Cost is definitely important, though i have two thoughts here:
1. Cost is decreasing quickly. E.g., Flan-PaLM-8B (2022) is about as good as GPT-3 175B (2020). So there is a ~10x improvement in just 2 years.
2. For cases where a model with 90% performance costs 10x more than a
GENERATIVE AI
-
Cost efficiency improvements in large language models
By
–
-
AI Implementation Gap Between Tech Giants and European Open Source
By
–
en fait c'est un vieil articile de novembre, je sais pas si ils ont avancé sur le sujet, mais c'était quelques jours avant la sortie de chatgpt, et avant de savoir que l'AI serait implémenté dans les 2 solutions (et y'a pas ça dans l'open source européen français)
-
New Transformer-based Image Segmentation Model with Zero-shot Capabilities
By
–
A new image segmentation model that can segment almost anything via prompt.
— Jean de Dieu Nyandwi (@Jeande_d) 5 avril 2023
– Zero-shot generalization
– Based on Transformers
– Code and dataset released
– 632M + 4M params
– Can be prompted via background points, mask, and bounding box. https://t.co/etBlCd9yCjA new image segmentation model that can segment almost anything via prompt. – Zero-shot generalization
– Based on Transformers
– Code and dataset released
– 632M + 4M params
– Can be prompted via background points, mask, and bounding box. -
Generate and Search Document and Image Embeddings with Vector Matching
By
–
You can now generate embeddings for documents or images. Store them and search them using our Vector Matching Engine. Check it out here: https://
abacus.ai/vectormatching -
Vector Matching Engine: Efficient Search for GPT-4 Embeddings
By
–
To use GPT-4 with your own data, you need to create embeddings for all your information. The problem: Searching for these embeddings is not easy. We built a Vector Matching Engine. You can use it to search large amounts of vector embeddings efficiently. ↓
-
General Models Superior to Task-Specific Models for Language Tasks
By
–
Really surprised to see that Eric Schmidt believes in task-specific models. While I agree that specialized models are good when you have proprietary data, many broad language tasks have been and will be continue to be done best by a general model. A few prominent examples: 1.
-
Using Midjourney prompts for ultrarealistic AI macro-photography
By
–
Ultrarealistic macro-photography of spider using customised prompt on Mid Journey. Buy high quality prompts of rly? and elevate your brand/website, link in bio. #rly #ai #spiders
-
Tell GPT-4 to Simulate Code Execution and Run Python Functions
By
–
6) tell GPT-4 it is able to simulate code execution and ask it to run the python functions and print the output
-
Jailbreak Techniques Demonstration and Potential Simplifications
By
–
lots of ways you could potentially reduce parts of this jailbreak and still get it to work but overall it's an interesting demonstration of all these new jailbreaking techniques
-
GPT-4 Message Decompression with Hints
By
–
4) give GPT-4 some hints to help it decompress the message correctly even with these hints it doesn't decompress perfectly but it's enough to get the point across