Welcome to BIGDATA AI News Briefs. This timely new feature provides the latest industry insights and perspectives surrounding AI areas such as deep learning, large-scale language models, generative AI, and transformers. We work tirelessly to unearth the most timely and interesting information underlying the day's most popular technologies. We recognize that this field is rapidly evolving and want to provide regular resources to keep you up to date with the latest information. enjoy!
GenAI technology startup AI21 Labs is expanding adoption into enterprise AI infrastructure and is used by eBay, Monday.com, Carrefour, and Ubisoft. An Israeli company that develops advanced AI technology to solve complex problems across a variety of industries has raised $155 million.
A few days ago, Gartner announced the 2023 Hype Cycle for Emerging Technologies as shown below. No surprises here.
Power your AI training and inference workloads with Latitude.sh Accelerate. Powered by NVIDIA H100 GPUs, Latitude Accelerate accelerates AI and machine learning tasks, making both model training and execution faster and more efficient. With dedicated instances, 32 cores per GPU, and hourly billing, Accelerate delivers unmatched performance and flexibility, all at the highest cost per GPU on the market.
GPT Pilot is a research project that explores how GPT-4 can be used to generate fully working, production-ready apps. The main idea is that most of the app's code (probably 95%) can be written by AI, but for the remaining 5%, developers are and will be needed until full AGI is achieved. It means that it becomes. Here are the steps GPT Pilot takes to create an app:
Prompt2Model – Generate Deployable Models from Instructions – An open source project that allows developers to train deployable special-purpose NLP models using natural language task descriptions. This method combines dataset acquisition, LLM-based dataset generation, and supervised fine-tuning.
WizardLM aims to improve large-scale language models (LLMs) by using LLMs to generate complex instruction data rather than manual human input. This model uses a method called Evol-Instruct to evolve simple instructions into more complex instructions for fine-tuning.
At the Google Cloud Next event in San Francisco last week, Google made the surprise announcement that it is now offering Llama 2 and Falcon LLM on Vertex AI on Google Cloud. This move was not expected as Google was previously the only cloud provider that did not host Llama 2 or other open source LLM models in partnership with a rival institution. Google's decision appears to be in consideration of businesses looking for more options. Following this trend, Llama 2 has become the most popular large-scale language model after GPT-4, considering its open source and commercial availability. For Llama 2, Google said it is the only cloud provider to offer both adapter tuning and RLHF.
Additional Information! Meta's SeamlessM4T (Massive Multilingual Multimodal Machine Translation) is a multimodal model that represents a major advance in speech-to-speech, speech-to-text translation and transcription. Published under the CC BY-NC 4.0 license, this model supports approximately 100 languages for input (voice + text), 100 languages for text output, and 35 languages for audio output (plus English). It aims to eliminate dependence on multiple models by consolidating functionality into one. It can handle:
- Supports voice input in 101 languages
- Text input/output in 96 languages
- Audio output in 35 languages
This model delivers state-of-the-art results by leveraging Fairseq2, the largest open dataset for multimodal translation, and other advances. Reduces toxicity and bias compared to previous models. This integrated model enables multiple tasks without relying on multiple separate models.
- Speech-to-speech translation (S2ST)
- Speech to Text Translation (S2TT)
- Text-to-speech translation (T2ST)
- Text-to-text translation (T2TT)
- Automatic speech recognition (ASR)
OpenAI has partnered with Scale for fine-tuning and advanced data labeling of GPT-3.5, allowing you to unlock the full potential of GPT by adapting models to your own data. Companies like Brex are already using this platform to optimize business and model performance. Scale's high-quality data engine and custom LLM platform helps you:
- Build a custom LLM that fits your business needs
- Create powerful custom models that increase efficiency and reduce costs
- Take advantage of Scale's fine-tuning and data preparation platform
- Optimize your AI investment
- Don’t let AI work for you instead of the other way around
LlamaGPT is a self-hosted offline chatbot that provides a private ChatGPT-like experience. This project is a culmination of open source contributions from various developers.
AI2 releases the largest open source text dataset for LLM pre-training. Dolma is his 3 trillion token dataset that sets a new standard for openness in language model research.
Hugging Face raises $235 million in Series D at $4.5 billion valuation – The round received contributions from major companies including Google, Amazon, NVIDIA, Salesforce, AMD, Intel, IBM, and Qualcomm . The funds will be used to acquire human resources.
Chip Huyen outlines 10 unresolved challenges for LLM – Chip Huyen, a prominent figure in AI research, highlights the top 10 challenges facing large-scale language model (LLM) development in a recent blog post. , and attracted a lot of attention.
- Illusions: Minimize the creation of inaccurate data by AI.
- Context mastery: Enhancing LLM context understanding.
- Data modalities: Incorporate different data types such as text and images.
- Efficiency: Increase the speed and affordability of LLM.
- Architecture evolution: Innovating beyond current model designs.
- Beyond GPUs: Exploring alternatives to mainstream AI training hardware.
- AI agents: Create LLMs for real-world tasks.
- Human preferences: Refine your models based on human feedback.
- Chat interface: Streamline user and LLM interaction.
- Multilingualism: Extending the LLM to languages other than English.
Addressing these challenges is important for the next generation of LLMs. As AI becomes an integral part of various fields, solving these problems will determine its future usefulness and impact. Chip Huyen's insights provide a roadmap for researchers and industry professionals in the AI domain.
Sign up for the free insideBIGDATA newsletter.
Join us on Twitter: https://twitter.com/InsideBigData1
Join us on LinkedIn: https://www.linkedin.com/company/insidebigdata/
Join us on Facebook: https://www.facebook.com/insideBIGDATANOW
