Where does your enterprise stand on the AI adoption curve? Take our AI survey to find out.
The AI talent grab is real. This year alone, Pinterest CTO Vanja Josifovski jumped ship to Airbnb, while Pinterest hired Walmart CTO Jeremy King to head up its engineering team. Moreover, all the big tech companies, including Google and Apple, have for some time been vacuuming up AI talent through acquisitions — a recent CB Insights report noted 635 AI acquisitions since 2010, topped by Apple with 20 acquisitions. Elsewhere, Microsoft turned to online education platforms to help train a new generation of AI students.
But while the AI talent pool may be growing, a significant shortage remains. Those with the necessary skills — ranging from robotics to computer vision and natural language processing — are in high demand, and they face a clear choice: Take a paycheck at one of the big tech companies, chase a longer-term moonshot at an emerging startup, or launch their own business from scratch.
Throughout 2019, tech companies have ramped up their efforts to secure the best AI talent and technology. Here, we take a look back at some of this activity, with a focus on the “big 5”: Facebook, Amazon, Apple, Microsoft, and (Alphabet’s) Google (FAAMG).
Back in February, news emerged that Facebook had snapped up GrokStyle, a San Francisco-based visual search startup founded in 2016. GrokStyle had developed an app that can automatically detect decor and home furniture from a photo, and it partnered directly with retailers to help users find items to buy.
Facebook swiftly shuttered GrokStyle, a clear indicator that this was more of an acqui-hire than anything else. As to how its technology or talent will be incorporated, Facebook simply said ” … their team and technology will contribute to our AI capabilities.” A separate GrokStyle statement at the time said it would continue to use its AI to build “great visual search experiences for retail.”
GrokStyle’s smarts could resurface inside Facebook Marketplace, the platform’s peer-to-peer platform for buyers and sellers. Or it could be applied to an as-yet undisclosed service that allows Facebook users to find brands through visual search.
In September, Facebook announced that it had bought New York-based Ctrl-labs, a startup founded in 2015 that developed a wristband to translate neuromuscular signals into machine-interpretable commands. It’s basically a brain-machine interface, and its prototype Ctrl-kit uses EMG to translate mental activity into action. Some 16 electrodes monitor the signals amplified by the muscle fibers of motor units, from which they measure signals. With the help of AI algorithms trained using Google’s TensorFlow, it can distinguish between the individual pulses of each nerve.
Facebook said it planned to fold Ctrl-labs into its Reality Labs division, a unit that is chiefly concerned with VR and AR. This deal was perhaps more about Facebook finding new ways to interact with devices and technology in general than a strict AI acquisition.
Also in September, news leaked that Facebook had bought Servicefriend, a four-year-old Israeli startup that creates messaging bots for customer service teams. Facebook hasn’t said much about the acquisition, electing to release only a vague confirmation that the deal took place, but at least one of Servicefriend’s cofounders now works for Facebook’s Calibra digital wallet team.
Back in July, Facebook announced controversial plans to spearhead a new digital currency called Libra, an alternative blockchain-based financial system that bypasses traditional banks to lower the costs of transferring money online and “lower the barrier [to entry]” for businesses. In tandem, Facebook is also developing the Calibra digital wallet built on Libra’s infrastructure, which will likely require round-the-clock customer support. So it seems likely that at least some members of the Servicefriend team are working on AI-enabled bots for Calibra.
TSO Logic (Canada)
Way back in January, Amazon’s cloud computing offshoot, Amazon Web Services (AWS), bought TSO Logic, a Vancouver-based startup that builds cloud spending analysis tools.
Founded in 2013, TSO Logic is all about helping companies work out the cost of running their current workloads in the cloud — clearly a useful tool in AWS’ bid to win new customers.
From an AI perspective, TSO Logic ingests “millions of data points,” such as the age, generation, and configuration of a company’s hardware and software, and then creates a granular statistical model to demonstrate how much the company is spending and where it could cut costs by transitioning to the cloud. The TSO Logic platform uses machine learning algorithms and pattern matching to determine the best fit for each workload out of the myriad options available across the public and private cloud.
Amazon acquired mesh network startup Eero in February for a rock-bottom price of $97 million — not much more than the $90 million it had raised from outside investors, which suggests the company was not in great shape.
Founded in 2015, San Francisco-based Eero has offered a home Wi-Fi system that uses multiple access points to extend connectivity to every corner of a building. A companion mobile app allows users to share their network, set up parental controls, run speed tests, and more. Although Eero is on the surface a hardware company, software plays a big role — its TrueMesh software was developed by applying machine learning to data collected from thousands of homes, allowing it to optimize its routing algorithms to ensure maximum network coverage.
For Amazon, acquiring Eero makes a lot of sense, given that the startup already offered a bunch of connected home devices. In September, Amazon launched the first Eero product since the acquisition, one that enables more granular Alexa-powered voice controls — including the ability to deactivate Wi-Fi for a specific device or turn on a guest network.
This was probably more of a hardware and IoT play than an AI move, but the acquisition was still notable in terms of supporting Amazon’s grand plans for leveraging AI in the connected home.
Canvas Technology (U.S.)
In April, Amazon bought warehouse robotics startup Canvas Technology for an undisclosed sum. Founded in 2015, Boulder, Colorado-based Canvas was an obvious candidate for Amazon, given the ecommerce giant’s growing move toward automation in its fulfillment centers. Subsidiary Amazon Robotics, which manufactures robotic warehouse technologies, was the result of Amazon’s $775 million acquisition of Kiva Systems back in 2012.
Canvas, which uses AI and computer vision to enable its robotic vehicles to navigate warehouses autonomously, has now effectively been swallowed up by Amazon’s robotics unit.
Back in May, Amazon made a small but notable move to buy some of the technology belonging to a 20-year-old New York-based advertising technology company called Sizmek. While the deal may have been more about snapping up engineering talent from a struggling, recently bankrupt ad tech company, it also signaled Amazon’s intentions in an online advertising space dominated by the likes of Facebook and Google.
Sizmek’s Ad Server is a direct competitor of Google Marketing Platform, once known as DoubleClick, and it helps marketers optimize and measure their online ad placements. Amazon revealed that it planned to continue serving Sizmek’s customers once the deal was concluded, and the Sizmek platform is now being sold as “Sizmek by Amazon.” Amazon also bought Sizmek’s Dynamic Creative Optimization (DCO) platform that allows brands to create hundreds of individual ads tailored for specific audiences — Sizmek leverages AI to ensure the right ad is delivered to a target market.
Amazon’s longer-term plans for Sizmek aren’t clear, but the ecommerce giant has been investing more in its advertising technology, and it certainly has the reach and muscle to take on the two major incumbents in online advertising. AI is expected to power around 75% of of ad impressions by 2023, according to a recent report by Juniper Research, with Amazon taking 8% of that spend.
News emerged back in March that Apple had bought a Silicon Valley machine learning startup called Laserlike, though the acquisition likely closed the previous year, judging by the founders’ LinkedIn profiles.
Laserlike used machine learning to scan information from across the web and deliver personalized results based on a user’s natural language searches. The results were displayed so that the user could browse them like news.
The technology could be used to help Siri deliver more personalized results, or even to surface tailored results through the new Apple News+ subscription service.
Laserlike was founded in 2015 by former Googlers who may have been tempted to jump ship, given that Apple had hired former Google AI chief John Giannandrea the previous year. Whatever Apple has in store for the Laserlike tech and team, it has managed to snag some quality AI talent in the process.
We’ve known Apple has been developing some form of vehicle for a while, and the company has reportedly been doubling down on its autonomous car efforts. Back in June, news emerged that Apple had snapped up the assets and some employees from struggling self-driving car startup Drive.ai, which was apparently in the process of shutting down four years after launch.
This doesn’t necessarily mean that Apple is any closer to building a self-driving car, but the company does now have more engineering and product talent with expertise in that field. Make of that what you will.
A few months back, reports emerged that Apple had acquired a Swiss computer vision startup called Fashwell. Apple has not confirmed the acquisition, but Fashwell’s founder and several senior executives joined Apple in January of this year, with several others joining later.
Founded in 2014, Fashwell worked with companies to integrate its visual search technology into their apps and websites, allowing users to search using photos rather than keywords. It’s all about making images “shoppable.”
It’s not clear what products Fashwell’s team is working on at Apple, but this felt like more of a talent grab than anything else.
Microsoft has made plenty of AI-focused acquisitions in the past, but the Seattle tech titan was fairly quiet on this front for 2019. A couple of its acquisitions did lean on automation and machine learning-based technologies, though.
In August, Microsoft announced that it had bought Java performance-tuning tool Jclarity. Founded in 2012, Jclarity uses machine learning techniques in its Illuminate offering to automatically find and fix performance issues in apps that rely on Java.
Also in August, Microsoft acquired PromoteIQ, an automated product marketing platform that lets brands run sponsored ads on retailers’ websites. PromoteIQ, which was founded in 2012, will continue to operate as a division of Microsoft’s advertising unit, and an announcement seemed to confirm that PromoteIQ and Microsoft plan to pool their respective AI and targeted ad placement smarts.
“Microsoft brings industry leading AI and machine-learning capabilities, as well as a strong global retail market footprint,” PromoteIQ said at the time.
A few days into the new year, news emerged that Google had acquired the founding team behind Superpod in what was largely seen as an acqui-hire of the founders, as well as a move to annex some of the startup’s assets. The acquisition actually closed in late 2018, according to Superpod cofounder and CTO William Li’s LinkedIn profile.
Founded in 2016, Superpod was essentially a crowdsourced question-and-answer app, similar to Quora. An obvious place for Superpod’s technology would be in Google Assistant. Amazon launched a crowdsourced Q&A service to tackle questions Alexa is unable to address, so it’s certainly feasible that Google could be planning something similar for its virtual assistant.
A few months back, New York-based education technology platform Socratic announced that it had been acquired by Google, though the deal actually closed way back in March, 2018, according to Socratic CEO Christopher Pedregal’s LinkedIn profile.
Founded in 2013, Socratic made a Q&A app that uses text and speech recognition to surface the most relevant learning resources based on a user’s question, with a focus on school subjects, such as science, math, literature, and social studies. While Google itself can obviously be used to research most school subjects, Socratic is guided by teachers and focused specifically on the school curriculum, removing much of the “noise” that comes with a standard search engine.
Socratic is very much being kept alive as a standalone product, and it is now branded as “Socratic by Google,” making this much more than a simple acqui-hire. However, the technology could easily be reappropriated by Google, and perhaps used to help train Google Assistant.
Looking at some of the AI-focused acquisitions of the past year, it’s clear that most of the big tech companies have sought to bolster their in-house teams through acqui-hires. This was pretty much true of each of Facebook’s and Apple’s AI acquisitions this year.
In some cases, it wasn’t always obvious whether the AI element of a product was what the acquirers were looking for — in other words, AI may only have been tangentially related to the acquisition. Facebook’s purchase of Ctrl-labs was perhaps more about finding new ways to interface with machines, for example, while Amazon’s Sizmek acquisition may have been chiefly about acquiring an ad server. But machine learning and automation were still a core part of those acquisitions, and certainly contributed to the respective companies’ appeal.
A broad look at the various AI acquisitions throughout 2019 also highlights the average age of the startups. Most were founded between 2013 and 2016, and of the 14 acquired companies we looked at, six were founded in 2015. This suggests that four years may be the sweet spot for scouting AI talent or general engineering talent. It’s just enough time for a startup to get a viable product into the market without becoming so big that it commands an astronomical buying price.
A quick peek across the industry reveals notable AI acquisitions beyond those of the “big 5,” and their founding dates tend to reaffirm this trend. Twitter acquired Fabula AI (founded in 2018), a machine learning startup that helps spot fake news; audio giant Sonos snapped up privacy-focused conversational AI company Snips (2013) for $37.5 million; Tesla nabbed computer vision startup DeepScale (2015) to develop driverless vehicles; VMware bought machine learning acceleration startup Bitfusion (2015); Matterport acquired Arraiy (2016), an AI startup that automates special effects processing in movies; and Intel shelled out $2 billion for AI chip startup Habana Labs (2016).
Even “non-tech” companies got in on the AI acquisition action. Nike, for instance, bought out Celect (2013), a Boston-based predictive analytics platform that helps establish consumer demand. And McDonald’s acquired two AI-focused companies, Dynamic Yield (2011) and Apprente (2017), to bring voice-enabled digital agents to drive-thrus alongside dynamic menus that change based on environmental factors. For several years already, people have proclaimed that every company is now a technology company, and this is certainly evidenced here. However, these acquisitions also lend credence to the notion that all companies will eventually be AI companies.
All of this activity casts some light on AI’s potential reach across sectors — from ecommerce and driverless cars to education and customer service. And acquisitions will continue to play a pivotal role in big tech’s AI strategy — after all, nabbing the brightest brains may be the best way to stay ahead of the curve.
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
- networking features, and more