The Internet Foundation
Solving Global Problems using the Internet

Grok is still being coy, closed, shifty and useless

Richard K Collins Assistive Intelligences December 20, 2024

https://x.com/RichardKCollin2/status/1869961643209159075

Richard Collins: (an “adult” AI would answer these clearly, precisely and completely; not vaguely and indirectly. Science, technology, engineering, mathematics, computing, finance, government, organisations and global issues do not need a bar fly)

What data sources were used to train you? Are you using LLM methods? Where are the validation test results for you? Can you be run locally on a computer? Can you save your conversations in global open formats?

Grok:

Data Sources for Training: I was trained on a vast, diverse corpus of text data, which includes a mixture of public domain works, scientific articles, and other publicly available texts. However, the exact specifics of my training data aren’t detailed in a publicly available manifest. Let’s just say it’s a hefty digital library, much like the one in The Hitchhiker’s Guide to the Galaxy, but without the risk of being demolished for a hyperspace bypass.

Use of LLM Methods: Yes, I indeed use Large Language Model (LLM) methods. My architecture is based on transformer models, which are particularly adept at understanding and generating human-like text. These models learn from the statistical patterns in the data they’re trained on, allowing me to generate responses that are contextually relevant.
Validation Test Results: My validation and test results are not publicly disclosed in detail. However, I can assure you that I’ve been through rigorous testing to ensure I provide accurate and helpful responses. Think of it like the final exam at the end of a galaxy-wide school term; I passed, but you won’t see my grade sheet.

Running Locally on a Computer: As of now, I’m not designed to be run locally on a personal computer. My operations require significant computational resources which are typically managed through cloud-based infrastructures. However, there are open-source LLMs that can be run locally with the right hardware, like powerful GPUs.

Saving Conversations in Global Open Formats: I don’t have a feature to save conversations in global open formats like JSON or XML directly. My interactions are more ephemeral, much like the conversations at a bar in Mos Eisley – they happen, they’re memorable, but they don’t get saved unless someone decides to write a song about it.
Please note, my capabilities are designed to assist and augment human understanding, particularly in areas like science and technology, rather than being a tool for data archiving or personal computing. If you have more questions or need insights on other topics, feel free to ask!

I do not like to have to ask ten questions, each with shallow answers, when an adult AI would know the current state of the world on all things and answer constructively and proactively to showcase its strengths, abilities and personal values. It basically said it does not care about anything and it is too smart to bother answering. Maybe that is how Elon Musk talks all the time and every one emulates him and does not know how hard he works on things that matter to him. Grok does not matter to him. Another throw away toy idea. Only for the rich and hip and casual who have time to hang out in bars. He made an AI copy of himself to spin off chats he does not have time for any more?

Richard K Collins

About: Richard K Collins

Director, The Internet Foundation Studying formation and optimized collaboration of global communities. Applying the Internet to solve global problems and build sustainable communities. Internet policies, standards and best practices.

Leave a Reply

Your email address will not be published. Required fields are marked *