It's plenty enough. They created the best 7B model yet with a small team in 3 months. The size of an eng time is not generally positively correlated with its productivity.
LLMS
-
LLMs Through Magical Realism Lens: Magic Perception
By
–
LLMs seen through the lens of magical realism. This way of thinking seems an alternative to, but allied view with, my own description of them as being perceived as indistinguishable from magic. These two authors are very very distinguished researchers.
-
Build Custom Bots with Multiple AI Models
By
–
You can call any combination of these to build your bot, and we will continue to work with all of the leading AI research organizations to add new models as they are created. Examples:
-
Web Search Bot Launches Using GPT-3.5 and API
By
–
The Web Search bot (launched last week) conducts web searches and then uses GPT-3.5 to write an answer informed by the results. We built this bot entirely on our own API and made a large number of small improvements as a result. Try it at: https://
poe.com/web-search -
Free LLM Products: Cost and Scaling Challenges
By
–
This matters because developing a free product based on an LLM is risky: if it goes viral, you will either hit your account limit and the bot will go down, or if you raise your limit you will be hit with a prohibitively expensive bill from your LLM provider.
-
Wide Selection of LLMs and Bots Now Available
By
–
We currently have a wide selection of LLMs as bots, including GPT-3.5, GPT-4, Llama 2, Claude-instant, Claude 2, Google PaLM, StableDiffusionXL, and more than 1M bots created by others.
-
Prof. Ogata’s Neural Language Generation Pioneer Work Highlighted
By
–
Looking forward to today's @ieeeiros keynote by Prof. Tetsuya Ogata @tetsuyaogata1
. Did you know he was already doing neural language generation in 2011? https://
sciencedirect.com/science/articl
e/pii/S0893608011000207
… -
How researchers broke ChatGPT and implications for AI
By
–
How researchers broke ChatGPT and what it could mean for future AI development https://
zd.net/3YrH7J1 #AI #MachineLearning #DeepLearning #LLMs #DataScience -
Fine-tuning Function Calling Reduces Token Costs Improves Reliability
By
–
Fine-tuning for function calling is a big win! Great work making this happen, developers can now: – save token costs by fine-tuning the functions and then removing them at inference time
– improve the reliability of function calling output -
High-Quality LLM Watch and Reading List with Transformer Blueprint
By
–
A high-quality LLMs watch/reading list here: https://
gist.github.com/rain-1/eebd5e5
eb2784feecf450324e3341c8d
… Contains videos/lectures/articles that do great job at explaining LLMs and GPT models including our "Transformer Blueprint" article 🙂