Is GPT4 capable of self-improving? Are we heading for AGI or AI doom?
18Recently, I have noticed plenty of arguments against AI and even a petition curtailing the training of LLMs bigger than GPT4. Will AI eventually lead us to our doom? Is GPT4 capable of self-improving?
I do a simple experiment on trying to find out the nth prime number as an illustration to show that the model is ultimately limited by its own generation. To improve it, it will need to be given the right teaching signal. Perhaps tools can help to improve it, but we will still be limited by the capabilities of the model and tools.
Rather than be worried by whether AGI will destroy us, I feel like we are still far from achieving AGI. AlphaZero only works on a narrow domain such as Go or chess, and it will take a long while more before we can get general self-improving algorithms. And that may not even be possible.
Furthermore, we can just avoid linking powerful AI to an action space which has negative implications for us, such as weapons. Instead of limiting technology, we can focus on how to use it to enhance our daily lives.
~~~~~~~~~~~~~~~~~~~~~~~
Related links:
Petition to pause GPT-4 experiments: https://futureoflife.org/open-letter/pause-giant-ai-experiments/
How ChatGPT works: https://www.youtube.com/watch?v=wA8rjKueB3Q
How Visual ChatGPT works (LLMs + Tools): https://www.youtube.com/watch?v=J1Xj0xXmtHU
AlphaGo/AlphaZero Introduction: https://youtu.be/a4lrFMSwGWQ
Learning, Fast and Slow (Memory to imbue better performance in AI systems): https://www.youtube.com/watch?v=Hr9zW7Usb7I
0:00 Introduction
1:05 Prime Number experiment
4:07 Discussion on Self-Improvement
5:16 Comparison to Alpha Zero
6:34 A way for improvement via tools
7:25 How to prevent AI Doom
8:23 Memory as the way ahead
9:25 Concluding note on AGI
~~~~~~~~~~~~~~~~~~~~~~~
AI and ML enthusiast. Likes to think about the essences behind breakthroughs of AI and explain it in a simple and relatable way. Also, I am an avid game creator.
Discord: https://discord.gg/fXCZCPYs
LinkedIn: https://www.linkedin.com/in/chong-min-tan-94652288/
Online AI blog: https://delvingintotech.wordpress.com/.
Twitter: https://twitter.com/johntanchongmin
Try out my games here: https://simmer.io/@chongmin