Source-linked AI summary
Human-in-the-loop Artificial Intelligence
Fabio Massimo Zanzotto
TL;DR
AI-driven automation may contract employment while learning systems extract knowledge from workers’ interactions. The paper proposes HIT-AI, which tracks knowledge provenance so revenue-generating AI decisions can repay legitimate knowledge producers, while acknowledging identification and privacy challenges.
Problem
AI may cause job-market contraction while learning systems extract workers’ knowledge without including its producers in the resulting wealth redistribution.
Method
The paper proposes HIT-AI, using explainability and knowledge-lifecycle tracking to identify knowledge sources and allocate repayment to their producers.
Results
HIT-AI is presented as a fairer AI approach that gives back a large part of machine-generated profit to the legitimate owners of the knowledge used.
Takeaways & Limitations
Responsible AI should keep knowledge producers in the loop by providing credit and revenue for knowledge used in profitable machine decisions.
Takeaways & Limitations
Implementing HIT-AI requires difficult individual identification and raises substantial privacy, technological, moral, and legal issues.
Abstract
from arXiv · showhide
Little by little, newspapers are revealing the bright future that Artificial Intelligence (AI) is building. Intelligent machines will help everywhere. However, this bright future has a dark side: a dramatic job market contraction before its unpredictable transformation. Hence, in a near future, large numbers of job seekers will need financial support while catching up with these novel unpredictable jobs. This possible job market crisis has an antidote inside. In fact, the rise of AI is sustained by the biggest knowledge theft of the recent years. Learning AI machines are extracting knowledge from unaware skilled or unskilled workers by analyzing their interactions. By passionately doing their jobs, these workers are digging their own graves. In this paper, we propose Human-in-the-loop Artificial Intelligence (HIT-AI) as a fairer paradigm for Artificial Intelligence systems. HIT-AI will reward aware and unaware knowledge producers with a different scheme: decisions of AI systems generating revenues will repay the legitimate owners of the knowledge used for taking those decisions. As modern Robin Hoods, HIT-AI researchers should fight for a fairer Artificial Intelligence that gives back what it steals.
1 Introduction
The introduction presents AI as both a source of helpful automation and a threat of mass unemployment. It argues that everyday interactions provide training knowledge and proposes HIT-AI to credit and financially repay its producers.
- AI systems are portrayed as increasingly automating everyday activities, including driving, cleaning, and conversational assistance.
- The authors warn that AI-driven automation may cause dramatic mass unemployment before an unpredictable job-market transformation.
- People’s routine online and workplace interactions generate training data that learning systems transform into machine knowledge, potentially undermining their own jobs.
- The paper proposes HIT-AI as a responsible paradigm that gives rightful credit and revenue to the people whose knowledge supports AI decisions.
2 Human-in-the-loop AI: Enabling Paradigms
This section contrasts programming with autonomous learning as ways of transferring knowledge to machines. It presents explainability and links between symbolic and distributed representations as mechanisms for keeping knowledge producers in the loop.
- 2.1 Transferring Knowledge to Machines with Programming vs. with Learning from Repeated Experience: Autonomous learning allows machines to learn from experience and addresses everyday problems that programming-based systems handle poorly.
- 2.1 Transferring Knowledge to Machines with Programming vs. with Learning from Repeated Experience: Programming clearly identifies programmers as teachers who can be paid, whereas autonomous learning extracts knowledge from data produced by unaware people.
- 2.2 Explainable Artificial Intelligence and Explainable Machine Learning: Understanding machine decisions remains an open debate, despite growing research attention to explainable machine learning.
- 2.2 Explainable Artificial Intelligence and Explainable Machine Learning: Explainable machine learning can keep people in the loop by preserving human decision authority or identifying the data sources behind machine decisions.
- 2.3 Convergence between Symbolic and Distributed Knowledge Representation: Because distributed representations approximate symbols, HIT-AI could track symbolic knowledge through the knowledge lifecycle and reward its producers.
3 Human-in-the-loop AI: a simple proposal for a better Future
The proposal frames AI’s job-market risk as a consequence of extracting workers’ knowledge while concentrating resulting revenues among machine owners. HIT-AI would track knowledge provenance so profit-making decisions repay its producers, but implementation raises identification, privacy, technological, moral, and legal challenges.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: AI systems learn from traces left by skilled and unskilled workers during ordinary jobs, creating a large-scale extraction of their knowledge.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: Machine owners may receive long-term revenues from workers’ knowledge, while the knowledge owners remain excluded from wealth redistribution.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: HIT-AI seeks to return part of machine-generated revenues to unaware knowledge producers.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: The proposal requires explainable models and complete knowledge-lifecycle tracking from training examples to machine decisions and profit recipients.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: Implementing ownership tracking requires identifying people online while addressing substantial privacy, technological, moral, and legal issues.
- 3 Human-in-the-loop AI: a simple proposal for a better Future: The proposed ecosystem includes symbiotic representations, trusted technologies, virtual identities, privacy-preserving mechanisms, and possible copyright extensions.
4 Conclusions
AI-driven job market contraction is framed as a consequence of knowledge extraction from workers. HIT-AI is proposed as a fairer approach that returns AI-generated profits to legitimate knowledge owners.
- AI systems extract knowledge from skilled and unskilled workers’ everyday interactions, contributing to job market contraction.
- HIT-AI is proposed as a fairer AI approach for addressing the knowledge theft associated with profitable artificial intelligence systems.
- HIT-AI calls for returning a large part of AI-generated profit to the workers whose knowledge supports those systems.