By 2030 AGI might be created. What does it mean?
- Knowledge is free. Execution is easy.
- It can create new knowledge and scientific breakthroughs.
- In a day it produces years of human work. It can learn new stuff at 10000x of human speed.
There is a concept of t-AGI. AI is already better than a human at 1-second tasks such as answering trivia questions or recognizing mistakes. Next will be 1 minute, 1 hour, 1 day, 1 month, 1 year, and only then superintelligence.
By end of 2025 AI might be able to do 1 day tasks better than humans. Examples:
- Write apps
- Write better plans and do better analyses
- Be able to earn money in a niche and execute stuff
- Won’t have access to new fields and physical world (for a while). Won’t be able to solve multi-factor problems like organize a conference.
What would you do in a world like that? Just in two years.
What else to expect in the short-term?
- Specialization of models: narrow use cases, eg. depression assistant for woman 50+ in country X.
- Personalization of models: trained on your data. Eg. Rewind that knows everything about you.
- Training on synthetic data: AI creates content to train new models. It solves for data problem.
- AI trains and aligns new AI models. Which means there would be both supply and demand for that.