The Solution To AI's Problems? Humans!
There is a simple solution to many of the problems that AI encounters from collecting data at random and giving equal weight to answers from experts, and voluminous and poliferous nut cases (Aside from Moi of course).
We need to use the resource of MILLIONS of experts who have DECADES of Experience and would enjoy telling someone they are wrong, and then backing it up. Especially if they knew they were making the world a better place, and the results were available to not just Groq, Alphabet, and ChatGPT
R - Recruiting: People with proven expertise in various fields to participate in the human-in-the-loop improvement process. Not just trolls under internet bridges or other AIs to create spam that then feedbacks into bad results.
E - Expertise: Emphasizing the experts' extensive knowledge and experience in math, science, and other technical fields to improve AI model performance. If you can't solve a set of problems that the expert would enjoy solving anyway, you probably aren't qualified to teach an AI how to solve problems.
T - Training: Providing training and guidance to users on how to effectively interact with AI models and provide high-quality feedback. Right now, we are experimenting on people who don't realize they are being experimented on, and think the AI actually knows something.
I - Insights: Integrating the Experts' insights and expertise into the AI models to improve their performance and accuracy.
领英推荐
R - Review: Regularly reviewing and refining the AI models' performance based on the Experts' feedback and expertise.
E - Evaluation: Evaluating the effectiveness of the human-in-the-loop approach and making adjustments as needed to optimize the process.
E - Engagement: Encouraging and incentivizing Experts to participate in the process, providing a sense of purpose and fulfillment. If they were compensated by the amount of time and effort they put it, it would add almost zero costs compared to the inflated values of the AI of the month, with results that are easily reproduced by the next startup with access to a million chips, which are going to flood the market next year. Or mine the year after that.
S - Support: Providing ongoing support and resources to Experts throughout the process, ensuring their success and satisfaction.
Solving this problem will take the value of AI to the next level, highlights the importance of recruiting and engaging experts to improve AI model performance in math, science, and other technical fields.
The operational excellence catalyst.
5 个月Why do you believe the Socratic Method is best for AI?
Director of Strategic Program Development at ExoAnalytic Solutions
5 个月It is important to note that power expended in the use of an AI solution is inversely proportional to the amount of human insight that can be exploited. Analytical solutions when available require very little computational power. Conversely when we don't know how to do something ... we can throw a massive model at the problem and learn something. Connecting the two so that learning more efficient solutions will be key to sustainably employing AI and enabling AI in domains where compute may be limited.