- I'm often asked, how to keep up? LLM genAI is evolving faster than HTML in the mid-90s. I constantly talk with people: developers, founders, investors, etc.; I go to events; I speak at events, I teach university courses; and I talk with lots of other people.
- If you see a useful article or whatever about AI, let me know. I'll look at it and if it fits, I'll add it.
- I post notes, updates, and stuff about AI at Blue Sky at @Andreas.com, in Twitter at @Andreas_Ramos and LinkedIn.
- Clément Schneider, a former student, is building multi agent systems (MAS) in Paris for companies. Talk with him.
My eBooks and Webpages
- How an LLM AI works. Written for students. Three pages in clear English.
- How to create Multi Agent Systems MAS. By me. Free ebook.
- How to Write Prompts for AI. By me. Free ebook.
- How to make a GPT AI with OpenAI ChatGPT-4. By me. Free ebook.
- How I added an AI chatbot to my website.
News and Events in AI
- Twitter is the main place for AI news. Stuff is announced or discussed in tweets. Here's my list of useful AI accounts in Twitter. BTW, I moved to Blue Sky at @Andreas.com
- See the OpenAI Forum for community discussion. Ask questions. Get answers. Go community.openai.com.
- AI Newsletters: AlphaSignal, The Deep View, and The Rundown AI, the Batch, Mindstream, Neuron, and oneusefulthing.org.
- Jeremiah Owyang's List of AI Events. Over 300 meetings or conferences every month in Silicon Valley (Palo Alto and San Francisco). There are also events in New York NYC, Seattle, and London. Come to SV for five days and go to lots of events. You'll learn more in five days here than a year elswhere.
- Agenda Hero has additional events in Silicon Valley.
Useful Books and Articles about AI
- God, Human, Animal, Machine by Meghan O’Gieblyn. Intelligence and sentience have a complex history in philosophy, theology, and science. What do these mean? The best book on the origins and implications of these ideas.
- Tim Lee explains LLMs in clear language. Large Language Models Explained. This is a Must-Read to get a basic understanding of AI LLM.
- A survey of 2,778 AI Researchers (January 2024) who had published peer-reviewed research in the prior year in six top AI venues (NeurIPS, ICML, ICLR, AAAI, IJCAI, JMLR) on their predictions of AI.
- What We’ve Learned From A Year of Building with LLMs: A practical guide to building successful LLM products, covering the tactical, operational, and strategic by Eugene Yan et al (June 2024).
- Leopold Aschenbrenner's essay Situational Awareness. On AI, Silicon Valley, and militarization of AI.
- If you see something useful, let me know and I'll add it.
My Courses for Digital Marketing and AI
- Courses in Digital Marketing with AI. I teach on the use of AI in digital marketing.
- Courses in SEO with AI and digital ads with AI.
- These university-level courses are hands-on and practical. I show the prompts and results. You get everything: code, examples, links, etc. I've been teaching since January 2023.
The Top LLM GenAI for Chat and Images (July 16,2024)
- The current top LLM AIs are Anthropic Sonnet (free), Anthropic Claude-3 ($20/month), OpenAI ChatGPT-4o ($20/month), Microsoft Copilot, Meta AI , Google Gemini. Tip: Microsoft Copilot in Creative mode uses GPT-4 for free.
- The maximum input context window (tokens) (how much you can enter into one AI chat): Claude 3.5 and Sonnet = 200,000, ChatGPT-4o = 128,000, Copilot: = 128,000, Gemini = 32,000 (Gemini Pro = 1 million), Meta AI = 4,096. This changes constantly (July 2924). The maxium output token window: Claude and Sonnet: 4,096, ChatGPT-4o: 4,096 to 8,192, Copilot 4,096 to 8,192, Meta AI: not listed, Gemini Pro 1.5: 8,000 tokens.
- These AI have a similar chat interface but there are differences in options, quality, and capabilities. These are called “foundation models” or "frontier models". Some 70,000 AI tools are built on top of these (such as Perplexity, etc.).
- Rayan also adds Mistral, an AI from France.
- The top image AI: Dall-e, Microsoft Copilot, Microsoft Designer, ImageFX, and Meta Image. Try also Pikaso
- Perplexity.ai is also a very good AI for search. It combines its own LLM AI, along with OpenAI ChatGPT, Claude-3, Mistral, and Llama-3 AIs with search data from the Microsoft Bing search engine.
- This list changes frequently. ChatGPT-5 may appear in the next few months. I'll try to keep this up-to-date.
- If you see something useful, let me know and I'll add it.
Multi-Agent System (MAS)
- Multi-Agent System (MAS): Combine multiple AI persona to create virtual teams with roles such as CEO, CTO, CFO, directors, managers, reviewers, testing, documentation, and so on. The team of AI persona collaborates to process tasks and produce results.
- What can an MAS do? Here are two examples of using an MAS to create a translations company or a software company. See arxiv.org/pdf/2405.11804 and arxiv.org/abs/2307.07924.
- What does this mean for you? You'll be able to create an MAS agent to duplicate what you do. It can do your work, learn for you, attend meetings on your behalf, make decisions, receive and send messages, and produce reports, software, translations, and so on.
- You'll be able to buy or rent MAS agents. Need software to do something in your company? Rent an MAS, describe what you need, and it makes custom software (with design, code, testing, and user manual) in 410 seconds (6.8 minutes) at a cost of US$0.30 (thirty cents). (Think about the implications for the software industry and the hundreds of thousands of programmers.)
- An MAS won't be just five or ten agents. Langchain expects one MAS may have 200 to 300 agents, each with specialized roles, tasks, and responsibilites.
- Whatever can be done remotely can be turned into an MAS.
- What's the advantage of an MAS? Chat-based LLM is restricted by the user's ability to use chat LLM, along with the user's knowledge, creativity, speed, and time. Agentic AI doesn't have those limits.
- How to create Multi Agent Systems MAS By me. Free ebook.
- Interesting papers on MAS: https://arxiv.org/abs/2405.18111, https://arxiv.org/abs/2404.17017, https://arxiv.org/abs/2309.03736, https://arxiv.org/abs/2403.07769, https://arxiv.org/abs/2402.03578, https://arxiv.org/abs/2402.01108, https://arxiv.org/abs/2402.01968, https://arxiv.org/abs/2405.12472
- See 800 research papers on MAS at arxiv.org
- If you see something useful, let me know and I'll add it.
Libraries and Frameworks for AI Projects
Yohei Nakajima @yoheinakajima asked AI builders of their favorite go-to libraries and frameworks for AI projects. A Hive Mind Survey Summary.
MAS Multi Agent System Frameworks:
- AgentOpsAI: This tool simplifies the debugging and monitoring of MAS multi agent system operations, making it easier to read through prompts and visualize spend. @AlexReibman
- AgentForge: Designed for rapid iteration of cognitive architectures with support for multiple LLM APIs per prompt and easy VectorDB implementation via ChromaDB. @JohnSmith4Reel
- crewAIInc: A flexible and powerful framework for enabling MAS multi agent system task completion with support for any OpenAI API-compatible endpoints. @whoabuddydev, @as_cybersamurai
- craftgen: Offers flexibility with its actor model and event-driven architecture, making human-in-the-loop workflows easy. @Necmttn
- Lyzr: An MAS multi agent system framework mentioned for its capabilities in building multi-agent systems. @theAIsailor
- Policy Synth: An open-source multiscale MAS multi agent system library. @robertbjarnason
- GraphAI: Used for data flow programming to build genetic apps, enabling LLMs to generate genetic applications. @snakajima