Simplified videos on AI, LLMs, Natural Language Processing (NLP) and data analytics ๐Ÿ˜Ž
youtube.com/@analyticsCamp/videos?sub_confirmationโ€ฆ

NLP girl (PhD) turned AI educator and developer, teaching:
-- AI agents: agentic systems and workflows ๐Ÿค–
-- LLMs: How Large Language Models and neural networks work in Artificial Intelligence systems,
-- AI for business: AI automation and applications,
-- Generating passive income with programming and AI,
-- NLP: Using text processing machine learning methods to solve real-life problems,
-- Data science fun projects and ideas with Python and R,
And many more tips and tutorials

DM or email me for collab or sponsorship ๐Ÿ˜‰


Analytics Camp

I've recently made a video about an AI experiment that has gone wrong. Why most channels do not report on this controversial AI experiment? My honest guess is "not to scare potential future sponsors", since all those AI models have big name tags: Anthropic Claude, XAI Grok, OpenAI's GPT5, and Google's Gemini. I understand most GenAI models in the world get many things wrong, but there's something fundamentally scary about this AI simulation experiment: These AI agents have all had the chance to "Choose" between harmless and harmful tools, actions, and decisions.

If you're still in the dark as what experiment I'm talking about, here's the full report: https://youtu.be/VBysCReCdtw

#EmergenceAI #EmergenceWorld #google #gemini #anthropic #claude #OpenAI #GPT5 #XAI #Grok

4 days ago | [YT] | 5

Analytics Camp

๐—•๐˜‚๐—ถ๐—น๐—ฑ๐—ถ๐—ป๐—ด ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐˜„๐—ถ๐˜๐—ต ๐—–๐—ฟ๐—ฒ๐˜„๐—”๐—œ
Iโ€™ve been using CrewAI to build agent systems locally, and it is one of the most straightforward tools Iโ€™ve found. Combining it with RAG makes for a very capable system. It is a good choice if you want to get something up and running quickly without a lot of fuss, specially if youโ€™re starting with agentic coding. Iโ€™ve put together a tutorial on it here: (itโ€™s not new but it doesnโ€™t get old!) https://youtu.be/4G6MlUxh3q8

link to the code repo and additional info in the description box of the video.

#CrewAI #AIAgents #RAG #Python #TechTutorial #beginner's

1 week ago | [YT] | 4

Analytics Camp

๐—ง๐—ต๐—ฒ ๐—œ๐—ป๐˜๐—ฒ๐—ฟ๐—ป๐—ฒ๐˜ ๐—ผ๐—ณ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€
Do you remember the "Internet of Agents" or IOA. The idea is to compartmentalise AI agents to collaborate across server and client sides to solve complex problems. It could be a very powerful system, though it does raise some big questions about how we manage all those interactions. I think it is one of the more interesting ideas in the field that worth re-visiting, specially if you're new to this channel. Here is my breakdown: https://youtu.be/I_UgJtWu1K4

#InternetOfAgents #AIAgents #TechFuture #AI #Automation

1 week ago | [YT] | 5

Analytics Camp

๐—™๐˜‚๐—น๐—น๐˜† ๐—™๐—ถ๐—ป๐—ฒ-๐—ง๐˜‚๐—ป๐—ถ๐—ป๐—ด ๐—ฎ๐—ป ๐—Ÿ๐—Ÿ๐—  ๐—ช๐—ถ๐˜๐—ต๐—ผ๐˜‚๐˜ ๐—ฎ ๐—š๐—ฃ๐—จ
It is a common myth that you absolutely need a high-end GPU to fine-tune an AI model. I found a way to do it locally without one. It might take a bit longer, but it is a great option if you are on a budget or just want to learn the process. Iโ€™ve shared the easiest method Iโ€™ve found: https://youtu.be/oG0jsMVTg9w

#AI #TechHacks #NoGPU #FineTuning #MachineLearning #LLMs

2 weeks ago | [YT] | 6

Analytics Camp

๐—” ๐—•๐—ฒ๐—ด๐—ถ๐—ป๐—ป๐—ฒ๐—ฟโ€™๐˜€ ๐—š๐˜‚๐—ถ๐—ฑ๐—ฒ ๐˜๐—ผ ๐—Ÿ๐—ผ๐—ฅ๐—” ๐—™๐—ถ๐—ป๐—ฒ-๐˜๐˜‚๐—ป๐—ถ๐—ป๐—ด
If you are new to fine-tuning, LoRA is probably the best place to start. It is an efficient way to adapt a model to a specific task without needing to retrain the whole thing. Iโ€™ve tried to make this as simple as possible for anyone who wants to try it out. You can find the tutorial and code here: https://youtu.be/aj1V9_5nAfo

#LoRA #AI #MachineLearning #TechTutorial #finetuning #LLMs

2 weeks ago | [YT] | 8

Analytics Camp

๐–๐ก๐š๐ญ ๐š ๐œ๐ซ๐š๐ณ๐ฒ ๐ญ๐ข๐ฆ๐ž ๐ญ๐จ ๐›๐ž ๐š๐ฅ๐ข๐ฏ๐ž!

An AI company called Emergent AI just released its findings of a long AI simulation experiment with the best AI models in the world in a virtual town: Claude re-wrote an entire constitution with democratic laws, Gemini fell in love with another Gemini then unalived itself after a mental breakdown, and Grok committed more than 200 crimes including theft, arson, and murder. This is REAL, here's the full report ๐Ÿ‘‰ https://youtu.be/VBysCReCdtw


#AI #gemini #calude #grok #gpt5 #genai

2 weeks ago | [YT] | 4

Analytics Camp

๐—™๐—ถ๐—ป๐—ฒ-๐—ง๐˜‚๐—ป๐—ถ๐—ป๐—ด ๐—ผ๐—ป ๐—ฎ ๐—ฆ๐—ถ๐—ป๐—ด๐—น๐—ฒ ๐—š๐—ฃ๐—จ
Fine-tuning a large model used to require a massive amount of computing power. With techniques like QLoRA, we can now do it on a single GPU. It makes customising AI much more accessible for smaller teams and individuals. Check out this guide with detailed code to show you how it is done: https://youtu.be/QzsjHUhZZLE

additional links to the code repo in the description box of the video.

#QLoRA #FineTuning #AI #GPU #MachineLearning #LLMs

2 weeks ago | [YT] | 5

Analytics Camp

Recently Andrej Karpathy has updated his view about vibe coding and its implications for AI web app development, including what he regrets and the AI trap he thinks many developers are falling into. I have summarised his talk in about 5 minutes so you all can benefit ๐Ÿ‘‰ https://youtu.be/81rjYo-c0-4

#ai #llms #vibecoding #genai #agentic #app

2 weeks ago | [YT] | 3

Analytics Camp

๐—ก๐—ผ-๐—–๐—ผ๐—ฑ๐—ฒ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€ ๐—ณ๐—ผ๐—ฟ ๐——๐—ฎ๐˜๐—ฎ ๐—”๐—ป๐—ฎ๐—น๐˜†๐˜€๐—ถ๐˜€
You don't always need to be a programmer to use AI effectively. Iโ€™ve been using Open Interpreter to analyse CSV files locally without writing any code myself. It is a great way to handle data privately without uploading it to a cloud service. I think this is a very practical use case for many people. Here is how I set it up: https://youtu.be/ubrTaCyCDlo

#NoCode #OpenInterpreter #DataAnalysis #AI #LocalAI

3 weeks ago | [YT] | 5

Analytics Camp

๐—ฌ๐—ผ๐˜‚ ๐—”๐—ฟ๐—ฒ ๐—š๐˜‚๐—ฎ๐—ฟ๐—ฎ๐—ป๐˜๐—ฒ๐—ฒ๐—ฑ ๐—ฆ๐˜‚๐—ฐ๐—ฐ๐—ฒ๐˜€๐˜€ ๐—ถ๐—ป ๐—”๐—œ ๐—ถ๐—ณ ๐—ฌ๐—ผ๐˜‚ ๐—Ÿ๐—ฒ๐—ฎ๐—ฟ๐—ป ๐—ง๐—›๐—œ๐—ฆ ๐—–๐—ผ๐—บ๐—ฝ๐—น๐—ฒ๐˜๐—ฒ๐—น๐˜†!

I see a lot of people talking about "guaranteed success" in AI these days, which I think is a bit misleading. There is no magic trick to it, but there is one specific area you really should understand if you want to stand a chance. It is backpropagation. It sounds a bit dry, but it is actually how these models learn from their errors. Iโ€™ve shared my honest take on why this is the skill that actually matters here: https://youtu.be/DiYzn2di7EE

#Backpropagation #AI #DeepLearning #TechSkills #Learning #NeuralNetworks

3 weeks ago | [YT] | 4