Philippe Wellens (@philwellens) profile picture

Co-Founder at Autodm | Ex-C3 AI

Paris, France
Joined April 2014
Photo by Philippe Wellens (@philwellens)Photo by Philippe Wellens (@philwellens)Photo by Philippe Wellens (@philwellens)
Impressed by @OpenAI 's dev releases. Faster, cheaper and vision-enabled GPT 4 Turbo with 128k context window. A better and cheaper GPT 3.5 Turbo with 16k (4x previous size). More accessible pricing for fine-tuned models. New opportunities for @autodm #OpenAIDevDay #ChatGPT #ai
1
1
529
First preview of @Autodmai 👀
You new generative AI conversational agent is coming soon. In the meantime take a look at some of the key features of Autodm’s first release. Want to see more? Join our waitlist at autodm.ai #GenerativeAI #chatbot #ChatGPT
3
159
Philippe Wellens retweeted
Announcing Chainlit 🔗💡- Build & share Python LLM apps in minutes⚡️ Chainlit is an open-source 🐍 Python package that lets you create ChatGPT-like UIs on top of your existing code effortlessly! `pip install chainlit` github.com/Chainlit/chainlit 🧵👇
20
177
13
828
215,837
Philippe Wellens 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.
142
902
83
5,757
1,466,514
Philippe Wellens retweeted
I decided to outsource my entire personal financial life to GPT-4 (via the @donotpay chat we are building). I gave AutoGPT access to my bank, financial statements, credit report, and email. Here’s how it’s going so far (+$217.85) and the strange ways it’s saving money. (1/n):
761
2,788
1,027
17,322
7,252,908
Philippe Wellens retweeted
Generate videos with nothing but words. If you can say it, now you can see it. Introducing, Text to Video. With Gen-2. Learn more at research.runwayml.com/gen2
154
880
471
3,775
1,488,171
Philippe Wellens retweeted
Why are tropical waters so transparent? Because they're dead. This is connected to why some scientists fear global warming could make Europe *colder* And freeze 80% of all humans above the 60th parallel:
108
2,633
320
11,439
2,386,214
Philippe Wellens retweeted
ok, after many tests, I've managed to get my Additive Prompting structure working for interior design/architecture photography with 90%+ coherence to the prompt in Midjourney. Examples & prompts in thread. Read them as you look at the images 🧵 Thread
155
482
76
4,457
1,400,667
Philippe Wellens retweeted
Space Elevator
Replying to @SpaceX
On launch day, when the crew arrives at the pad, they’ll ride the tower elevator to the access arm and board Dragon for their flight to space
4,818
6,497
627
113,527
74,263,675
Philippe Wellens retweeted
The future of LLMs will be open source models trained on customer specific data and securely deployed within the customers own cloud infrastructure.
True story when I joined PwC in the 90’s, employees weren’t allowed to access the internet from their PCs due to security concerns. Each era brings its own set of challenges which are rapidly overcome 😊 wsj.com/articles/jpmorgan-re…
11
56
11
296
146,556
Great read about decision making when overwhelmed with high availability of advisors or information advice. blog.ycombinator.com/what-to…
1
Thank you @CBinsights for painting the food delivery industry competitive landscape goo.gl/hnzgQ7
2
Thanks @podcastwire for writing about @jenny_blake's pivot method "The better we get at pivoting, the more it becomes a continuous process”
In a normal winter u expect 4-5 record high temperatures. Last winter there were 35 record high temp goo.gl/XrnMbI #climatechange