The article "Revelations: A Decidable Class of POMDP with Omega-Regular Objectives" won an Outstanding Paper Award at the AAAI 2025 conference, a prestigious international conference about artificial intelligence. This year, only three papers received such an award out of 3,000 accepted and 12,000 submitted! This recognition crowns the results of research initiated in Bordeaux (France) within the Synthèse team at the Bordeaux Computer Science Research Laboratory (LaBRI), where four of the authors work: Marius Belly, Nathanaël Fijalkow, Hugo Gimbert, and Pierre Vandenhove. The work also involved researchers from Paris (Florian Horn) and Antwerp (Guillermo A. Pérez). The article is freely available on arXiv, and this post outlines its main ideas.

If you had a PC in the 2010s, you probably owned a copy of Purble Place. The gaudy kids' game came with every copy of Windows Vista and 7. It was a simple, three-title package: Purble Pairs was a basic tile memory game; Purble Shop had the player design a mystery character using logic and deduction; and the last game of Comfy Cakes had kids playing line cook for the Purble Chef while juggling orders on a conveyor belt. And for many online teens, the legacy of these games easily equals that of Minesweeper and Solitaire, the more venerable pack-in games of PCs past. Yet nobody knows who made it.

Last week, the New York Times noticed something that federal regulators hadn't about Tesla: that thanks to a software update this summer, occupants can now play video games on the dashboard console while the car is in motion. This would seem to be the latest example of Tesla charging ahead with a feature that could potentially endanger people in and outside its cars, and sure enough, the feds are now looking into the matter. I was also curious about the feature. So on Sunday, I went to a Maryland Tesla showroom to test drive a Model 3 while gaming, a stunt that other Tesla owners have demonstrated on YouTube. The showroom was located in a mall, which had a parking lot where I felt safe enough trying to maneuver the car.

Many Tesla vehicles allow drivers to play a selection of games on the infotainment system while the car is in motion, according to a report by The New York Times. The company rolled out an update in the summer that reportedly let drivers play Solitaire, jet fighter game Sky Force Reloaded and strategy title The Battle of Polytopia: Moonrise while on the road. The touchscreen is said to display a warning before a game of Solitaire starts. "Solitaire is a game for everyone, but playing while the car is in motion is only for passengers," the message reads, according to the Times. That indicates Tesla knows the game is playable while the car's moving.

Morpion Solitaire is a popular single player game, performed with paper and pencil. Due to its large state space (on the order of the game of Go) traditional search algorithms, such as MCTS, have not been able to find good solutions. A later algorithm, Nested Rollout Policy Adaptation, was able to find a new record of 82 steps, albeit with large computational resources. After achieving this record, to the best of our knowledge, there has been no further progress reported, for about a decade. In this paper we take the recent impressive performance of deep self-learning reinforcement learning approaches from AlphaGo/AlphaZero as inspiration to design a searcher for Morpion Solitaire. A challenge of Morpion Solitaire is that the state space is sparse, there are few win/loss signals. Instead, we use an approach known as ranked reward to create a reinforcement learning self-play framework for Morpion Solitaire. This enables us to find medium-quality solutions with reasonable computational effort. Our record is a 67 steps solution, which is very close to the human best (68) without any other adaptation to the problem than using ranked reward. We list many further avenues for potential improvement.

When it comes to competitive games, AI systems have already shown they can easily mop the floor with the best humanity has to offer. But life in the real world isn't a zero sum game like poker or Starcraft and we need AI to work with us, not against us. That's why a research team from Facebook taught an AI how to play the cooperative card game Hanabi (the Japanese word for fireworks), to gain a better understanding of how humans think. Specifically, the Facebook team set out to instill upon its AI system the theory of mind. "Theory of mind is this idea of understanding the beliefs and intentions of other agents or other players or humans," Noam Brown, a researcher at Facebook AI, told Engadget.

So your Marshmallow-or-better Android phone just got Google Assistant, and when you long-press the Home button for the first time after the update, your new digital helper asks, "Hi, how can I help?" Read on for eight easy ways to get started with Google's chatty servant, from telling it what to call you, to making a shopping list, to turning on Bluetooth and even playing Solitaire. "Open the Kindle app," "make an appointment for noon today," "remind me to buy milk," and "play some jazz"--all familiar Android voice commands from way back when, and most of them will work just fine with Google Assistant. If you're not sure where to start with Google Assistant, try a familiar Android voice command--anything from "create a calendar event" to "set a timer." Just tap and hold the Home button, wait for Google Assistant to pop up, and then say something like, "send a text message to my wife," "send a mail message," or "what's the fastest way home?"

Tucked within Google's unremarkable Allo messaging app is a real treasure: Google Assistant, which injects Google Now with an eager-to-please personality that finally provides the give-and-take other digital assistants lack. We've always talked about Apple's Siri, Microsoft's Cortana, and Google Now as the three digital assistants from the top smartphone platforms. But the truth is that Google Now was little more than a series of informative cards, while Siri and Cortana preferred a text-based approach with a bit of sass. Google Assistant retains its visual approach, but within a messaging context that really nails it in how you interact with the app itself. Google announced Google Assistant this past May, and the preview version of it is live in Allo, which itself can be used on Android 4.1 (Jelly Bean) on up.

Monte Carlo tree search (MCTS) methods have had recent success in games, planning, and optimization. MCTS uses results from rollouts to guide search; a rollout is a path that descends the tree with a randomized decision at each ply until reaching a leaf. MCTS results can be strongly influenced by the choice of appropriate policy to bias the rollouts. Most previous work on MCTS uses static uniform random or domain-specific policies. We describe a new MCTS method that dynamically adapts the rollout policy during search, in deterministic optimization problems. Our starting point is Cazenave's original Nested Monte Carlo Search (NMCS), but rather than navigating the tree directly we instead use gradient ascent on the rollout policy at each level of the nested search. We benchmark this new Nested Rollout Policy Adaptation (NRPA) algorithm and examine its behavior. Our test problems are instances of Crossword Puzzle Construction and Morpion Solitaire. Over moderate time scales NRPA can substantially improve search efficiency compared to NMCS, and over longer time scales NRPA improves upon all previous published solutions for the test problems. Results include a new Morpion Solitaire solution that improves upon the previous human-generated record that had stood for over 30 years.

Despite its ubiquitous presence, very little is known about the odds of winning the simple card game of Klondike Solitaire. The main goal of this paper is to investigate the use of probabilistic planning to shed light on this issue. Unfortunatley, most probabilistic planning techniques are not well suited for Klondike due to the difficulties of representing the domain in standard planning languages and the complexity of the required search. Klondike thus serves as an interesting addition to the complement of probabilistic planning domains. In this paper, we study Klondike using several sampling-based planning approaches including UCT, hindsight optimization, and sparse sampling, and establish lower bounds on their performance. We also introduce novel combinations of these approaches and evaluate them in Klondike. We provide a theoretical bound on the sample complexity of a method that naturally combines sparse sampling and UCT. Our results demonstrate that there is a policy that within tight confidence intervals wins over 35% of Klondike games. This result is the first reported lower bound of an optimal Klondike policy.