution is pretty rapid. The marginal 15% - the magic behind ML - takes a huge amount of effort, data, & tuning. |
Older messages
SaaS Competitive Advantage Through Elegant LLM Feedback Mechanisms
Tuesday, October 3, 2023
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. SaaS Competitive Advantage Through Elegant LLM Feedback
Centaurs & Cyborgs : The Jagged Frontier of AI
Monday, October 2, 2023
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. Centaurs & Cyborgs : The Jagged Frontier of AI Last week,
Artisanal Emails
Wednesday, September 27, 2023
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. Artisanal Emails Artisanal chocolate, artisanal candles,
Avoiding the PLG Trap : Office Hours with Oliver Jay
Monday, September 25, 2023
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. Avoiding the PLG Trap : Office Hours with Oliver Jay On
Building Applications with AI - Lessons from LangChain, Hearth, & Context.ai
Thursday, September 21, 2023
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. Building Applications with AI - Lessons from LangChain, Hearth,
You Might Also Like
🗞 What's New: YouTube just launched Hype to help small creators get discovered
Thursday, September 19, 2024
Also: Running newsletter ads!
SaaSHub Weekly - Sep 19
Thursday, September 19, 2024
SaaSHub Weekly - Sep 19 Featured and useful products Schedul Threads logo Schedul Threads Boost your Threads following, reach monetization status fast and automate your Threads content publishing for
95 new Shopify apps for you 🌟
Thursday, September 19, 2024
New Shopify apps hand-picked for you 🙌 Week 37 Sep 9, 2024 - Sep 16, 2024 New Shopify apps hand-picked for you 🙌 What's New at Shopify? 🌱 New granular staff permissions for Gift cards Improvement ⸱
[SaaS Club] How firing customers built a 7-figure SaaS
Thursday, September 19, 2024
The SaaS Club Newsletter Hey Reader Here's a quick round up of what's been going on at SaaS Club: In this week's newsletter: 🎧 Discover how firing customers fueled SaaS growth 📈 Turn
The MrBeast Guide
Thursday, September 19, 2024
Today's newsletter is brought to you by LCA. We help companies build their future. From new products to websites, apps, and AI solutions. We work with businesses doing $10M-$10B in revenue. We
Writing Software for Robots
Thursday, September 19, 2024
Tomasz Tunguz Venture Capitalist If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here. Writing Software for Robots In a few years, most feature flags
KoppieOS, ScrapX, Notch, Magic Inspector, and Entyx
Thursday, September 19, 2024
KoppieOS, where Ai meets Personality. BetaList BetaList Daily Magic Inspector Automate browser testing in natural language KoppieOS KoppieOS, where Ai meets Personality. ScrapX Monitor your competitors
“I’m waiting for the right time”
Thursday, September 19, 2024
Read time: 1 min. 03 sec. "Hey Pat, can I talk to you for a second?" My buddy pulled me aside at a wedding last weekend. I already knew what he wanted to ask... "I really want to do what
Mastering Portfolio Construction
Thursday, September 19, 2024
How to build a strategy that maximizes returns. ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏
join me: Big Tech in Fintech
Thursday, September 19, 2024
thought you would be interested Hi there, Laura here, Principal Analyst at CB Insights. Thought you would be interested in this new briefing on big tech's activity in financial services. In under