Every week, Invector Labs publishes a newsletter that covers 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: The Challenges of Scaling AI Startups
In recent years, the world of software was shaped by two fundamental movements: cloud computing and software as a service(SaaS). Both paradigms became the fundamental mechanisms to deliver software solutions and established new patterns for scaling companies. While before, software startups needed meaningful human and computing resources to scale, the cloud and SaaS movements removed that friction ushering a new era of innovation in company building. The natural assumption is that artificial intelligence(AI) will follow similar patterns but is that the case?
Scaling AI startups is very different than previous businesses and far from being an easy endeavor. High computing costs, professional services requirements and a constant competition from the largest tech companies in the world are some of the challenges that should be considered when evaluating building or working with AI startups. A few days ago, the famous venture capital firm Andreessen Horowitz published a brilliant essay outlining some of the challenges of scaling AI startups. AI is a completely new paradigm that requires new investment and company building theses.
Now let’s take a look at the core developments in AI research and technology this week:
Amazon Customer Service Bots
Amazon researchers published a paper describing the architecture of two neural networks used for customer service interactions
Transfer Learning with T5
Google Research published a research paper outlining a new transformer model called T5 that uses transfer learning in natural language tasks.
Developmental Psychology to Predict Object Relationships
Researchers from MIT, Harvard, DeepMind and IBM, published a paper introducing a method that leverages concepts of psychology to improve object recognition.
Cool AI Tech Releases
Einstein Reasoning Paths
Salesforce open source a framework that can retrieve evidence paragraphs from Wikipedia to answer open domain questions.
Multi-Armed Bandits Book
Microsoft Research published a new book about the famous multi-armed bandit problem for machine learning.
Google Research open sourced a new version of OpenImages, the largest annotated dataset for image analysis tasks.
AI in the Real World
AI Startups Metrics
Venture Capital firm Andreessen Horowitz published a masterpiece about the metrics for scaling AI startups.
The Pope and AI
Pope Francis joined technology giants in advocating for AI regulation.
D&D and AI
Wired published a great article outlining the role that adventure games can play in advancing AI.