​AI Tech Circle​

Hey Reader!

I am back this week after taking a break for the Eid holidays. Recently, the UAE experienced the heaviest rainfall in its history, which posed significant challenges. The UAE leadership stepped forward to address the situation, and civil service departments and volunteers worked tirelessly day and night to support the residents affected by the heavy rain.

This week started with the event 'MachinesCanSee' at the Museum of the Future, Dubai. It was quite a good gathering focused on Artificial Intelligence, and many industry leaders spoke to the audience.

This week, The #AIIndex2024 Report published through an independent initiative at the Stanford Institute for Human-Centered Artificial Intelligence (HAI), led by the AI Index Steering Committee, an interdisciplinary group of experts from across academia and industry, presents a comprehensive look at global trends in AI.

Key Takeaways:

1. AI beats humans in on some tasks but not on all
2. Industry continues to dominate frontier AI research
3. Frontier models get way more expensive
4. The US leads China, The EU, and the UK as the leading source of top AI models.
5. Robust and standardized evaluations for LLM responsibility are seriously lacking.
6. Generative AI investment skyrockets
7. The data is in: AI makes workers more productive and leads to higher-quality work
8. scientific progress accelerates even further, thanks to AI
9. the number of AI regulations in the United States sharply increases
10. People across the globe are more cognizant of AI's potential impact and more nervous

Here are a few charts that can give you the highlights of the AI Index report (source of these charts)

Adoption of AI, organizations are in the race to implement and get benefits. The survey's findings state that 55% of organizations were using AI in 2023, up from 50% in 2022 and 20% in 2017

Impact on jobs: most people expect AI to change their careers, and 36% expect AI to replace them.

The most expensive training is for Google’s Gemini Ultra, which costs an estimated $191 million worth of compute infrastructure to train, and OpenAI’s GPT-4 costs an estimated $78 million.

AI and Humans: This year’s index report tracked several benchmarks, like tasks in coding, advanced reasoning, and agentic behavior.

Generative AI is taking a significant share of the overall investment in AI from 2023

Risks from AI concern businesses worldwide. Organizations' top AI-related concerns include privacy, security, and reliability. Although organizations have started taking steps to mitigate these risks, globally, most organizations have mitigated only a small number of these risks.

You can review the AI Index from the IEEE '15 Graphs That Explain the State of AI in 2024'. The AI Index tracks the generative AI boom, model costs, and responsible AI use. The source of the above charts is available here.

---

Weekly News & Updates...

This week's AI breakthroughs mark another leap forward in the tech revolution.

1. AI Policy: The German Federal Office for Information Security publishes the report "Generative AI Models - Opportunities and Risks for Industry and Authorities. Download and go through it from here​
2. ​Meta Llama 3 is released with 8B and 70B models that deliver new capabilities, such as improved reasoning, and set a new state-of-the-art for models of their sizes. This version significantly improves Llama 2 and demonstrates SOTA performance on various industry benchmarks. The models also achieve reduced false refusal rates, improved alignment, and increased diversity in model responses. It has also enhanced reasoning, code generation, and instruction following.
3. ​The Building Blocks of AI: Decoding the Role and Significance of Foundation Models

The Cloud: the backbone of the AI revolution

• ​Oracle and Palantir Join Forces to Deliver Mission-Critical AI Solutions to Governments and Businesses
• ​Serving LLMs using HuggingFace and Kubernetes on OCI
• ​How to build a RAG solution using OCI Generative AI
• ​New infrastructure for the era of AI: Emerging technology and trends in 2024

Favorite Tip Of The Week:

Here's my favorite resource of the week.

• ​Many-shot jailbreaking research from Anthropic on a “jailbreaking” technique that developers of large language models (LLMs) can use to evade safety guardrails.
• ​But what is a GPT? Visual introduction to transformers. an excellent visual representation from 3Blue1Brown

Potential of AI

• ​Udio, is an app for music creation and sharing that allows you to generate fantastic music in your favorite styles with intuitive and powerful text-prompting.
• ​Stable Audio 2.0 is a new model that produces high-quality, full tracks with coherent musical structures that are up to three minutes long from a single prompt.

Things to Know

• ​A video on Llama 3 RAG Demo with DSPy Optimization, Ollama, and Weaviate
• Build a RAG app using MetaAI's Llama-3; access detailed steps over here​
• Reliable, fully local RAG agents with LLaMA3, watch the complete tutorial over here​
• ​Explainer: What Is Retrieval-Augmented Generation?

The Opportunity...

Podcast:

• This week's Open Tech Talks episode 132 is "Navigating the Rise of Generative AI: What Every CTO Needs to Know with Kevin Surace"

Apple | Spotify | Google Podcast | Youtube

Courses to attend:

• ​Local LLM RAG with Unstructured and LangChain [Structured JSON] is an excellent video that enables you to replicate this.
• ​Deep Learning Foundations: Large Language Models from Soheil Feizi, CS Prof at UMD, ML/AI, MIT Alum.
• ​Hands-on Data Science: Complete your First Project from Mısra Turp
• ​Advanced RAG with LlamaParse and Reranker collab lab and video

Events:

• ​GISEC, Global, 23-24 April, Dubai, UAE
• ​Oracle’s Data and AI Forum, 2nd May 2024, Theatre of Digital Art (TODA), Dubai, UAE

Tech and Tools...

• ​Aligning Diffusion Models to Human Preferences. The text-to-image model uses human feedback for state-of-the-art alignment, marking a new era in AI creativity.
• ​Uni2TS is a PyTorch-based library for research and applications related to Time Series Transformers. This library aims to provide a unified solution to large-scale pre-training of Universal Time Series Transformers
• ​Torchtune is a PyTorch-native library for quickly authoring, fine-tuning, and experimenting with LLMs.

Data Sets...

• ​Ego-Exo4D dataset for video learning & multimodal perception research. Ego-Exo4D V2 includes more videos, Aria machine perception services, and new annotations.
• ​Open datasets for OCR for extracting text from pdf/images

Other Technology News

Want to stay on the cutting edge?

Here's what else is happening in Information Technology you should know about:

• Mark Zuckerberg on Fighting in the UFC, Meta AI, Legacy, Metaverse & More. A short interview of Mark covering more comprehensive topics, worth listening over here​
• Forbes’ sixth annual ForbesAI50 companies list is published; check the complete list over here​

Earlier Edition of a newsletter:

That's it!

As always, thanks for reading.

Hit reply and let me know what you found most helpful this week - I'd love to hear from you!

Until next week,

Kashif Manzoor

---

The opinions expressed here are solely my conjecture based on experience, practice, and observation. They do not represent the thoughts, intentions, plans, or strategies of my current or previous employers or their clients/customers. The objective of this newsletter is to share and learn with the community.