Artificial Intelligence — 2017 accomplishments
I spent time understanding that new phenomenon called “Artificial Intelligence”. Here are the key accomplishments that, I believe, reflect the increasing impact of AI across industries and businesses (transportation, healthcare but also agriculture among others). It is not specifically centered around scientific research but more about the increasing role of AI as it relates to industries. You will note the role of leading players like Microsoft, Apple, Google or others and their major announcements in 2017.
Please do not hesitate to comment or write to me to discuss further.
January — Microsoft doubles down on AI
Microsoft CEO, Satya Nadella, expressed his vision of the role of Artificial Intelligence at the DLD Digital conference in Germany: “The fundamental need of every person is to be able to use their time more effectively, not to say, ‘let us replace you.’ This year and the next will be the key to democratizing AI. The most exciting thing to me is not just our own promise of AI as exhibited by these products, but to take that capability and put it in the hands of every developer and every organization.”
Indeed, the year started with a multiplication of articles about AI leading to the end of the human race as we know it and worries that we will all be replaced or work for robots in the near future.
Democratizing AI by giving everyone the tools to work with and understand AI was a core breakthrough in 2017, initiated in January.
For more info — see article here
February — Ford gets all in with a $1b investment in self driving cars
As all automotive manufacturers, Ford is in the race to produce automated vehicles. Through a $1b investment in Argo AI, Ford promised to start production as soon as 2021.
Argo AI, founded by Google veterans late 2016, is based in Pittsburgh and is working to develop a virtual driver software platform. Argo is tasked with developing the entire “virtual driver system,” which means all of the sensors like cameras, radar, light detection, and ranging radar known as LIDAR, as well as the software and compute platform. The main challenge for autonomous vehicles is to identify elements of their environments (bikes, pedestrians etc.) and then adequately predict their behaviors. Training deep neural networks is one of the solution being implemented today.
At the time of the investment in February, everything had to be done. After a year, the company has 250 employee (see details here). In October 2017, Argo AI also bought a company that makes laser systems needed to operate cars without humans (LiDAR sensors), called Princeton Lightwave.
March 2017 — AI supporting cancer detection
On Friday March 3rd 2017, the Google research blog announced that it had achieved significant results in using artificial intelligence to identify breast cancer. Using deep learning to analyze thousands of slides of cancer cells, Google hopes its technology will, in the future, help pathologists better treat patients.
Objective was to implement an automated detection algorithm, that would help doctors go through huge amount of information and thereby reduce the diagnosis variability.
April 2017 — Machine Learning, a $20B market?
Here are the big numbers.
2017 was the year when AI moved from being a research project to a meannigful business opportunity. Recent estimates mention that the global machine learning as a service market is set to increase from $1B in 2016 to nearly $20B in 2025.
Main opportunities rely in professional services to help big corporations unlock value from legacy large datasets. Enterprise move to the cloud could also act as a catalyst to unlock value from Machine Learning as a Service.
In 2016, the MLaaS market is driven by Amazon, IBM, and Microsoft, with the three tech giants holding more than 73% of the overall market. However innovation in machine learning techniques are fierce and we expect new players to also take a share of the pie in the coming years.
May 2017 — Faster AI solutions for enterprises
In May 2017, H2O.ai announced a collaboration with Nvidia to offer its machine learning algorithms powered by GPU (graphics processing units) acceleration.
One quote that stands out from Adam Wenchel, VP of AI and Data Innovation at Capital One: “Future advancements in machine learning will unlock opportunities for us to create breakthrough consumer experiences in ways that we can’t even imagine today. As users of the H2O.ai and NVIDIA platforms, we see GPU acceleration of machine learning as a transformative development for the enterprise distributed machine learning community.”
H2O.ai is famous for being one of the first company to provide an open source platform to help enterprises derive value from data with machine learning and deep learning. Enterprises can use this new end-to-end solution to operate on large data sets, iterate faster, deploy quickly, and gain real-time insights. As of Q1 2017, about 9,000 enterprises and a third of Fortune 500 companies are using H2O.ai’s solutions. And we expect to see more …
June 2017 — When we lost control of chatbots…
Everyone probably remembers that story. While you were having breakfast and getting ready to go to work, researchers from the Facebook Artificial Intelligence Research lab (FAIR) made an unexpected discovery as they were trying to improve chatbots. Their AI bots accidentally invent a new LANGUAGE while training to negotiate with one another. Yes this happened … scary …
Instead of mimicking human speech, the bots developed their own machine language spontaneously.
Up until that time, chatbots were handling simple tasks, basic decisions. Attempt was made to have them dialogue and “to engage in start-to-finish negotiations with other bots or people while arriving at common decisions or outcomes”. Which they did … but leaving people out of it.
July 2017 — Apple is getting in the race
Finally! it was long due but in July 2017, Apple announced it launched a blog focused on machine learning research papers and sharing the company’s findings.
Historically the company has been more secretive than others on its machine learning research and previously did not allow its researchers to publish in academic journals.
After a few months, the site contains ~8 articles around face detection algorithms, training models to label synthetic images efficiently, challenges beyond adding new languages in Siri etc.
Apart from sharing technical breakthroughs, it is clear that Apple wants to leverage the platform to recruit more engineers (big link in the footer to screen job openings at the firm).
Here is the blog
August 2017 — AI is getting to gaming
“Google’s AI Declares Galactic War on StarCraft”
It all started with Lee. In May 2016, Lee Sedol, one of the best player of Go in history, lost to an artificial intelligence called AlphaGo, the machine built by researchers at Google’s DeepMind lab.
From there, DeepMind started to expand to other games, more complex and closer to real life situations. As such, researchers started feeding StarCraft data to their learning software. And in August 2017, they released a bot that can compete against humans in Starcraft. The beginning of an untapped market opportunity (the US gaming industry is estimated to be ~$20b in 2017, its second largest market behind China). And also the playground to test advanced technologies that can be used in real life in a few years.
September 2017 — Machine Learning machine technology for the agricultural industry
Increasing productivity is one of the key agricultural challenges of our century. Blue River technology, a start up based out of Sunnyvale, California, applies computer vision and artificial intelligence to observe, identify, and precisely spray specific plants. The target’s technology is intended to assist crop production by reducing the volume of chemicals by up to 90 percent, according to Blue River.
That company was acquired by John Deere,a wholesale distributor of agricultural machinery and other equipment, in September 2017, for a total of $305m, in an effort to build the next generation of smart agriculture equipment.
Press release here
October 2017 — AI getting easier with Gluon
The two giants, Microsoft and Amazon, unveil a partnership to open source a programing library called Gluon, in an effort to give developpers ready to use machine learning tools.
Applying deep learning algorithms requires expertise and time to optimize content. Gluon offers building blocks on which developpers can create their own applications.
Microsoft’s CEO had said it in January: the key is to democratize Artificial Intelligence. Such partnerships are the perfect example.
November 2017 — re:Invent
November has now become the AWS month. The most attented (30,000+ people) AWS conference took place end of November in Las Vegas.
AWS is an $18b business, growing at 40% year on year. Each year, the company releases its major announcements at its conference. Among the ones I noted related to Machine Learning:
- Amazon Translate — real time translation service
- Amazon Transcribe — smart speech to text service
- Alexa for Business — help office workers boost their productivity but also get the tools and resources they need to deploy Alexa within conference rooms etc. — for me, this is the most exciting announcement!
- Machine Learning Solutions Lab — pairing your team with AWS experts to help you design and optimize your machine learning algorithm.
Amazon is moving from selling building blocks to selling end to end solutions and this is very promising!
December 2017 — NIPS, the Neural Information Processing Systems conference
In the world of Artificial Intelligence and Machine Learning, NIPS is the conference that matters in December.
It was first organized in 1986 by The California Institute of Technology and Bell Laboratories, and originally designed for researchers to discuss artificial neural networks.
In 2017, in the wave of the Artificial Intelligence hype, it became the place to be seen and most importantly the place to hire talents. With more than 8,000 attendees, the conference is now the place for the best researchers to come present papers on machine learning, artificial intelligence and statistics.
Interesting summary by Bloomberg here