Hive – AI for Enterprises Meets Human-in-the-Loop
We’re not going to lie to you. Sitting around all day researching the crap out of various technology topics, and then giving said research away for free, is a bit disheartening at times. No, it’s not the fact that we have to do the research. That’s the best part. We come across incredibly interesting facts, talk to some of the most fascinating people on the planet, and get consistently blown away by just how promising some technologies actually are. The worst part is that we can’t afford to pay for some decent tools that could make our research better. Take CrunchBase for example. They’re a tool we use to look up basic information about startups like funding taken in so far, size of last funding round, year founded, and the like.
We use CrunchBase to produce many of our “top 10 lists” in which we usually provide some sort of ranking of companies based on funding taken in so far. We usually use CrunchBase to produce these lists because the price is right (60 bones a month or so), but what we’re finding is that not everyone keeps CrunchBase updated. Take, for example, the below feedback we received from a reader on our “Top-10 AI Startups in Australia and New Zealand” article:
Looks like they don’t use CrunchBase in the Land Down Under. In a previous article, we talked about how the best startup data source to use is CB Insights. Unfortunately, we don’t have thousands of dollars laying around to pony up for the gold standard of startup data, so until Anand decides to show us how much he loves us with a subsidized CB Insights subscription, we’re left with having to rely on you, our readers, to point out areas where we’re lacking when it comes to basic startup data. Since the most promising startups out there aren’t even updating CrunchBase because they are far too busy kicking ass and taking names, it means that sometimes we come across interesting companies that we should have written about a long time ago. Take a startup called Hive.ai for example. A quick look at CrunchBase tells us absolutely nothing about Hive except that they were founded in 2017 and that “Hive Predict provides AI capabilities to enterprises.” Let’s talk about that.
AI for Enterprises or Enterprise AI?
We’ve touched on enterprise AI before, a topic that largely involved looking at companies like Noodle that hope to involve AI in corporate decision making. That certainly could be considered “Enterprise AI”, or solving a small number of large problems using AI. Then there’s this notion of “AI for enterprises” which is more about providing a set of tools to solve smaller problems in large numbers. In order to sell either of these solutions, you’ll need:
- Something that’s easy to use and doesn’t require dipping into the travel budget or boring your employees to tears when they’re forced to attend some brown-bag training session
- Something that’s already being used by big corporations and therefore has been “blessed” by someone else. (Yes, we see the catch-22 here, and it’s something that everyone deals with when trying to sell enterprise software solutions.)
- Something with an easy-to-understand demo that can immediately wow the dinosaur in senior management who always objects to implementing new and exciting technologies in favor of the “this is how it’s always been done” option.
Hive does all of this right, and they immediately started wowing us with some slick demos in their case study section. The first one shows a video playing along with analysis of the exposure time that various brands receive during the news clip. Forget about the tech, we were just blown away by how much branding is being shoved in people’s faces while they try to watch the news:
Watch the sporting example next and you’ll see just how much advertising happens during sports, even ones as uninspiring as hockey. After that, the use cases start getting a bit more interesting. There’s a “not safe for work” algorithm that can detect how offensive a video or photo might be for those people in the office who don’t know how to have fun. There are algorithms that can detect gunshots, explosions, or fights, and then detect commercials when they come on just based on the content that’s being displayed. Real-time object detection can identify things in videos and photos, and there’s even functionality we’ve talked about before called “planogram” which can be used for merchandising use cases. It’s all there for you to take a look at, and you can upload your own media to test the various algorithms out with an easy-to-use interface. Should you want to train your own algorithms, they’ll help you do that as well.
Hive Data – Data Labeling as a Service
So how did we first come across Hive? Turns out they’re doing some massive human-in-the-loop projects that are on par with the likes of Mechanical Turk. It’s called Hive Data and it’s all about paying thousands of people pennies to do extremely mundane data labeling tasks. The company’s 200,000+ workers can generate millions of labels within a few hours, and the number of workers can be increased by utilizing “surge pricing” in order to ensure faster turnaround times. It’s data labeling as a service, and allows you to label large quantities of data at dirt cheap prices. When someone in academia says your service is cheap, that’s when you know you’ve arrived at rock bottom pricing:
Hive has a past that involves academia, as the startup was founded by two PhD students at Stanford who were frustrated with the difficulty of generating quality datasets for machine learning research. Now, Hive is working with some of the world’s best universities that are studying machine learning in a quest to produce 1,000 academic papers on machine learning that all use data that’s been labeled by Hive. If you have data that needs labeling, they’ll sort you out. They’ll also help you develop custom machine learning models. Here are a few examples:
In addition to custom machine learning models, there’s also Hive Enterprise.
Hive Enterprise – AI for Enterprises
Hive Enterprise is currently composed of four industry verticals as follows:
- Hive Media – Currently, Hive Media ingests over 200 channels of broadcasted TV across the US and Europe, and our deep learning models identify and tag the footage with brands, celebrities, and objects in real-time. They have the world’s largest archive of TV broadcast content – ingesting 5,000+ hours per day – and they’re adding over 500 million data labels per day.
- Hive Security – A fully self-contained visual intelligence Software Development Kit (SDK) that is incorporated into physical server boxes and optimized for security use cases. It’s been trained on 20 million faces and can also spot objects.
- Hive Auto – Building the definitive visual perception layer for autonomous vehicles and operating the largest data labeling platform for the self-driving car industry. Nearly 100% accurate and 50x cheaper than competitive offerings.
- Hive Retail – Planogram model helps to determine: brand of a beverage, can or bottle, and front or back facing
Expect that list to grow. It’s also easier for Hive to make some select acquisitions and integrate them into each vertical without disturbing the rest – or even create new verticals via acquisition.
Hive has plans to grow their human-in-the-loop model to be a whole lot bigger. According to an article published yesterday by Business Insider:
Silicon Valley-based AI startup Hive has announced its plans to pay more than 600,000 people to help it label photos for its image recognition AI system. The workers, who are being paid pennies per image, will help the company create an AI the firm’s CEO Kevin Guo claimed would be “whole orders of magnitude better than Google.” “What’s the alternative? Playing Candy Crush Saga and losing money,” Guo asked.
The man has a point. The article goes on to talk about how Hive has taken in around $30 million in funding so far from investors that include Peter Thiel.
There are some pretty cool companies out there like Hive.ai that are flying under the radar and building some really cool AI-powered tools that today’s enterprises can start to implement quite easily. We’ve looked at quite a few use cases for Hive’s technology, and it just goes to show that you don’t need to be raising $100 million rounds every other day to have a viable business. Nor do you need to be keeping CrunchBase updated apparently.
Despite what the pundits say, FAANG stocks (Facebook, Apple, Amazon, Netflix, Google) don't give you real exposure to AI. Read about 5 stocks that give you true pure-play exposure to AI in our guide to investing in Computer Vision companies, freely available to Nanalyze subscribers.