Відео

Excellent video! Thank you for putting this together.

This was well done and I thoroughly, enjoyed it!

Great video, well done.

You made a video using notebookLM, perplexity in your process would you recommend using notebookLM, Gemini Deep research and Deepseek? I make UA-cam documentaries so need to do deep dives on a person/subject to help me write the script - wondering what most efficient/accurate process will be

Wow, this makes a lot of sense and clears up my confusion on this-thought there wasn’t as much transparency to work off of. Thanks for video, good observation on the response time of DeepThink (R1) comparatively to competitors as of now-still fairly good especially when considering the cost/effectiveness of this framework approach for their AI, but something to consider. And great editing.

In making the presentation which is a great one BTW, is there any useful guidance of % data sources used eg 20% UA-cam, 10% paper, etc ?

Great Presentation. I would like to get that prompt guide

Awesome, could you share the full notebook?

You are amazing, as always. Thank you..

I have found Perplexity Pro with Deepseek R1 to produce far better results than alternative searches. Thanks for showing why!

🎯 Key points for quick navigation: 00:00 *🚀 Overview of DeepSeek and its Impact* - DeepSeek is shaking up the AI industry, showing significant potential and causing major financial shifts. - The company was founded recently by former hedge fund manager Liang Wifing. - DeepSeek's R1 model is noted for its cost-effectiveness and innovative training methods. 02:00 *💡 Training Innovations and Techniques* - DeepSeek utilizes pure reinforcement learning, which allows the AI to learn through experience rather than just being told correct answers. - Their training includes an “aha moment” where the model displays self-awareness during problem-solving. - The use of a hybrid approach combining reinforcement learning with supervised fine-tuning enhances output coherence. 05:00 *🔄 Four-Stage Training Process Explained* - The training process includes four stages that build a foundation for reasoning and problem-solving abilities. - The stages involve structured guidance, reinforcement learning, rejection sampling, and aligning the model with human preferences. - This methodical approach creates a highly capable and user-friendly AI model. 07:00 *⚙️ Efficiency and Cost-Effective Innovations* - DeepSeek employs Group Relative Policy Optimization and Mixture of Experts to ensure high efficiency in training. - These techniques allow the model to focus on specific tasks without over-utilizing resources. - This results in substantial cost savings, making powerful AI accessible for less. 09:00 *📦 Model Distillation and Portability* - The technique of distillation allows DeepSeek to transfer knowledge from a large model to smaller, more accessible versions. - This makes it feasible to run intelligent AI on everyday devices with limited resources. - Improvements in performance efficiency help decrease the required hardware size and cost. 10:30 *🧠 Advanced Reasoning and Hardware Choices* - DeepSeek implements chain of thought reasoning to enhance the model's analytical capabilities. - The choice of hardware, specifically less powerful GPUs, is supported by significant infrastructure optimizations. - R1's performance is competitive, despite using more modest technological resources. 11:30 *📊 Benchmarking and Industry Implications* - DeepSeek's R1 model has achieved impressive scores on various benchmarks, particularly in mathematics. - Its pricing is significantly lower than that of competitors like OpenAI, raising questions about industry investment strategies. - This shift could prompt reevaluation of hardware reliance and pricing strategies for established AI firms. 12:30 *⏳ Observations on Performance* - Despite cost benefits, DeepSeek R1's response time is slower compared to some current models. - This aspect may affect its practical applications, despite offering robust reasoning capabilities. - Balancing speed and cost-effectiveness is crucial for future development. Made with HARPA AI

Lovely use of NotebookLM and nice presentation. This works for me thanks. 🙏👍

Honestly the best Information about Notebook lm I've heard so far. Thanks for your work. Especially the extension is so valuable for me 🙏🏽

Great video, thank you. You have a nice balance between illustration and text. You voice has excellent enthusiasm and your sentences are concise. Is there a way for me to jump to any number? Could you list the definitions by number, name and link to the video?

This is really bad on notebook lm with the images. Thank you for pointing it out. But it’s going to make my work so much harder. Hope they fix this.

Good guide.Thank you so much for these prompts - they are very helpful!

coloca seus vídeos para serem dublados amigos. o Brasil precisa de vc . put your videos to be dubbed friends. Brazil needs you

Wonderful video. I have watched a lot of videos about Notebook LM and did not hear these tips. I loved the tips about using Chrome extensions. If I create a knowledge base about a topic using these techniques how then do I share this knowledge base with my community?

ElevenLabs is costly crippleware for those building botcasts as a service. Read their terms of service. Ergo, I don't know why but this fella is avoiding discussing the serious fundamentals.

Iluminate output is flat and boring compared to the personable notebook output. Different voices do not change this and the three podcast per day limit only applies to projects you leave on their server. Ergo this fella's comments are not all that useful.

Is there a way to get it to do just one female solo voice? I used your prompt for female instead of male but it still came out male.

cool! Thank you. Can you share the prompt you used for notebook lm keep the references at the end of the answer?

So few resources on this topic. Very helpful; thanks.

Really interesting about the Google slides ocr vs pdf. That is a super helpful tip! Did you convert the pdf to Google slide format in one go? I ask because conversions from pdf to other formats often have a lot of data lost and reformatted. Great video thank you!

how did you get to perplexity gives you a lot of sources? Whan i try it gives me like 10 sourcers being pro

**Data Curation** - Effective research starts with a solid foundation, which involves gathering relevant sources like websites, documents, and videos. - NotebookLM is a powerful tool for data organization, but its effectiveness is contingent on the quality of the sources provided. - Using AI-powered search engines like Perplexity and Google Deep Research can significantly streamline the process of collecting high-quality materials. **AI Tools for Research** - Perplexity excels in understanding user needs, compiling research from multiple websites, and providing citations, although sometimes those citations may not appear in the final summary. - Google Deep Research acts similarly to a human researcher, digging deeper into web pages and following links to gather comprehensive data, albeit at a slower pace than Perplexity. - Both tools can serve as valuable assistants in preliminary research, but they lack the nuance of human judgment, necessitating manual verification of the gathered information. **Transforming Research into Presentations** - Once data is compiled, NotebookLM can convert research into professional presentations quickly, allowing users to create structured slides with detailed speaker notes. - Gamma AI enhances this process by offering various options for presentation creation, including pasting in text, generating content, or importing existing documents. - The AI capabilities of Gamma include automatic content splitting, layout adjustments, and image generation, resulting in visually coherent and professional presentations. **Comparative Analysis of AI Tools** - Both Perplexity and Google Deep Research have their strengths; Perplexity tends to deliver faster and broader results while Google provides a more detailed, albeit slower, research output. - Users should be cautious as Google Deep Research's outputs from recent years might not be applicable for all research contexts, especially rapidly evolving fields like fintech. - Cost considerations reveal Perplexity offers a free basic version, while Google Deep Research is part of a paid subscription, making it essential for users to weigh the benefits against potential expenses. **Finalizing Research Outputs** - Organizing research into smaller, manageable chunks improves accuracy and reduces errors, especially when using AI tools to assist with data gathering. - Maintaining a critical eye on the data quality from AI tools ensures that final outputs are reliable and fact-checked. - The final presentation should encapsulate essential findings while providing depth through speaker notes, ensuring a comprehensive understanding of the research topic.

about perplexity sources, you can directly upload. i know it isnt the best but it is something

I can't thank you enough ❤

I tried podcast for Hindi language using method you told ..I got website and summary in Hindi ..but podcast is generating in English....Any way to read out loud option to hear the summary instead of reading long paragraphs?

Thank for sharing

You are really genius...

This is the best video... especially I like how to remove language barriers... please share such magics...

Video is great!! I installed WebSync full site importer extension but it does nothing. What could I be doing wrong? It seems simple but....???? I go to a UA-cam video list and try to import but nothing happens.

You have done all the work testing and improving the default podcast, you achieved amazing results. THANK YOU!

Love it! This will make my custom podcasts so much more enjoyable and to the point!

Awesome

Amazing! Thank You!

And the images you generated for the presentation screams AI, and I suspect that investors who are generally very smart and skeptical people, would also assume that your presentation itself was AI generated. That reduces your credibility and also implies that you have not been doing your homework. As an investor I would immediately ask you hard questions about your methodology and how you reached your conclusions etc. if you cheat and don’t do your homework, you will be caught out. And no cigar for you.

I never had a limit on the audio generations 😮

🫡

I guess most or even all of these require notebooklm's pay-to-play version?

It helped me a lot, even for my native language, which is Spanish. Thank you!!!

Thanks for this @The_AI_News9 Sadly Illuminate isn't open to all right now. I just tried to access from London, UK and had to join the waitlist. Just scanning the comments and it appears I'm not the only one.

Finally someone tackled this riddle brilliantly! Thank you😀

Is the best workflow (for the most up to date information generation) using NotebookLM to have a library or agreggation of resources in a file inside of google drive and bulk upload (after updated)?

I’ve been one of your early followers, your videos are so helpful. I know that you will become viral one day 🙏. Thank you.

The most comprehensive NotebookLM playlist on the web ! - ua-cam.com/play/PL-HkokgcYrl5SrKYeVo28JA4OMPbslhA8.html Please help grow this small channel and community! Spread the word by sharing the playlist, Leave your comments, Contribute and learn from others through comments! Please consider subscribing if you found the content useful - www.youtube.com/@The_AI_News9/?sub_confirmation=1 !

The most comprehensive NotebookLM playlist on the web ! - ua-cam.com/play/PL-HkokgcYrl5SrKYeVo28JA4OMPbslhA8.html Please help grow this small channel and community! Spread the word by sharing the playlist, Leave your comments, Contribute and learn from others through comments! Please consider subscribing if you found the content useful - www.youtube.com/@The_AI_News9/?sub_confirmation=1 !

The most comprehensive NotebookLM playlist on the web ! - ua-cam.com/play/PL-HkokgcYrl5SrKYeVo28JA4OMPbslhA8.html Please help grow this small channel and community! Spread the word by sharing the playlist, Leave your comments, Contribute and learn from others through comments! Please consider subscribing if you found the content useful - www.youtube.com/@The_AI_News9/?sub_confirmation=1 !