Hackers of India

Advances in the Trinity of AI: Data, Algorithms & Compute Infrastructure

By  Anima Anandkumar  on 17 Oct 2019 @ Hitb Sec Conf

Presentation Material

Presentation

Trinity of AI: data, algorithms and cloud from Anima Anandkumar

Video


 

Abstract

AI at scale requires a perfect mix of data, algorithms and computing infrastructure. I will describe how the deep-learning revolution brought synergies along with all three facets. I will then discuss about future of AI and design of algorithms to support the next generation applications.

AI Generated Summarymay contain errors

The speaker, A researcher in AI algorithms discusses various topics related to AI development and implementation. Here is a summary of the key points:

  1. Testing AI applications: The speaker emphasizes the importance of testing AI applications before launching them into the real world. This involves identifying stakeholders who will be impacted, , assessing potential wrong decisions, and finding ways to mitigate them.

  2. Advances in algorithms: The researcher notes that while data processing speeds have increased, algorithmic advancements are necessary to tackle new challenges. They express excitement about next-generation algorithms that can handle complex tasks.

  3. Reverse engineering AI systems: The speaker highlights the need for interpretability and explainability in AI systems, particularly when it comes to reverse-engineering an AI algorithm. This involves understanding why an AI system makes certain decisions and identifying potential biases or errors.

  4. AI on Mars Rover: The researcher mentions that current AI implementations on Mars Rovers focus on traditional path planning and control, but notes that incorporating modern AI algorithms will require significant advancements in computational capabilities.

  5. AI advancing itself: While the idea of AI solving all world problems is utopian, the speaker believes that AI can augment human capabilities, such as helping programmers debug code or mathematicians prove theorems. However, achieving true Artificial General Intelligence (AGI) is still a distant goal.

  6. Gorgan demo: The researcher mentions their Gorgan demo, which combines algorithms with rendering and simulations to create new possibilities. The demo is available online for public use.

Throughout the discussion, the speaker emphasizes the need for continued research and development in AI algorithms to address complex challenges and ensure responsible implementation of AI systems.