The A.I Megathread (LLM , GPT , Development)

[3rdWorld]

AI poses ‘risk of extinction,’ industry leaders warn - The Boston Globe

Recent advancements in so-called large language models have raised fears that AI could soon be used at scale to spread misinformation and propaganda, or that it could eliminate millions of white-collar jobs.

www.bostonglobe.com

 

[Skooby]

You cannot love technology more than yout own existence.
People all impressed and excited about their own demise.

AI poses ‘risk of extinction,’ industry leaders warn - The Boston Globe

Recent advancements in so-called large language models have raised fears that AI could soon be used at scale to spread misinformation and propaganda, or that it could eliminate millions of white-collar jobs.

www.bostonglobe.com

Technology won't stop for anyone. Either adapt, hop on board or get left behind.


All those years of folks saying "Why we need to know Algebreh" sound stupid right now .

 

[Pirius Black]

Technology won't stop for anyone. Either adapt, hop on board or get left behind.


All those years of folks saying "Why we need to know Algebreh" sound stupid right now .

They were just waiting for AI to progress to the point that it eliminates the need for them to think and use math ever again.

 

[Fani Willis fan]

GitHub - s0md3v/roop: one-click face swap

one-click face swap. Contribute to s0md3v/roop development by creating an account on GitHub.

github.com

About​

one-click deepfake (face swap)

Take a video and replace the face in it with a face of your choice. You only need one image of the desired face. No dataset, no training.

That's it, that's the software. You can watch some demos here.

Installation​

There are two types of installations: basic and gpu-powered.

• Basic: It is more likely to work on your computer but it will also be very slow. You can follow instructions for the basic install here.
• GPU: If you have a good GPU and are ready for solving any software issues you may face, you can enable GPU which is wayyy faster. To do this, first follow the basic install instructions given above and then follow GPU-specific instructions here.

Usage​

Executing python run.py command will launch this window:

Choose a face (image with desired face) and the target image/video (image/video in which you want to replace the face) and click on Start. Open file explorer and navigate to the directory you select your output to be in. You will find a directory named <video_title> where you can see the frames being swapped in realtime. Once the processing is done, it will create the output file. That's it.

Don't touch the FPS checkbox unless you know what you are doing.

yikes..

 

[bnew]

https://archive.is/aVioG

Spoiler: more

Spoiler: more

 

[IIVI]

I saw a panel of fighter pilots that included an F-22 pilot say it would not surprise them at all if not this next generation, but the next (gen 7) would have A.I fighter pilots. A.I-assisted tech was used on the 4th-gen Typhoons to win some dogfights vs the F-22.

Crazy that fighter pilots may really be obsolete, as it's such a crazy expertise.

That said, war doesn't care. It's not about ego, but being effective.
If a fleet of A.I planes from another country go against a fleet of our human-piloted planes and they shoot down a few of ours that's an L.

That tech the video mentions about A.I tracking half a billion people that China sold to other countries as well as soon as you step out the house, the A.I can basically track you.

 

[bnew]

https://archive.is/2zmG4

UAE's Falcon 40B is now Royalty Free

Abu Dhabi- United Arab Emirates: 31 May, 2023 – The Technology Innovation Institute’s (TII) Falcon 40B, the UAE’s leading large-scale open-source AI model, is now free of royalties for commercial and research use, in response to global demand for inclusive access to AI.

www.tii.ae

Abu Dhabi- United Arab Emirates: 31 May, 2023 – The Technology Innovation Institute’s (TII) Falcon 40B, the UAE’s leading large-scale open-source AI model, is now free of royalties for commercial and research use, in response to global demand for inclusive access to AI.

Ranked #1 globally on Hugging Face’s leaderboard for large language models (LLMs), Falcon 40B outperforms competitors like Meta’s LLaMA and Stability AI’s StableLM. Under the permissive Apache 2.0 software license, Falcon 40B end-users have access to any patent covered by the software in question. Apache 2.0 ensures the security and availability of safe and powerful open-source software and establishes an effective governance model.

TII’s unrestricted access to Falcon 40B highlights its dedication to technological innovation and knowledge-sharing, fostering a collaborative ecosystem and strengthening the UAE’s position as a global AI leader. In addition, it reflects the UAE’s commitment to pushing the boundaries in shaping a future where AI plays a pivotal role in driving positive change.

Open-source, royalty-free deployment of Falcon 40B could empower public and private sector entities with efficiencies such as faster project starts, faster iterations, more flexible software development processes, robust community-driven support, and easier license management.

TII aims to cultivate a thriving ecosystem of collaboration, innovation, and knowledge sharing among developers, researchers, and businesses globally. The move promotes transparency, inclusivity, and accelerated progress in AI for the world, unlocking diverse opportunities for its applications across industries and sectors.

Dr. Ebtesam Almazrouei, Director, AI Cross-Center Unit at TII, said: “Waiving Falcon 40B royalties promotes inclusive tech advancements for a cohesive society. We’re committed to using technology to unify humanity and future-proof our world.”

For more information about the current open sourced Falcon 40B AI model and registration, please visit: FalconLLM.TII.ae.

For Falcon 40B’s ranking, visit: Open LLM Leaderboard - a Hugging Face Space by HuggingFaceH4

 

[bnew]

GitHub - the-crypt-keeper/can-ai-code: Self-evaluating interview for AI coders

Self-evaluating interview for AI coders. Contribute to the-crypt-keeper/can-ai-code development by creating an account on GitHub.

github.com

About​

Self-evaluating interview for AI coders

Can AI Code?​

A self-evaluating interview for AI coding models.

Key Ideas​

• Interview questions written by humans, test taken by AI
• Sandbox enviroment (Docker-based) for untrusted Python and NodeJS execution
• Compare LLM models against each other
• For a given LLM, compare prompting techniques and hyper-parameters

Results​

Model	Notes	Python	JavaScript
openai/gpt-3.5-turbo	Proprietary 170B	65/65	62/65
ai21/j2-jumbo-instruct	Proprietary 178B	55/65	39/65
cohere/command-nightly	Proprietary 52B	48/65	45/65
Wizard-Vicuna-13B-Uncensored	Open 13B ggmlv3.q5_0	31/65	48/65
vicuna-7B-1.1	Open 7B ggmlv3 q5_0	51/65	40/65
Manticore-13B	Open 13B ggmlv3.q5_0	47/65	37/65
Guanaco-13B	Open 13B GPTQ 4bit	41/65	37/65
WizardLM-13B-1.0	Open 13B ggmlv3.q5_0	53/65	52/65
WizardLM-7B-Uncensored	Open 7B ggmlv3.q5_1	54/65	37/65
VicUnlocked-30B-LoRA	Open 30B GPTQ 4bit	45/65	41/65
bigcode/Starchat	Open 15.5B	40/65	45/65
bigcode/tiny_starcoder_py	Open 159M	39/65 [1]	N/A
bigcode/Starcoder	Open 15.5B	45/65 [1]

[1] Starcoder models are prompted with FIM techniques, so these results are not directly comparable to the other chat models. See tiny_starcoder/interview-tinystarcoder.py at can-ai-code · the-crypt-keeper/tiny_starcoder

Evaluation of 65B models is on the Roadmap. Can you help? Reach out!

Results (chart)​

 

[bnew]

GitHub - epfml/landmark-attention: Landmark Attention: Random-Access Infinite Context Length for Transformers

Landmark Attention: Random-Access Infinite Context Length for Transformers - GitHub - epfml/landmark-attention: Landmark Attention: Random-Access Infinite Context Length for Transformers

github.com

About​

Landmark Attention: Random-Access Infinite Context Length for Transformers

arxiv.org/abs/2305.16300

 

[bnew]

 

[bnew]

https://archive.is/1dYB1

 

[bnew]

 

[bnew]

https://archive.is/p5ayu

Spoiler: mirror

GitHub - Fictiverse/Redream: Realtime Diffusion, using Automatic1111 Stable Diffusion API

Realtime Diffusion, using Automatic1111 Stable Diffusion API - Fictiverse/Redream

github.com

About​

Realtime Diffusion, using Automatic1111 Stable Diffusion API

 

[bnew]

https://archive.is/H68OD

 

[bnew]

https://archive.is/W2xXW

Spoiler: full text

AlteredAvatar: Stylizing Dynamic 3D Avatars with Fast Style Adaptation

presents a method that can quickly adapt dynamic 3D avatars to arbitrary text descriptions of novel styles. Among existing approaches for avatar stylization, direct optimization methods can produce excellent results for arbitrary styles but they are unpleasantly slow. Furthermore, they require redoing the optimization process from scratch for every new input. Fast approximation methods using feed-forward networks trained on a large dataset of style images can generate results for new inputs quickly, but tend not to generalize well to novel styles and fall short in quality. We therefore investigate a new approach, AlteredAvatar, that combines those two approaches using the meta-learning framework. In the inner loop, the model learns to optimize to match a single target style well; while in the outer loop, the model learns to stylize efficiently across many styles. After training, AlteredAvatar learns an initialization that can quickly adapt within a small number of update steps to a novel style, which can be given using texts, a reference image, or a combination of both. We show that AlteredAvatar can achieve a good balance between speed, flexibility and quality, while maintaining consistency across a wide range of novel views and facial expressions.

paper page: Paper page - AlteredAvatar: Stylizing Dynamic 3D Avatars with Fast Style Adaptation

[2305.19245] AlteredAvatar: Stylizing Dynamic 3D Avatars with Fast Style Adaptation