AIGuys Digest | August 2024
🌟 Welcome to the AIGuys Digest Newsletter, where we cover State-of-the-Art AI breakthroughs and all the major AI news🚀. In this thrilling edition of August 2024, we’re diving headfirst into the ever-evolving universe of Artificial Intelligence. 🧠✨
Don’t forget to check my new book on AI, it covers a lot of AI optimizations and hands-on code:
🔍 Inside this Issue:
- 🤖 Latest Breakthroughs: This month it’s all about Agents, LangChain RAG, and LLMs evaluation challenges.
- 🌐 AI Monthly News: Discover how these stories are revolutionizing industries and impacting everyday life: EU AI Act, California’s Controversial SB1047 AI regulation act, Drama at OpenAI, and possible funding at OpenAI by Nvidia and Apple.
- 📚 Editor’s Special: This covers the interesting talks, lectures, and articles we came across recently.
Let’s embark on this journey of discovery together! 🚀🤖🌟
Follow me on Twitter and LinkedIn at RealAIGuys and AIGuysEditor.
Latest Breakthroughs
Are Agents just simple rules? Are Agents just enhanced reasoning? The answer is yes and no. Yes, in the sense that agents have simple rules and can sometimes enhance reasoning capabilities compared to a single prompt. But No in the sense that agents can have a much more diverse functionality like using specific tools, summarizing, or even following a particular style. In this blog, we look into how to set up these agents in a hierarchal manner just like running a small team of Authors, researchers, and supervisors.
How To Build Hierarchical Multi-Agent Systems?
TextGrad. It is a powerful framework performing automatic “differentiation” via text. It backpropagates textual feedback provided by LLMs to improve individual components of a compound AI system. In this framework, LLMs provide rich, general, natural language suggestions to optimize variables in computation graphs, ranging from code snippets to molecular structures. TextGrad showed effectiveness and generality across various applications, from question-answering and molecule optimization to radiotherapy treatment planning.
TextGrad: Improving Prompting Using AutoGrad
The addition of RAG to LLMs was an excellent idea. It helped the LLMs to become more specific and individualized. Adding new components to any system leads to more interactions and its own sets of problems. Adding RAG to LLMs leads to several problems such as how to retrieve the best content, what type of prompt to write, and many more.
In this blog, we are going to combine the LangChain RAG with DSPy. We deep dive into how to evaluate the RAG pipeline quantitatively using RAGAs and how to create a system where instead of manually tweaking prompts, we let the system figure out the best prompt.
How To Build LangChain RAG With DSPy?
As the field of natural language processing (NLP) advances, the evaluation of large language models (LLMs) like GPT-4 becomes increasingly important and complex. Traditional metrics such as accuracy are often inadequate for assessing these models’ performance because they fail to capture the nuances of human language. In this article, we will explore why evaluating LLMs is challenging and discuss effective methods like BLEU and ROUGE for a more comprehensive evaluation.
The Challenges of Evaluating Large Language Models
AI Monthly News
AI Act enters into force
On 1 August 2024, the European Artificial Intelligence Act (AI Act) enters into force. The Act aims to foster responsible artificial intelligence development and deployment in the EU. The AI Act introduces a uniform framework across all EU countries, based on a forward-looking definition of AI and a risk-based approach:
- Minimal risk: most AI systems such as spam filters and AI-enabled video games face no obligation under the AI Act, but companies can voluntarily adopt additional codes of conduct.
- Specific transparency risk: systems like chatbots must clearly inform users that they are interacting with a machine, while certain AI-generated content must be labelled as such.
- High risk: high-risk AI systems such as AI-based medical software or AI systems used for recruitment must comply with strict requirements, including risk-mitigation systems, high-quality of data sets, clear user information, human oversight, etc.
- Unacceptable risk: for example, AI systems that allow “social scoring” by governments or companies are considered a clear threat to people’s fundamental rights and are therefore banned.
EU announcement: Click here
California AI bill SB-1047 sparks fierce debate, Senator likens it to ‘Jets vs. Sharks’ feud
Key Aspects of SB-1047:
- Regulation Scope: Targets “frontier” AI models, defined by their immense computational training requirements (over 10²⁶ operations) or significant financial investment (>$100 million).
- Compliance Requirements: Developers must implement safety protocols, including the ability to immediately shut down, cybersecurity measures, and risk assessments, before model deployment.
- Whistleblower Protections: Encourages reporting of non-compliance or risks by offering protection against retaliation.
- Safety Incident Reporting: Mandates reporting AI safety incidents within 72 hours to a newly established Frontier Model Division.
- Certification: Developers need to certify compliance, potentially under penalty of perjury in earlier drafts, though amendments might have altered this.
Pros:
- Safety First: Prioritizes the prevention of catastrophic harms by enforcing rigorous safety standards, potentially safeguarding against AI misuse or malfunction.
- Incentivizes Responsible Development: By setting high standards for AI model training, the company encourages developers to think critically about the implications of their creations.
- Public Trust: Enhances public confidence in AI by ensuring transparency and accountability in the development process.
Cons:
- Innovation Stagnation: Critics argue it might stifle innovation, especially in open-source AI, due to the high costs and regulatory burdens of compliance.
- Ambiguity: Some definitions and requirements might be too specific or broad, leading to legal challenges or unintended consequences.
- Global Competitiveness: There’s concern that such regulations could push AI development outside California or the U.S., benefiting other nations without similar restrictions.
- Implementation Challenges: The practicalities of enforcing such regulations, especially the “positive safety determination,” could be complex and contentious.
News Article: Click here
Open Letter: Click here
MORE OpenAI drama
OpenAI co-founder John Schulman has left the company to join rival AI startup Anthropic, while OpenAI president and co-founder Greg Brockman is taking an extended leave until the end of the year. Schulman, who played a key role in creating the AI-powered chatbot platform ChatGPT and led OpenAI’s alignment science efforts, stated his move was driven by a desire to focus more on AI alignment and hands-on technical work. Peter Deng, a product manager who joined OpenAI last year, has also left the company. With these departures, only three of OpenAI’s original 11 founders remain: CEO Sam Altman, Brockman, and Wojciech Zaremba, lead of language and code generation.
News Article: Click here
Apple and Nvidia may invest in OpenAI
Apple, which is planning to integrate ChatGPT into iOS, is in talks to invest. Soon after, Bloomberg also reported that Apple is in talks but added that Nvidia “has discussed” joining the funding round as well. The round is reportedly being led by Thrive Capital and would value OpenAI at more than $100 billion.
News Article: Click here
Editor’s Special
- The AI Bubble: Will It Burst, and What Comes After?: Click here
- Eric Schmidt Full Controversial Interview on AI Revolution (Former Google CEO): Click here
- AI isn’t gonna keep improving Click here
- General Intelligence: Define it, measure it, build it: Click here
🤝 Join the Conversation: Your thoughts and insights are valuable to us. Share your perspectives, and let’s build a community where knowledge and ideas flow freely. Follow us on Twitter and LinkedIn at RealAIGuys and AIGuysEditor.
Thank you for being part of the AIGuys community. Together, we’re not just observing the AI revolution; we’re part of it. Until next time, keep pushing the boundaries of what’s possible. 🚀🌟
Your AIGuys Digest Team