Artificial General Intelligence

AI Enablers, Artificial General Intelligence, Core AI
Machine Learning Natural Language Processing

Business Overview

AI21 was formed by AI luminaries and veterans of the Israeli technological intelligence unit, with the mission of marrying existing machine learning and NLP technologies with the knowledge representation technologies of the 1980s, giving rise to AI with an unprecedented capacity to understand, analyze and visualize complex information. The company was founded by seasoned entrepreneurs Prof. Yoav Shoham, a CS Professor Emeritus at Stanford University and Ori Goshen, ex-(Israeli) intelligence.

Operating Status
November 2017
Business model
Offering type
Funding stage
Business stage
Revenue status
Revenue Generating
Total funding
$34.5 Million
AI Enablers, Artificial General Intelligence, Core AI, Custom, Technologies


Ori Goshen
Serial Entrepreneur
Yoav Shoham
Serial Entrepreneur
AI Expert
Amnon Shashua
Serial Entrepreneur
AI Expert

Funding Rounds

Calculated as the average ratio between the current funding round amount raised and the previous funding round amount raised. Note that 'plus' rounds are summed together. i.e. Series A = Series A and Series A+.
Funding Growth Multiple

New wpDataTable

Date Announced

Funding Round

Amount Raised


AI Technology Stack

AI Description

AI21 develops the next generation of AI software “ AI that augments existing Machine Learning technology with an innovative capacity to understand, analyze and visualize abstract knowledge structures such as time, space, causality and agent-relative beliefs, intentions and preferences.

AI employees
AI application
Horizontal AI
AI types
Machine Learning, Natural Language Processing
ML types
Semi-Supervised Learning, Supervised Learning, Unsupervised Learning
AI algorithms
Bayesian Networks, Inductive Logic Programming, Neural Networks, Representation Learning, Rule-Based
AI tools
Proprietary, Pytorch/Torch
C++, Java, Python
AI hardware


Masking tokens uniformly at random constitutes a common flaw in the pretraining of Masked Language Models (MLMs) such as BERT. We show that such uniform masking allows an MLM to minimize its training objective by latching onto shallow local signals, leading to pretraining inefficiency and suboptimal downstream performance. To address...
We review the cost of training large-scale language models, and the drivers of these costs. The intended audience includes engineers and scientists budgeting their model-training experiments, as well as non-practitioners trying to make sense of the economics of modern-day Natural Language Processing (NLP).
The ability to learn from large unlabeled corpora has allowed neural language models to advance the frontier in natural language understanding. However, existing self-supervision techniques operate at the word form level, which serves as a surrogate for the underlying semantic content. This paper proposes a method to employ weak-supervision directly...

Sign up for free

  • Save profiles to your list
  • View all funding rounds
  • Advanced sorting filter access
  • Search investment rounds
  • Search exit events

Sign up for Pro

  • Export lists (CSV, Excel) of startups and investors
  • View newly added AI startups
  • View investor emails
  • View investment funds and rounds sources
  • Advanced filters, sorting and search
  • Starting at just $15 per month