Last Week in AI
Every week, my team at Invector Labs publishes a newsletter to track the most recent developments in AI research and technology. You can find this week’s issue below. You can sign up for it below. Please do so, our guys worked really hard on this:
From the Editor: AI Conquers the World’s Most Complex Poker Game
Poker remains one of the most challenging games for artificial intelligence(AI) agents. Among the things that make poker so challenging is that, at any given point, no player can plan a complete strategy as those highly depend on the cards of the other players. In AI theory, this is what is known as an incomplete-information environment. Among the variations of poker, multiplayer game of no-limit Texas Hold ’Em stands on a league of its own in terms of complexity. The game includes six players that have to constantly balance diverse options such as checking, raising or folding without knowing the actions of the other players. From the perspective of AI, this is nothing short of a nightmare.
This week, researchers at Carnegie Mellon University in Pittsburgh and the Facebook artificial intelligence lab in New York City unveiled an AI systems called Pluribus that was able to beat five world top poker players including four-time world champion Darren Elias. Pluribus mastered human-like skills like bluffing or calling a bluff and all that against five other human players. From the AI perspective, the skills shown by Pluribus can have practical applications in trading, auctions or political negotiations. This is no small milestone for AI.
Now let’s take a look at the core developments in AI research and technology this week:
Consciousness can spark the deepest passions in the AI community. AI researchers from Columbia University have bee working on deep learning models that can rebuild themselves which sorts of represents a mathematical form of consciousness.
>Read more in this article from Quanta Magazine
Ethics is one of the fundamental challenges of the next decade of AI. OpenAI published a thoughtful paper about the path to build responsible AI.
>Read more in this blog post from OpenAI
Data labeling and augmentation is the biggest challenge of modern AI solutions. Google AI researchers published a paper that showcases how to use unsupervised and semi-supervised learning for labeling and augmenting training datasets.
>Read more in this blog post from Google Research
Cool AI Tech Releases
Facebook and Carnegie Mellon teamed up to build an AI system that could beat humans in 6-player poker.
>Read more in this blog post from the Facebook AI Research team
The intersection of AI and blockchain is one of the most fascinating areas of research these days. Microsoft Research just open source framework that leverages blockchain to collaboratively build and train machine learning models.
>Read more in this blog post from Microsoft Research
Model serving remains an important challenge in most machine learning solutions. Google published a clever architecture of how to leverage Google Functions to streamline model serving.
>Read more in this blog from the Google Cloud engineering team
AI in the Real World
The race between the United States and China to dominate AI is on. The outcome of that race will be central to the global politic and economic power. Scientific American just published a thoughtful analysis about this dynamic.
>Read more in this article about Scientific American
Cloudera’s general manager Hillary Mason discusses how companies can avoid ruining machine learning projects.
>Read more in this coverage from VentureBeat
Former foreign and finance minister of Mexico, is partnering with MIT to spearhead an effort to launch international AI policies.