Persistence pays. Once I needed to encrypt a string, it gave me code with a hallucinated library. I came back and asked for one that wasn't hallucinated, code worked first try.
Thanks, as always! I have perplexity (and gigabrain – a great AI search for reddit) setup in firefox so I in teh task bar I type "p the_query" or "g the_query" and it gives the perplexity or gigabrain results (no logging into perplexity, clicking on a search box etc)
I think there are a couple of factors around why "features" are the big thing at the moment. 1. VC money arrived a while back. They want a return – legit usage and growth. Better models only placate them.wo long. 2. I think the really revolutionary model stuff will arrive after the US election is done. Right now is a very sensitive time.and they could end up blundering into a huge political scandal.
Self-driving computers will revolutionize productivity, but significant challenges remain in their development and implementation.
The Future of AI-Based Computing * Self-driving computers represent the ultimate interface for AI, capable of performing tasks like a remote worker. * As AI models improve, they will increasingly handle complex tasks traditionally done by knowledge workers. * Current AI technology is not perfect, but its potential is enormous as it learns to navigate desktop environments. * AI's ability to react to errors during tasks is a game-changer, enhancing its utility in real-world applications. * Equipping AI with advanced skills beyond basic mouse and keyboard functions is crucial for its effectiveness.
Security and Agency Concerns * Granting AI agency to operate computers raises security risks, but these can be managed with proper protocols. * Virtualized environments are already industry standards, suggesting that AI can safely manage tasks in these settings. * AI's ability to handle novel situations is a significant advancement over traditional automation methods. * Security concerns must be addressed as AI takes on more responsibilities in sensitive environments. * Empowering AI to manage tasks autonomously could lead to increased efficiency and productivity.
The Role of User Interfaces * Current user interfaces may not be optimal for AI, suggesting a need for AI-specific applications. * AI's reliance on existing interfaces complicates its ability to perform tasks effectively, similar to self-driving cars on complex roads. * Developing a universal API for AI could streamline its interaction with various software and systems. * AI must learn to perform tasks without excessive guidance to be truly useful in everyday applications. * Improving AI's ability to understand and execute user commands will enhance its practical applications.
Key Takeaways * Self-driving computers could transform job roles, particularly in data entry and routine tasks. * AI's ability to learn from mistakes is essential for its evolution and effectiveness. * Future AI applications will require a balance between user control and AI autonomy. * Real-world testing and iteration are critical for developing effective AI solutions. * As AI capabilities grow, so will its potential to replace or augment human jobs.
Are there any AI coding tools can have both the problem/project context and awareness of the current state of the actual code? I have been using the ChatGPT o1 models for some projects and it has helped me speed things up a fair amount. But at certain points it just forget things we talked about and rules I set up early on. It even forgets specific libraries that I use and suggest things that I specifically told it to avoid. Edit: If it had access to at least read the code at all times I think it would avoid some of those issues.
has anyone be able to get Computer Use to work in the docker container locally? I can't get it working, it would be incredibly useful for data sites that don't have APIs
Persistence pays. Once I needed to encrypt a string, it gave me code with a hallucinated library. I came back and asked for one that wasn't hallucinated, code worked first try.
great discussion:)
I am finding the 'No Yapping' addition to my prompts awesome…
Thanks, as always! I have perplexity (and gigabrain – a great AI search for reddit) setup in firefox so I in teh task bar I type "p the_query" or "g the_query" and it gives the perplexity or gigabrain results (no logging into perplexity, clicking on a search box etc)
I think there are a couple of factors around why "features" are the big thing at the moment. 1. VC money arrived a while back. They want a return – legit usage and growth. Better models only placate them.wo long. 2. I think the really revolutionary model stuff will arrive after the US election is done. Right now is a very sensitive time.and they could end up blundering into a huge political scandal.
Here is a quick digest to enhance your viewing:
Self-driving computers will revolutionize productivity, but significant challenges remain in their development and implementation.
The Future of AI-Based Computing
* Self-driving computers represent the ultimate interface for AI, capable of performing tasks like a remote worker.
* As AI models improve, they will increasingly handle complex tasks traditionally done by knowledge workers.
* Current AI technology is not perfect, but its potential is enormous as it learns to navigate desktop environments.
* AI's ability to react to errors during tasks is a game-changer, enhancing its utility in real-world applications.
* Equipping AI with advanced skills beyond basic mouse and keyboard functions is crucial for its effectiveness.
Security and Agency Concerns
* Granting AI agency to operate computers raises security risks, but these can be managed with proper protocols.
* Virtualized environments are already industry standards, suggesting that AI can safely manage tasks in these settings.
* AI's ability to handle novel situations is a significant advancement over traditional automation methods.
* Security concerns must be addressed as AI takes on more responsibilities in sensitive environments.
* Empowering AI to manage tasks autonomously could lead to increased efficiency and productivity.
The Role of User Interfaces
* Current user interfaces may not be optimal for AI, suggesting a need for AI-specific applications.
* AI's reliance on existing interfaces complicates its ability to perform tasks effectively, similar to self-driving cars on complex roads.
* Developing a universal API for AI could streamline its interaction with various software and systems.
* AI must learn to perform tasks without excessive guidance to be truly useful in everyday applications.
* Improving AI's ability to understand and execute user commands will enhance its practical applications.
Key Takeaways
* Self-driving computers could transform job roles, particularly in data entry and routine tasks.
* AI's ability to learn from mistakes is essential for its evolution and effectiveness.
* Future AI applications will require a balance between user control and AI autonomy.
* Real-world testing and iteration are critical for developing effective AI solutions.
* As AI capabilities grow, so will its potential to replace or augment human jobs.
Created with HorsyAI Chrome Extension
Thank you for using the censorship button. It makes me all fuzzy inside.
Yeah boys!
guys: you're still thinking inside the box. AI – OS. That's teh future
Are there any AI coding tools can have both the problem/project context and awareness of the current state of the actual code? I have been using the ChatGPT o1 models for some projects and it has helped me speed things up a fair amount. But at certain points it just forget things we talked about and rules I set up early on. It even forgets specific libraries that I use and suggest things that I specifically told it to avoid. Edit: If it had access to at least read the code at all times I think it would avoid some of those issues.
has anyone be able to get Computer Use to work in the docker container locally? I can't get it working, it would be incredibly useful for data sites that don't have APIs
💯% I love coding w/ cursor in the ide.