SemanticScholar

Soms komen trends bij elkaar.

SemanticScholar is hier een mooi voorbeeld van. Het combineert data mining, natural language processing en artificial intelligence om uit de almaar groeiende hoeveelheid publicaties datgene te vinden wat het meest relevant is.

SemanticScholar is gemaakt door AI2 en biedt een non-profit platform dat het mogelijk maakt om te zoeken naar academische publicaties. Het bevat op dit moment meer dan 40 miljoen publicaties op het gebied van computer science en neuroscience. Binnenkort komen daar de publicaties op het gebied van biomedical sciences bij. Bronnen die gebruikt worden zijn o.a. arXiv en PubMed.

De makers claimen dat hun product sneller en beter relevante publicaties vindt dan bijvoorbeeld met Google Scholar of PubMed mogelijk is. Wat betreft de content, dat door computers wordt opgehaald en verwerkt, zijn ze zich er gelukkig bewust van, dat dit niet waterdicht correct is. Daarom wordt de mogelijkheid geboden om correcties te melden via de FAQ.

Bij het zoeken naar een artikel biedt het programma mogelijkheden om via suggesties voor nieuwe termen de zoekopdracht te verfijnen. Ook kan er worden gefilterd op bijvoorbeeld jaar en documenttype. Vervolgens kunnen interessante artikelen worden bewaard in een leeslijst na inloggen met een bestaand Twitter-, Facebook- of Google account.

In een interne evaluatie door de TUDelft Library in 2015 was het toen beperkt aantal bronnen en wetenschappelijke werkvelden de reden om even af te wachten hoe dit platform zich zou ontwikkelen.
Intussen is er content beschikbaar vanuit meerdere bronnen en nu lijkt de tijd rijp om SemanticScholar opnieuw op waarde te schatten.

 

Smart Campus, voor het gemak

Verleden jaar werd de TUDelft Library vereerd met Deens bezoek, waarbij Lars Binau vertelde over de Smart Library als onderdeel van de Smart Campus van DTU (Technical University of Denmark) [1].

Onderdeel van dit project is het gebruikt van sensoren die in de lampen zijn ingebouwd, wat natuurlijk heel slim is. Lampen zijn doorgaans overal aanwezig waar mensen zijn en er is stroom.

Het prettige van sensoren is dat er zoveel mee gemeten kan worden. Er zijn klimaatsensoren voor het bepalen van bijvoorbeeld de temperatuur en luchtvochtigheid, maar ook sensoren die het geluidsniveau meten of de aanwezigheid van mensen.
Met de verzamelde gegevens is de logische volgende stap om iets met die gegevens te doen. Lampen en airco’s kunnen bijvoorbeeld worden aangestuurd, afhankelijk of er mensen zijn, waarbij ook de voorkeur van die groep mensen een factor kan zijn voor de werking.

Een lamp of airco zelf kan ook weer sensoren hebben om door te geven of het apparaat wel goed werkt.
Door gegevens van apparaten met elkaar te combineren, zoals van een airco met de aanwezigheid van mensen, kom je al snel uit op een samenspel van verbonden apparaten en dan komt de term IoT (Internet of Things)[2] om de hoek kijken.

Naast gemak voor mensen leveren al die sensoren veel gegevens op waar bedrijven graag op inspelen.
Een bedrijf [3] heeft plannen om gegevens vanuit een stofzuigerrobot, over de indeling van een huis, te verkopen. Een ander bedrijf [4] heeft een aardige verzameling artikelen geschreven over gebouwbeheer en AI (Artificial Intelligence)[5].

Diverse mensen zijn bezorgd over deze ontwikkelingen en Elon Musk, bekend van onder andere de Tesla, heeft dan ook een oproep gedaan aan de politiek om regelgeving rond AI te maken [6], voordat het te laat is.
Er zal sowieso steeds meer regelgeving komen vanuit de overheid; ook een Smart Campus zal een manier moeten vinden om optimaal gemak voor studenten en werknemers te bieden, zonder privacy aan te tasten.

[1] http://www.bibliotek.dtu.dk/english/servicemenu/visit/smart_library, 31-7-2017
[2] https://nl.wikipedia.org/wiki/Internet_der_dingen, 31-7-2017
[3] https://www.iculture.nl/nieuws/roomba-fabrikant-huisdata-apple/, 31-7-2017
[4] http://3bplus.nl/ai-en-sensoren-een-introductie/, 31-7-2017
[5] https://nl.wikipedia.org/wiki/Kunstmatige_intelligentie, 31-7-2017
[6] https://www.recode.net/2017/7/15/15976744/elon-musk-artificial-intelligence-regulations-ai, 31-7-2017

Google I/O 2016 The journey to AI-first

During the annual developer conference Google I/O, there were a lot of announcements made. Here is a little recap of the news till now on some renewed existing products and a few interesting new products.

Renewed existing products
Author Andrew Smith once said: “People fear what they don’t understand”. When Amazon launched Echo in 2014 the reactions were skeptical and people found the idea of a device that is always listing on command in your home, creepy [1]. Over time Echo became popular. Excitingly, one of the announcements in Google I/O 2016 was Google Home, which has a lot of similarities with Echo. It is a speaker you can instruct to listen and to control your house or, as many describe it, as ‘the always-present version of Siri’ or another virtual assistant software [1]. Google Home can be seen as a digital conversational partner with which you can interact in a hands-free way and one that can assist you with obtaining any intelligence you desire [1].

The virtual assistant software has also been adapted into Google’s newest messaging platform introduced as Allo, which is more than just another chat program. Allo makes it possible to send automatic replies or provide extra information when required by using a pattern of text and picture recognition in the context [1].

Google’s DUO is a video chat which goes a step further by providing an extra feature ‘Knock Knock’, which allows the recipient to see a preview of the face-to-face chat call before accepting the call [1].

New products
The days of using an app easier and quicker have arrived. Introducing: Instant Apps, which allows you to use the app (or a part of the app) instantly without any installation and without worrying about enough storage on your phone [1].

And then there is the innovation presented as DayDream, that takes you closer to all your VR (virtual reality) dreams on a mobile device [1]. DayDream is a development platform for mobile VR in order to create VR experiences (YouTube – Google I/O Highlights).



Google I/O Highlight (2 minutes)

Interestingly, most of the features emphasize the essence of AI (Artificial Intelligence). The importance of AI increases as it provides features to understand the context in order to assist the user more effectively. This gives the virtual assistant software a whole new meaning. Virtual assistant is not merely considered as an application or a product, it has become a kernel of every application and product. With the growth of AI inside of virtual assistant the impossibilities are becoming more and more possible. According to Google’s CEO Sundar Pichai “20 percent of queries on its mobile app and on Android devices are voice searches” [2]. Presently, by using Google products, we, as the users, are feeding AI with data, which allows AI to expand its potential beyond words. As Pichai points out, he is on a journey from “mobile-first to AI-first”, but meanwhile, “there are areas where we will be ahead, and there will be areas where someone points a way and we do it” [3].




Google I/O Keynote (almost 2 hours)


Sources:
[1] google io 2016 what you need to know from allo to daydream
[2] google reveals 20 percent queries voice queries
[3] googles ceo sums up his ai vision

TNW Europe (part 2), Google’s secret sauce

TNWConf2016
The 11th annual TNW Conference Europe took place on May 26 & 27 in Amsterdam. Described, as “The most intimate technology festival on the planet” by CNBC, over the years TNW Conference has become one of the leading technology events in the world. It’s organized by Dutch based online media company The Next Web and brings promising startups, investors, technology gurus, innovators and entrepreneurs together. It’s a great place to share their thoughts about the future of technology, marketing, talk about design or for networking, to get inspired or to boost your imagination.


google-new-logoGoogle’s director of product Aparna Chennapragada talked about how to build the mobile products of the future. Is there a recipe or special formula to do that? Google have a sauce, a secret sauce and it looks like this:

Secret sauce = AI+UI+I

Daily we use a lot of different products like Google Search, Google Translate, Goggle Now, YouTube (…. and many more). All of them have something in common. They are all powered by information (in many cases AI and machine learning in various forms). But why is this fact more relevant now than ever?
The answer is simple ”one word …Mobile” said Aparna.

27519172705_f575126d5c_h

Mobile is the game changer. With 3 billion plus phones we produce daily a massive input of useful data that can help improve these products. Mobile changes the game on the output side as well. Nowadays we always carry our phones and that gives us access to products and services in situations and context that we never had before (like in the car for example). If you combine these two things together almost every other real-world problem has a chance to turn into software or more importantly AI problem.

Google’s formula solves this problem.

AI (Artificial Intelligence)+UI (User Interface)+I (Personalization)

The formula in depth:

 AI (Artificial intelligence)

There are 3 observations about AI

  1. Use AI for tasks that are easy for the machines but difficult for people.
    Google Translate is a good example for that
  2. Wow vs. WTH (what the hell!!!) ratio – If AI sends you to the wrong airport gate and you mist your flight (WTH) the product will need a lot Wow’s to make up for that WTH moment.
  3. Training shapes the learning – AI is as good as the training data. When overtime the data improves, the AI improves too.

UI (Interface)

  1. UI needs to be proportionate to the confidence in AI – Strong AI needs less UI and vice-versa
  2. Magic vs. Prediction trade-off– people tend to choose for predictable but slower approach instead of faster “magical” solution.
  3. User feedback is very important for improvements of the system but hard to get nowadays. For example: Google Now question “Will it rain this weekend?” gives no feedback to the system.

I (Personalisation) 

  1. Make the benefits clear and immediate- people will not use products that only promise benefits in a long run
  2. Allow users to teach – ask you users to help (when possible!)
  3. Who are your users – it is important to know who your users are. What will work for specific users in US will not automatically work in India.

The whole presentation is now also available on YouTube
Photo credit:TNW

Trends 2016: Virtual Reality (VR), Augmented Reality (AR) and Artificial Intelligence (AI)

With all the technological developments from the last year IamWIRE says that if 2015 was the year of the sensors (products), 2016 will be the year of user experiences due to the release of high tech devices in the consumer market. They are referring to three of the most important technological developments that will emerge as trends for this year: Virtual Reality (VR), Augmented Reality (AR) and Artificial Intelligence (AI).

Trends:
  • Looking back at the last year’s developments in Virtual Reality, we saw in 2015 the headsets Gear VR, Oculus VR and Google Jump platform appear, the last one with the 16 GoPro camera’s that can create 360 degrees videos, a new type of content, now already popular on YouTube.

    All of them open new possibilities for Virtual Reality new experiences in gaming, training, and simulations. In the beginning of this year is scheduled the launch of Oculus Rift (already here for a price of $599), Sony’s PlayStation VR and HTC – Vive headset with VR technology. According to TrendForce, the expectations are as far as 43mil. users this year with 14 mil. VR devices to be sold, most of them to be used in gaming. Quirk’s is talking about Virtual Reality as of a democratising technology, but warns that its adoption by the industries and consumers will go far beyond 2016. It predicts that the economical sectors that are mostly based on experience – like tourism and health care – will adopt this technology faster than others.

  • In 2015 the investments in Augmented Reality went up to $700mil. (1). Its timeline didn’t begin with Google Glass (launched in 2013 and stopped in 2015), but it was surely considered well ahead of its time. The next generation, Google Glass 2 is expected in May 2016 (2). What else did Google do in 2015? They backed-up a startup company, Magic Leap, that raised a huge amount of investments for its 3D mapping system integrating real objects with interactive graphical “objects” (3). Let the race begin!

    In the spring of 2016 are announced as well the Microsoft’s HoloLens and ASUS’s AR headset.
    There is no promise yet for the consumer market, but everyone expects a serious name taking a stand in AR: Apple. Its interest in AR became visible through its moves (in acquiring in 2015 three AR companies) that suggest a long term goal. Latest news: Apple just hired beginning 2016 a top tech specialist in AR/VR (4).

Continue reading

Artificial Intelligence at work for website building

Would you like to design a website whiteout having to think to much? The people behind The Grid think that building a website should be as easy as feeding the website with text or photo’s.
The Grid is a website creation and hosting platform that uses AI to learn from you, the website creator. It learns about you, your habits and preferences, it adapts to you and evolves with you. How? Templates are replaced with layout filters, the algorithms pick up your color preferences from your online posts, and crops main figures from your photos for easy design. Just try it!