Serving the AI Gold Rush

May 9, 2018

We would like to share with you an article written by Mr. David (Dede) Goldschmidt, Vice President & Managing Director at Samsung Catalyst Fund – explaining in a very allegoric manner the contribution of allegro to the AI Gold Rush.

Allegro.AI: Serving the AI Gold Rush

David (Dede) Goldschmidt, Vice President & Managing Director, Samsung Catalyst Fund

During the California Gold Rush, prospectors flocked to the Sierra Nevada hillsides hoping to find gold. It was a risky bet, and only a few struck it rich. But there was another group of people who made a lot of money from the Gold Rush: the ones providing the picks and shovels, transportation to and from the mines, services to exchange gold for cash, and—most famously—the sturdy blue jeans the prospectors wore.

These Gold Rush infrastructure providers were necessary to enable miners to go out and look for gold. The providers of tools, supplies, and services didn’t exude the glamour or receive the same attention of the miners, but they were indispensable to the whole story.

That’s much like the way things are today in the world of artificial intelligence (AI). Companies trying to mine mountains of big data in search of AI insight “gold” are very much in the public eye. But behind the scenes are companies starting to make AI’s picks, shovels, and jeans.

Currently, developing new AI products poses serious challenges:

  • Companies spend too much time, sometimes up to 80-90% of their R&D resources, just building up custom tools and data frameworks. This greatly slows time to market and consumes resources that should focusing on building the product.
  • Frequently, these purpose-built tools lack the flexibility needed to expand into other products or future requirements. And if they are tied to the equipment they’re being run on, they may become obsolete as more advanced equipment comes into play.
  • Preparing the data used to train AI models is incredibly time- and labor-intensive. One company developing a vision-based advanced driver-assistance system (ADAS) employs more than 1,000 people whose sole job is to annotate data for its system’s ongoing training.
  • AI requires too much power and cost to build the models, preventing end products from updating themselves without high-bandwidth connections to bigger processors.
  • Customer concerns about data privacy and ownership – so much data is shipped to outside cloud services that companies rightfully worry about having control over their data.

So scalability, flexibility, infrastructure, and privacy are aspects of AI that make deployment of products at scale difficult.

Recently, we announced our investment in a company, Allegro.AI, who wants to change all this.

The founders of Allegro.AI—Nir Bar-Lev, Moshe Guttmann, and Gil Westrich—saw an opportunity to give companies a shortcut to capitalizing on advances in AI, especially those involving deep learning computer vision. The founders have applied their almost 60 collective years of leadership experience, technical expertise, and entrepreneurial spirit to Allegro.AI and its AI development and management platform.

The toolset that Allegro.AI provides is broad, covering the entire AI development lifecycle:

  • An end-to-end AI lifecycle management toolbox that allows organizations to streamline and scale the time-intensive processes of building datasets and models.
  • Automation of the labor-intensive process of labelling the data so the AI can be trained on it.
  • On-premise data processing. At a time when data privacy concerns are top-of-mind, customer data needs to stay secure and under your control. Because Allegro.AI’s tools can be run on your company’s premises and hardware, you don’t need to send the data back to an external cloud service.

Furthermore, the Allegro.AI platform addresses a specific weakness in many deployed AI-based products – static models. Their tools allow data processing on the end products themselves, continuously, based on data and context specific to the device. This gets around the typical static model required by the limitations of edge device processing, and allows the products to continue to improve and learn without the need to develop custom tools.

The Allegro.AI platform handles all the infrastructure and workflow-related aspects of deep learning, simplifying and streamlining the complex data management, modeling, development, production, and deployment of AI-based systems.

The potential impact of Allegro.AI spans multiple industries, ranging from medical imaging to automotive. It can help smart products become smarter, and it can extend AI technologies to more applications.

We’re particularly excited about this investment because, like Samsung, Allegro.AI is committed not just to developing this foundational technology, but also to building the open, collaborative ecosystem that is necessary to leverage AI for consumers in a meaningful way.

Until now, the modern miners of AI insights have been left to create their own tools and workflows. With Allegro.AI, data-economy prospectors can focus on accelerating innovation for the primary capabilities of their specific AI-based offerings, instead of building the underlying infrastructure.

And in the spirit of the California Gold Rush, here’s hoping both the prospectors and the providers of picks, shovels, and jeans can strike it rich.