All we are is words...
A newsletter on the commercialization of natural language understanding
Welcome to our first newsletter!!
This (weekly) newsletter exists to track the progress of natural language understanding (NLU) first companies and the impacts / opportunities they are having / creating commercially and socially.
How do we define NLU? An umbrella term of natural language processing (NLP), generation (NLG), querying (NLQ), text to speech (TTS) and speech to text (STT).
Quick step back for overview...For some time now we've been trying to create a non-human entity...box...thing...machine that can hear us, understand us and speak back to us, in a manner that is indistinguishable from a human. This was the original framing of the Turing test. No humanoid like figure to go along with the hearing, understanding and talking. Just a partition that separates a human from the indistinguishable non human that is talking back to us and the question of whether we could tell if it was human or not. In constrained scenarios it is certainly true that some current machines would pass this test. So why the focus on NLU first companies in particular and not NLU in general?
The Turing test or the singularity for that matter, are milestones that distort the attention and actual impact technical advances in artificial intelligence have had and will have. Does it really matter if we have machines as or more intelligent than us, if for some time prior to that point this capability has completely transformed the world economy? It is entirely possible that a vast portion of the current employment opportunities will disappear way before we even reach those milestones if indeed they are at all possible. Much value will be created and destroyed prior to these possible future events and we think this is ultimately a product of commercial ventures.
What is an NLU first company? While NLU capabilities are part of many businesses, this newsletter will focus on those companies whose business model is primarily dependent on some aspect of NLU. So while AWS, Azure and others make NLU capabilities available, it is not the core part of their business model and won't be a focus here. While these companies make certain capabilities ubiquitous, those capabilities are usually removed from the leading edge offerings of these NLU first companies. Rather it is our thesis that those companies, private and public, that are primarily focusing on NLU in their offering are the companies that will have the biggest impact on advancing the technology and impacting the daily lives of the people on this planet by way of our human-machine experiences or economic change and disruption. These companies also frankly offer the best opportunity to access the value generation to be created by commercial NLU. Additionally while non-commercial entities like Open AI and their GPT-3 solution are legit amazing (!!!), until it is commercially leveraged (and it is starting to be), it’s broader societal impact is minimal, so we won’t focus on these types of institutions either.
So how will we track & measure progress and the related impacts? Here are a few metrics we’ll be tracking at a minimum:
How many NLU first companies have we identified (if we miss some and we will, let us know)? Any new this week?
Where are they in their business maturity arc i.e. early stage start up, late stage, post start up, public etc? Any notable changes this week?
How much money have they raised? What raising occurred this past week?
Primarily though, beyond our metrics we’ll track:
Commercial happenings - What interesting commercial happenings occurred this week with links to the source articles
Technical happenings - What interesting technical happenings occurred this week with links to the source articles
Ethical or broader domain hot spots and points of interest - Finally any interesting ethical topics / debates or otherwise interesting articles that came up this week worth highlighting
A company or topic deep dive including:
What are the primary NLU capabilities of these companies?
What is the value to capability relationship?
Notable adoption and displacement impacts, either historical or of note this week
Basic market data we’ll track each week:
Happenings this week:
Commercial:
Cryptocurrency exchange platform Coinbase has acquired AI-powered support platform Agara, which has its operations in India and the US at an undisclosed amount. This is Coinbase’s first start-up acquisition in India. This acquisition is aimed at reinforcing Coinbase’s “commitment to delivering world-class support for customers” and providing Agara’s expertise in machine learning (ML) and natural language processing (NLP) to the company’s engineering team…read
QuestionPro, a global leader in online survey and research services has acquired Bryght AI, a “conversational intelligence” platform that helps…read
Walden Catalyst has announced that it has invested $20 million in Israeli deep tech company AI21 Labs. Founded by Israeli serial entrepreneurs Prof. Amnon Shashua, Prof. Yoav Shoham, and Uri Goshen, A121 Labs has developed Wordtune, which was launched about a year ago and already used by almost million users…read
This morning voice-powered doctor’s assistant Notable announced a whopping $100 million Series B funding round. ICONIQ Growth led the round with participation from Greylock, F-Prime Capital and Oak HC/FT…read
Laiye, planning to raise $100-150m in new funding round…read
AI Upskilling Innovator, Ahura AI Raises $3M Seed Round…read
Technical:
Microsoft and Nvidia team up to train one of the world’s largest language models (slightly older than this week but worth adding for the inaugural newsletter)…read
Apple AI Researchers Propose ‘Plan-then-Generate’ (PlanGen) Framework To Improve The Controllability Of Neural Data-To-Text Models…read
In the new paper Hierarchical Transformers Are More Efficient Language Models, a team from the University of Warsaw, OpenAI and Google Research proposes Hourglass…read
Baidu recently announced PLATO-XL, an AI model for dialog generation, which was trained on over a billion samples collected from social media conversations in both English and Chinese…read
Ethical & other:
We argue that the growing prevalence of statistical machine learning in everyday decision making – from creditworthiness to police force allocation – effectively replaces many of our humdrum practical judgments and that this will eventually undermine our capacity for making such judgments. read
Google AI has released GoEmotions. It is a human-annotated dataset of 58,000 Reddit comments extracted from popular English-language subreddits and labelled with 27 emotion categories; it includes 12 positive, 11 negative, and 4 ambiguous emotion categories and 1 “neutral” category. read
New research coming out of the Massachusetts Institute of Technology suggests that the underlying function of ‘next-word prediction’ computational models resembles the function of language-processing centers in the human brain…read
Looking forward, we’ll consider:
Week 2:
Usual sections +
What the distribution of these in-scope companies by age, how much raised and where they are in their maturity cycle
Week 3:
Usual sections +
Who are the major players at the moment i.e. raised >$100mm, and what are their capability distributions?
Week 4:
Usual sections +
Who are the middle rank players at the moment i.e. raised $10-$100mm, and what are their capability distributions?
Week 5:
Usual sections +
Who are the major public i.e. listed players at the moment and what are their capability distributions?
Week 6:
Usual sections +
Who do we follow and read for latest NLU technical insights?
Week 7:
Usual sections +
Who are the smaller players i.e. raised <$10mm, and what are their capability distributions?
Week 8:
Usual sections +
Ok….first company deep dive review
Thanks if you made it this far!!!