Hey | Welcome to this week's edition of AlphaSignal.
Whether you are a researcher, engineer, developer, or data scientist, our summaries ensure you're always up-to-date with the latest breakthroughs in AI.
Let's get into it!
Lior |
|
|
|
IN TODAY'S EMAIL | Read Time: 4 min 22 sec |
|
|
|
RELEASES AND ANNOUNCEMENTS | --------------------- A platform with three chat endpoints and an embedding endpoint, offering a balance of performance and cost. Their models, including a high-performing prototype, excel in multilingual support and code.
--------------------- The suite of AI tools includes Gemini Pro for developers, healthcare-focused MedLM models, and Duet AI for development and security operations.
--------------------- The guide details six strategies to optimize ChatGPT and LLM performance: including clear instructions, using reference texts, breaking down complex tasks, allowing 'thinking' time, utilizing external tools, and performing targeted prompt evaluations.
--------------------- Newly supported models include Meta's Llama 2, GPT-4 Turbo with Vision, and Mixtral 8x7b. Fine-tuning for GPT-3.5 Turbo and other models is now generally available.
--------------------- OpenAI has launched a $10 million grant program named "Superalignment Fast Grants" to fund research on aligning and ensuring the safety of superhuman AI systems, inviting applications by February 18. |
|
|
|
 | Lightning AI Studios: All-in-one AI development cloud platform | Welcome to Lightning AI Studios. Developed by the creators of PyTorch Lightning, this is where deep learning meets simplicity and speed. Discover a persistent GPU cloud environment that's ready whenever you are. Prototype, train, serve, and so much more – all from one place. Zero setup.
Key Features Collaborative Deep Learning: Develop collaboratively in the cloud at any scale. Persistent Cloud IDE: Fully persistent VSCode, JupyterLab with live sessions. Massive Data Handling: Up to 12TB per studio and live distributed metrics. Multi-node Jobs: Scalable and effortless, with no code changes required. Advanced File Management: Visualize and manage over 100M+ files easily. Flexible Connectivity: SSH, PyCharm, local VSCode integration, and AWS deployment. Curated Studios: Tailored environments for training, fine-tuning, and serving cutting-edge LLM models.
Lightning AI Studios is your portal to rapid-prototyping, training, and deploying AI models with unparalleled ease and speed.
Start today without a credit card and get 6 free GPU hours each month. | | Launch a free studio | partner with us |
|
|
|
NEWS | Google Announces Imagen 2: A DALL-E 3 Competitor |  | What's New? | Google’s Imagen 2 offers photorealistic text-to-image generation with improved accuracy and detail. It features enhanced capabilities like multilingual text and logo rendering, advanced inpainting, and outpainting. Available through Google Cloud Vertex AI, Imagen 2 relies on a natural data distribution for realistic outputs and improved image-caption pairing. | What's New? | Imagen 2 excels at generating realistic and aesthetic images, accounting for framing, lighting, and exposure. The model is also integrated with the SynthID watermarking toolkit, which makes it possible to distinguish AI generated images from natural images even after filtering and compression. | What's New? | High-Quality Image Generation: Aligns closely with user prompts for realistic results. Multilingual Capabilities: Supports text and logo rendering in multiple languages. Advanced Editing Tools: Includes inpainting and outpainting for enhanced image manipulation. Accessibility: Available via web and API for Google Cloud Vertex AI customers.
| | Read more |
|
|
|
NEWS | Microsoft Unveils Phi-2: A Remarkably Performant 2.7B Model |  | What's New? | Microsoft Research has unveiled Phi-2, a 2.7 billion-parameter language model, as the latest addition to their Phi series. Despite its smaller size, Phi-2 competes with larger models like Mistral-7B and Llama-2 7B. Like its predecessor Phi-1.5, Phi-2 was trained on "textbook-quality" real and synthetic data. Phi-2 was built by embedding the knowledge of the 1.3B Phi-1.5 through a technique called scaled knowledge transfer. | What's New? | At just 2.7B, Phi-2 outperforms 7B and 13B Mistral and Llama-2 models on composite benchmarks, and even beats the Llama-2-70B on complex reasoning tasks. What’s more, the model also shows improved handling of toxicity and bias, without explicit reinforcement learning from human feedback (RLHF). Phi-2’s success underscores the promise of training LLMs on textbook-quality data, and demonstrates the utility of high-quality synthetic data in LLM training pipelines. | What's New? | Efficiency: At only 2.7B, Phi-2 rivals 7B and 13B models across the board. Quality Training: Uses “textbook-quality” training data. Device Friendly: Runs efficiently on laptops and mobile devices
Accessibility: Available for research on Azure AI Studio.
| | Read more |
|
|
|
Tenyx recently announced a novel solution to one of the most significant challenges in AI: catastrophic forgetting during the fine-tuning of large language models (LLMs). With this groundbreaking methodology, Tenyx helps businesses adapt LLMs to their unique requirements without compromising foundational knowledge and protective safeguards. | | JOIN THE BETA ↗ |
|
|
|
NEWS | DeciLM-7B Beats Mistral 7B and Llama 2 7B |  | What's New? | Deci has unveiled DeciLM-7B, a new open-source language model with 7 billion parameters. The model is more accurate than Mistral 7B and Llama-2 7B, as evaluated on the Open LLM Leaderboard. Additionally, it is optimized for throughput. In a head-to-head PyTorch comparison, DeciLM-7B is 1.83x faster than Mistral 7B and 2.39x faster than Llama-2-7B for sequences of 2048 token input and output, and when combined with Deci’s inference SDK, the new model is 4.4x faster than Mistral 7B with vLLM. | What's New? | DeciLM-7B demonstrates that with neural architecture search, it is possible to simultaneously achieve state-of-the-art accuracy and throughput. These optimizations also translate into major cost savings. Inference with the instruction-tuned variant of DeciLM-7B costs just $0.18/million tokens, compared to $0.35/million tokens for the most cost-effective endpoint for Mistral-7B-instruct. | What's New? | Leaderboard Dominance: Sets new standards in the 7-billion-parameter LLM class. Innovative Design: Utilizes variable GQA for enhanced balance between accuracy and speed. Throughput Superiority: Faster and more cost-effective than other competitive 7B models.
Open-Source Accessibility: Available under Apache 2.0.
| | Try decilm-7b |
|
|
|
| | Want to promote your company, product, job, or event to 150,000+ AI researchers and engineers? You can reach out here. | |
|
|
|
|