The Sequence Scope: AI Inspired by Neuroscience

Weekly newsletter that discusses impactful ML research papers, cool tech releases, the money in AI, and real-life implementations.

Image for post
Image for post

The Sequence Scope is a summary of the most important published research papers, released technology and startup news in the AI ecosystem in the last week. This compendium is part of TheSequence newsletter. Data scientists, scholars, and developers from Microsoft Research, Intel Corporation, Linux Foundation AI, Google, Lockheed Martin, Cardiff University, Mellon College of Science, Warsaw University of Technology, Universitat Politècnica de València and other companies and universities are already subscribed to TheSequence.

Editorial: AI Inspired by Neuroscience

The human brain has always been considered the main inspiration for the field of artificial intelligence (AI). For many AI researchers, the ultimate goal of AI is to emulate the capabilities of the brain. That seems like a nice statement but it’s an incredibly daunting task, considering that neuroscientists are still struggling trying to understand the cognitive mechanism that powers the magic of our brains. It’s not a surprise that we are regularly seeing AI research that looks to recreate some of the capabilities of the human brain.

The idea of recreating structures that imitate many of the cognitive abilities of human cognition has been at the center of AI since its early days, but only recently have movements such as the emergence of deep learning made many of these ideas practical. In recent years, we have seen numerous deep neural network architectures that attempt to simulate many capabilities of the human brain that are still not well understood from the neuroscientific standpoint. The list is impressive: memory, common sense, abstractions, attention, imagination, and even forgetting! Just this week, researchers from Amazon published a very insightful reflection about the importance of encoding the “ability to forget” in deep neural networks.

What do you think? What abilities of human cognition should be relevant to deep learning systems? Use the comments section to share your thoughts. Share this post to get the discussion going.


🔺🔻TheSequence Scope — our Sunday edition with the industry’s development overview — is free. To receive high-quality educational content every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻

🗓 Next week in TheSequence Edge:

Edge#23: the concept of machine reading comprehension; the evaluation of the SQuAD 2.0 dataset from Stanford University; the introduction to the spaCy framework.

Edge#24: the concept of text summarization; the overview of PEGASUS, Google’s new research in abstractive text summarization; the exploration of Stanford’s CoreNLP framework.

Now, let’s review the most impactful developments in the AI industry this week.

🔎 ML Research

Better Imitation Learning

Google Research published a paper proposing an imitation learning method that can operate in low data environments ->read more on Google AI blog

Forgetting in AI

The Amazon Research team published a super insightful blog post about the importance of forgetting in modern neural networks ->read more on Amazon Research blog

AI to Play Physical Board Games

DeepMind published a remarkable paper proposing automated environments to train AI agents in board games, by understanding the physics of the environment (board, pieces, etc)->read more in the original research paper

🤖 Cool AI Tech Releases

Baidu Brain and PaddlePaddle

Baidu open-sourced the new version of its machine learning platform Baidu Brain 6.0 as well as its Paddle Paddle deep learning framework, which allows developers to build and deploy models more efficiently ->read more in Baidu’s press release

NLP in TensorFlow Lite

One of the most requested features of TensorFlow is finally here. TensorFlow Lite just added end-to-end support for NLP models ->read more on TensorFlow blog

💸 Money in AI

  • NVIDIA is going to pay $40 billion for the acquisition of processor architecture firm Arm from Softbank. In the company’s statement, Huang, CEO of NVIDIA, said, “In the years ahead, trillions of computers running AI will create a new internet-of-things that is thousands of times larger than today’s internet-of-people.” The deal still has to go through global antitrust approval, and if it does the combination of these two companies might seriously impact the distribution of the market.
  • AI gym equipment startup Tonal has raised a $110 million round. Tonal leverages A.I. to dynamically adjust the weights for each exercise in real-time for the most effective workout.
  • AI call center software startup has raised $54 million in its funding round. The team develops AI-powered solutions, which allows the extraction of information from audio conversations, accelerating the transcription and analysis of business calls, as well as automating and improving workflows.
  • Cloud-based data catalog startup Data.World raised $26 million in a venture capital funding round. Their focus is on creating a metadata management tool that helps to improve data discovery, governance, and access.
  • RapidAI, a neuroimaging stroke software platform, closed a $25 million round this week. The platform uses AI to create advanced images from CT and MRI scans, improving stroke assessment that significantly speeds up hospital work and facilitates better patient outcomes.
  • Another medtech startup powered by AI, Aidoc, raised an additional $20 million as part of its Series B extension round this week. The company provides AI solutions, which analyze and flag acute abnormalities across the body, helping radiologists prioritize life-threatening cases and expedite patient care.
  • Big data analytics service for data lakes Varada closed a $12 million round. The startup enables data architects to accelerate and optimize workloads, using dynamic analysis, adaptive indexing, and machine-learning tools to continuously track cluster performance and data usage.
  • Deepfake detection startup Sentinel has raised $1.35 million in a seed round. The team built a multi-layer system that utilizes large databases with verified deepfakes, an ensemble of neural networks classifiers and other AI tools to help governments and media organizations uncover AI-altered or manipulated images and videos.
  • Boston Dynamics, probably the most famous robotics company, shifts from the R & D phase to making its robotics business profitable. The company was founded in 1992, its current CEO Robert Playterprojects that the company will have a positive cash flow in 2023–2024.

Written by

CEO of IntoTheBlock, Chief Scientist at Invector Labs, Guest lecturer at Columbia University, Angel Investor, Author, Speaker.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store