Welcome to our monthly digest, where you can catch up with any AIhub stories you may have missed, peruse the latest news, recap recent events, and more. This month, we find out about aligning generative AI with technical standards, learn how machine learning can be applied to semiconductor manufacturing, investigate social choice problems, and hear about the return of the AI Song Contest. We continued our series meeting the AAAI/SIGAI Doctoral Consortium participants in this interview with Joseph Marvin Imperial. Joseph is based at the University of Bath, focusing on aligning generative AI with technical standards for regulatory and operational compliance. Amina Mević is another AAAI/SIGAI Doctoral Consortium participant, working at the intersection of machine learning, physics, mathematics, and semiconductor technology.

In their paper Reducing Leximin Fairness to Utilitarian Optimization, Eden Hartman, Yonatan Aumann, Avinatan Hassidim and Erel Segal-Halevi present a scheme for addressing social choice problems. In this interview, Eden tells us more about such problems, the team's methodology, and why this is such a fascinating and challenging area for study. The paper looks at social choice problems -- situations where a group of people (called agents) must make a decision that affects everyone. For example, imagine we need to decide how to divide an inheritance among several heirs. Each agent has their own preferences over the possible outcomes, and the goal is to choose the outcome that is "best" for society as a whole.

We consider a classic social choice problem in an online setting. In each round, a decision maker observes a single agent's preferences overa set of $m$ candidates, and must choose whether to irrevocably add a candidate to a selection set of limited cardinality $k$. Each agent's (positional) score depends on the candidates in the set when he arrives, and the decision-maker's goal is to maximize average (over all agents) score. We prove that no algorithm (even randomized) can achieve an approximationfactor better than $O(\frac{\log\log m}{\log m})$. In contrast, if the agents arrive in random order, we present a $(1 - \frac{1}{e} - o(1))$-approximatealgorithm, matching a lower bound for the off-line problem.We show that improved performance is possible for natural input distributionsor scoring rules. Finally, if the algorithm is permitted to revoke decisions at a fixedcost, we apply regret-minimization techniques to achieve approximation $1 - \frac{1}{e} - o(1)$ even for arbitrary inputs.