Ed Hagopian (@[email protected]) retweeted
Analysis aided by AI just revealed that sperm whale clicks are essentially whale equivalents to human speech sounds.
It's incredible to see AI doing things for research in so many different fields.
Genuinely curious to know how smart they are. teapawt.org/begusgasper/status/173…
RT @omarsar0: Toward Language Agent Fine-tuning
Explores the direction of fine-tuning LLMs to obtain language agents.
Finds that languag…
Ed Hagopian (@[email protected]) retweeted
ChatGPT-4V Multimodal decodes a Redacted government document on a UFO sighting released by NASA.
I have tested this on 100s of redacted documents and I can say we are in a new world.
Ed Hagopian (@[email protected]) retweeted
Want to understand how LLMs and Transformers work?
This is by far the best visualization I've seen.
The article makes you understand the fundamental concepts in 5 min.
ig.ft.com/generative-ai/
Ed Hagopian (@[email protected]) retweeted
The `rebel-large` model is awesome for relation extraction 🔗
Paired with CUDA, it’s blazing fast ⚡️. With @llama_index 🦙, we can now build a knowledge graph over any text data super quickly! 🕸️
Full Colab notebook showing how you can use it:
colab.research.google.com/dr…
Ed Hagopian (@[email protected]) retweeted
Faster sorting algorithms discovered using deep reinforcement learning
nature.com/articles/s41586-0…
Ed Hagopian (@[email protected]) retweeted
Neural Nets for NLP - Carnegie Mellon
This is a great course that covers neural networks techniques and algorithms for natural language processing. Covers neural network architectures for language modelling such as recurrent networks and transformers, tips and tricks for training and debugging language models, multilingual learning, and much more.
Lecture videos of Neural Nets for NLP: youtube.com/playlist?list=PL…
Ed Hagopian (@[email protected]) retweeted
Microsoft just dropped a bombshell.
All free ChatGPT users can browse the real-time internet using the 'Browsing with Bing' plugin soon.
5 insane browsing prompts I tried on GPT-4 that everyone can access soon:
Ed Hagopian (@[email protected]) retweeted
How to use #chatGPT for effective* learning?
A glimpse into the future of learning.
(Spoiler: The future is already here.)
*fun, engaging and memorable
**Snoop Dogg is going hard on me in this one! 😅
A mini-🧵 in 3 parts:
Ed Hagopian (@[email protected]) retweeted
This is wrong.
Developers and community are moats.
Like I said yesterday. You think @khanacademy is going to rip out GPT underneath its new AI-centric education system? And it is one of many building on top of GPT.
You are insane if you think that.
Google will never get Khan to switch.
That is a moat.
Leaked Google document: “We Have No Moat, And Neither Does OpenAI”
The most interesting thing I've read recently about LLMs - a purportedly leaked document from a researcher at Google talking about the huge strategic impact open source models are having
simonwillison.net/2023/May/4…
Ed Hagopian (@[email protected]) retweeted
Ed Hagopian (@[email protected]) retweeted
Oops haven't tweeted too much recently; I'm mostly watching with interest the open source LLM ecosystem experiencing early signs of a cambrian explosion. Roughly speaking the story as of now:
1. Pretraining LLM base models remains very expensive. Think: supercomputer + months.
2. But finetuning LLMs is turning out to be very cheap and effective due to recent PEFT (parameter efficient training) techniques that work surprisingly well, e.g. LoRA / LLaMA-Adapter, and other awesome work, e.g. low precision as in bitsandbytes library. Think: few GPUs + day, even for very large models.
3. Therefore, the cambrian explosion, which requires wide reach and a lot of experimentation, is quite tractable due to (2), but only conditioned on (1).
4. The de facto OG release of (1) was Facebook's sorry Meta's LLaMA release - a very well executed high quality series of models from 7B all the way to 65B, trained nice and long, correctly ignoring the "Chinchilla trap". But LLaMA weights are research-only, been locked down behind forms, but have also awkwardly leaked all over the place... it's a bit messy.
5. In absence of an available and permissive (1), (2) cannot fully proceed. So there are a number of efforts on (1), under the banner "LLaMA but actually open", with e.g. current models from @togethercompute, @MosaicML ~matching the performance of the smallest (7B) LLaMA model, and @AiEleuther , @StabilityAI nearby.
For now, things are moving along (e.g. see the 10 chat finetuned models released last ~week, and projects like llama.cpp and friends) but a bit awkwardly due to LLaMA weights being open but not really but still. And most interestingly, a lot of questions of intuition remain to be resolved, e.g. especially around how well finetuned model work in practice, even at smaller scales.
Ed Hagopian (@[email protected]) retweeted
Chinese authorities arrest ChatGPT user for generating fake news theverge.com/2023/5/9/237167… via @Verge
Ed Hagopian (@[email protected]) retweeted
4. Samsung bans ChatGPT from company devices
Samsung has decided to temporarily ban the use of generative AI tools, such as ChatGPT, on company-owned devices and internal networks following an accidental leak of sensitive data last month.
Ed Hagopian (@[email protected]) retweeted
"I'd far rather be happy than right any day" - #DouglasAdams
Ed Hagopian (@[email protected]) retweeted
the arc of technology is towards simplicity.
talking to a computer like we talk to a human is pretty simple.
we have come a long way from punch cards to natural language, but within the paradigm of natural language, we can now go so far.
Ed Hagopian (@[email protected]) retweeted
Look how quickly this little DeepMind robot gets back up after being knocked down 👀
Ed Hagopian (@[email protected]) retweeted
Spectacular news! GPT4All version2 will soon have access to more AI models with NO NEED OF INTERNET after download than any other system.
I will be writing an in-depth review of each local AI model and their attributes soon.
Ed Hagopian (@[email protected]) retweeted
Wrote about a wild trick I discovered today with the ChatGPT Code Interpreter alpha: you can upload binaries - like Deno or Lua - and it gains the ability to write code in those languages! til.simonwillison.net/llms/c…
Ed Hagopian (@[email protected]) retweeted
I told ChatGPT I was a baker that wanted to get into data science and (using the @noteable_io plugin) it made me a notebook with EDA, Viz, ML and Further Reading. And it explained every line of code and told me how it worked using baking metaphors.
It's gonna change everything.





























