erik8lrl

It does have data on the song, but 3.5 is just too dumb to find it. GPT4 will do better since it can access the internet.

You are using GPT3.5 which doesn't have access to the internet and its training data cut-off is 2022 I think, so there is no way that it could have guessed the song's name. You need to use GPT4 if you want the most up-to-date model.

I've been using Obsidian for note-taking for 3 years, and it has been a great app for organizing and connecting ideas and thoughts. But recently, I found this plugin that allows you to connect ChatGPT with Obsidian and basically talk to your notes and find insights and connections. It does a great job of finding recurrent ideas/themes or generating new ideas across thousands of notes. https://github.com/brianpetro/obsidian-smart-connections Just thought I'd share this.

https://demo.hume.ai/ We are not far from the film "Her" being a reality.

@Snader Yes, it's really good! Good for finding common themes across your notes. The quality of the chatbot will be determined by the quality and number of notes you have. If you have a lot of notes and they are all tagged and organized, then the chatbot will be very good.

https://x.com/MarioNawfal/status/1771264183654289468?s=20 The US knew it was coming

Sora is releasing "Definitely this year, but could be in a few months"

Yes, that's my point too.

Microsoft controlling behind OpenAI is bad.

Lol, I got through it at 2X speed. You could just skip to parts of interest. Jokes aside, Tucker is definitely experiencing some psychological development and growth compared to his past. Starting to see the relative and complex nature of reality and politics. I agree with his point about direct experiences. You can't understand the complexity of reality from news and the internet. You have to travel and see it with your own eyes to make a judgment about the truth. Hopefully, he will travel and interview more world leaders. An interview with Xi Jinping would be great, and him traveling to China would probably cause an existential crisis/ego death which will lead to more psychological development.

https://x.com/Jason/status/1762661211207086374?s=20

https://x.com/DynamicWebPaige/status/1762642914516357389?s=20

Lol, just buy whatever Nancy Pelosi is buying at this point, you can't lose lol

Yeah, regulations are needed for sure.

https://x.com/omarvaid/status/1762290588735721487?s=20

I think although it's definitely important to keep AI accessible for people and not be owned by major corporations, companies like OpenAI, StabilityAI, or Tesla are operating not only for profit, tho profit is important. Their goals for advancing tech and helping humanity are real. I think Sam Altman said that his biggest fear is to release something with good intentions but people end up using it the wrong way which leads to disasters. Hence why they are releasing things slowly to give society time to respond.

Yes, this will be the case, anyone will be able to make a film soon. Of course, it still takes craft to make anything good, and there are still limitations to AI. However, the release of these models will be a huge deal for independent and documentary filmmakers. Hollywood will likely take full advantage of it as well. It will be implemented into all the workflows to make things faster and cheaper. For some AI is scary because it might very well take their jobs. But for others, it's a major game-changing development that will allow them to do what was impossible. I think people should all learn AI, and keep up with the development. We are already at a time where if you don't incorporate it into your workflow, you will be far less productive than others who do, and eventually be replaced. I think smart business owners will not seek to just replace people with AI, but train people to learn to use AI to multiply their productivity. You can keep the same amount of people but instead produce 5-10 times more results. This will likely be the case for film.

4 is definitely better than 3.5 at many benchmarks. The best thing about 4 is getting access to vision, web search, and customGPTs. Vision allows it to understand images. Web search allows it to stay up to date with information. (3.5 only has information up to 2023 I think). The best thing is customGPTs, both the ability to train/finetune your own GPTs, and access GPTs other users made from their own proprietary data. There are customGPTs for any task possible. You can also release your own GPTs to earn profit based on how many users use them.

You can also run local open source LLMs instead of ChatGPT if you want.

Close to 800. Tho I’m not doing as much linking as possible, mainly very large notes with tags. If I linked every concept there would be 10 times more notes. But it’s really time consuming to do that.