AI, ethics and what the cows have to do with it
On day 3 of my sabbatical I was allowed to speak for the first time on the Ignite Talks of 12min.me in Hamburg. Oliver Rößling had asked me if I would even like to speak. Very cool format. 12min pitch, then discussion.
It felt a bit strange that the first time as a non-“official” Microsoftie after over 5 years to speak in public. But I can – I think – get used to it. The Slidedeck you can download at the end of the blog post. And yes there is a lot of content from MSFT in there. Not because I did not find anything else, but because he’s just good ?
AI (= AI = Artificial Intelligence), ethics and the impact on society have been my personal hobby for some time. I believe that we are currently in a world that is changing at a rapid rate. I myself was allowed to experience the whole start time of the internet live and in color and very directly. Whether as a student or entrepreneur. I was there from the beginning. And it was already very fast then how the business models have changed or even the topics. I think that many of you can still remember Boris Becker’s advertisement for AOL “Am I already in it?!?”. Yes, that was not really self-evident at the time. But no matter. We now want to talk about AI, ethics and cows.
Today, every company should be a data driven – data driven company. It does not matter if it produces cars, makes newspapers, sells food or any services.
Because data is the new gold, oil and natural gas!
Important here is the difference between Big Data and Smart Data. Big Data usually describes a collection of many structured and unstructured data that is collected and collected without the company knowing what to do with it. “You could use the data later”. Would you handle your gold that way? And is it ethically justifiable to collect all data (and evaluate it sometime)? Hmmmmm. I think that could lead to some discussions.
If we take a look at the whole Big / Smart Data, AI, Infrastructure Cloud Map, we can see that the topic is now quite extensive:
How can we become the whole Lord?
Well then there are AI and ML and the “new gods in white” the Data Scientists. These are exactly the colleagues who can tell us exactly what we always wanted to know but could never ask because we did not know what we wanted to know. Say the data scientists are able to master the big data of this world and use machine learning (ML) and artificial intelligence. But where does all this data come from?
In 2014, Cisco commissioned a study at IDC to analyze how much data collectors will be available in the near future (until 2018) to build their infrastructure-based business model. IDC has predicted that by 2018 – in a few months – approximately 20MW of data collectors will be installed worldwide (and then connected to Cisco infrastructure). But that’s not all. Any movements on the Internet are now logged. The Amazon, Zalandos, Netflixe and Google’s of this world now know better what you want to buy and buy in the future than you.
Unfortunately, the advertising on the Internet – Ad rotation etc – is not really geared to it. Unfortunately, you still get the articles that you have just ordered or that you certainly do not want. Unfortunately, we are not that well yet with Ingenious Technologies (I’m in AR). We can even track you via devices and you do not even have to be logged in all the time. Scary? Yes. Ethics? Hmmm…
But let’s see what machine learning (ML) is about:
“The goal of machine learning is
to program computers
to use example data or past experience
to solve a given problem. “
(Introduction to Machine Learning, Second Edition, MIT Press)
So the goal is to program computers so that they should solve a current problem with example data or on the basis of historical data (= experience).
ML is the (data) basis for AI. What does that mean when translated?
The artificial intelligence is only as smart as the database with which you fed them.
Why this can lead you to see Tay. This is the chatbot that Microsoft unleashed on March 23, 2016 on the European and American Twitter users – after a very successful test in Asia – and had to be shut down after 16h for releasing sexist and racist tweets. Tay has become so “famous” that he even has his own Wikipedia page (https://en.wikipedia.org/wiki/Tay_(bot)) and its own Hashtag #FreeTay has gotten.Ethics? None. Incidentally, IBMs Watson had a similar problem after “reading” the “Urban Directory”.
Bullshit in – bullshit out.
Old Computer Scientist Wisdom.
If we look at it all a bit more pragmatically, it looks like this:
A good and illustrative example of this is the recommender engine from Netflix. Surely you’ve wondered why Netflix’s suggestions are actually really good, or at first the recommender tries to understand what happened. So which films and series I have looked at, which I have canceled and which possibly even geliked. Next, the Recommender tries to find out why you did these actions in the movies and series. So why did you quit “Orange is the new Black” and why did you watch “Designated Survivor” and maybe even more than once? Next, the recommender tries to anticipate (say) what you’re going to do next. Are you actually looking at “Designatd Survivor” as well as “House of Cards” – which is pretty close. In this phase the Recommender does not show its suspicions to the users yet. This remains completely hidden in the system and Netflix tries to learn. The last step – and all steps are based on probabilities – is that the recommender – if he has collected enough data, so the probability has exceeded a certain threshold – suggestions to you to make the Netflix experience as good as possible for you. And you’ll find that the suggestions from the Recommender – or Recommendation Engine – are so good that, as a rule, they actually meet your taste. He has me, for example After “House of Cards” and “Designated Survivor” offered “White Collar” – and what can I say – I’m just at the end of the second season ?
How does the whole technology behind AI, Smart Data and ML look like?
it should be different – nothing works without the cloud. For the amounts of data that are collected and evaluated, it simply takes huge capacities that are meaningfully no longer kept in your own data center – even if this is critical or personal data – but centrally from one of the 3 relevant providers – Google, Amazon or Microsoft – or its partners related. Why even with personal data? The data centers of the 3 providers are partly already in Germany or at least in the EU and thus the offerers are subject to the GDPDR or in German the “basic data protection regulation” (DSGVO). Thus, at least once no Schindluder can be operated. Whether you still need the upgrade with the national cloud as offered by Microsoft, the user must decide for himself. For almost all application scenarios, the “European” cloud should be more than sufficient.
There is now a multitude of different services available on the individual cloud systems. Many are similar, none are alike. Even basic things like Docker or Kubernetes or a mySQL service on Google are different to AWS like Azure. Even if you rely on an open source “standard” you can not easily migrate your cloud application systems from GCP to Azure to AWS. This makes the selection in the beginning really difficult. My recommendation would be to use a Multicloud Service Partner (or Managed Service Partner – MSP) as an intermediary between the systems.
The basic services for Smart Data and ML are Spark, Hadoop and “R” as the query language. In theory, on all three clouds, I personally think that Microsoft is slightly ahead of Azure thanks to the acquisition of “Revolution Analytics” – the professional manufacturer of “R” in 2015. The Azure ML Studio is looking for the same on the other two platforms. That would be us then already in the application layer. On the one hand, the apps should make the use of the infrastructure and services as easy as possible, and on the other hand, the ML and Smart Data functions should be included in as many end users as possible – exactly what you have recognized – as much data as possible collected while using the apps. Again, Microsoft uses its strong position in the business customer environment. The AI and ML platform integrates with almost all business apps – be it Dynamics 365, CRM Online or just Microsoft Office – making apps smarter.
Unfortunately – and with that we come to the intelligent assistants – Microsoft overslept the consumer market and perhaps even gave up. Bing has no relevance in Europe – despite big deals with Apple and others – a smart speaker is looking in vain – if and when the harmanKardon “Invoke” comes to Europe or Germany is in the Stars and a mobile platform is not available anymore. Google with its home system has come in my view, much too late on the market and shines – as always – with a data greed of his peers: Google Home mini simply sends permanent information into Google HQ. Of course, you apologize for the hardware defect that has been fixed so immediately with a software patch (Diesel cheating ….). If you believe it will be blessed (Source: http://winfuture.de/news,100042.html) .
What’s going on the speaker: Exactly voice-controlled bots. So apps. In 2016 I got quite a stir with a talk during the “Explained Conference” where I said “Bots are the new apps”. This was first taken up by the dpa and via heise online and Focus is the saying then landed in the “power rotation” for a few days at the departure gates at the airports (https://www.heise.de/newsticker/meldung/Microsoft technicians bots are-the-new-Apps 3334108.html). So are bots really the new apps? I do not think so and I have to correct myself a little bit. Currently it is unfortunately so that the different Canvases – so the messenger / websites etc in which the bots are running – all have different functions and presentation methods. I feel strongly reminiscent of the “Browser Wars” in the late 90s where you had to program endless browser switches to see which features are supported and which are not. In addition, the development of bots in recent months has not really made progress. Intelligent is different. Currently, most bots are designed to follow a tree structure or, for computer scientists, very long If … then … else loops, and if you want to correct something in the 6th turn, you start all over again. Really, they do not do anything the bots and intelligent is different. But many call center functions can also be automated with “If … then … else”.
Therefore, I believe that the bots also need a major overhaul – similar to the web in 1998. My bot vNext looks like that bots bundled with voice user interfaces, ML and smart data will be the measure of things. And I think that here very fast a 2-3 fight of the system between Facebook, Google and Microsoft will start. Although AWS / Amazon makes a top VUI, in my view the real visionaries lack ideas and starting points on how to actually use vNext bots outside of eCommerce. In the Wired Backchannel, a top AI article was released in June that sheds some light on what Microsoft has in mind (https://www.wired.com/story/inside-microsofts-ai-comeback). The article is very worth reading.
All in all we live in an exciting time: Almost every aspect of our lives is being transformed by technology – and at a breathtaking pace.
I believe that today we are only at the beginning of a new technological revolution. It promises us a change in the way we live, work, communicate and learn. And at a pace and extent that has never experienced humanity before. A new generation of technological innovation gives us opportunities that promise new ways of unlocking economic opportunities and solving some of humanity’s most burning issues.
According to a recent World Economic Forum study, more than 75 percent of IT and communications executives believe that Within 10 years, we will see robot pharmacies, 3D printed cars, and 3D printed liver transplants that 10 percent of all cars will be on the road without drivers, and 10 percent of people will be wearing networked clothes via the Internet Managers in the IT and communications sector deceive. Serious.
I do not believe that this development takes another 10 years and then only for example. 10% of all cars drive autonomously. I believe the newly elected federal government must deal in the upcoming legislative period with how it actively uses the impact on the economy, the world of work and life to make Germany even stronger. In particular, the world of work is facing significant changes. Many of today’s standard jobs will not be in the same scale as today in 3-5 years and will be gone in 10 years. The fourth industrial revolution has now also affected occupational groups in which so far many employees work as knowledge workers on the » certain side.
Exactly these professional groups are in the focus: from the clerk in the insurance up to the auditor. The phases of refusal of reality , resistance  described in the change process of Kotter  to the adaptation and positive design are to be expected. But you can also positive effects. If activities that do not generate immediate human benefits but are attributable to transaction costs are largely technicalized, then you do not have to regret this and on the contrary try to speed up this process. The benefits of these changes could be huge. Now a not-too-distant future is imaginable in which there is no more poverty, wiping out diseases that have tormented humanity for millennia, finding a solution to climate change and making new forms of communication and cooperation a reality But looking at the same technological revolution, we can also ask ourselves: do not we just go to a bleak future where robots and automation are destroying millions of jobs, where income inequality leads to an unbridgeable gap As there is a constant threat to public safety, and our privacy is being undermined by aggressive surveillance and uncontrolled personal information gathering, a Chapman University study reveals that technology is the only issue of concern to people second is ranked. By contrast, cyberterrorism, the fear of corporate and government private-data gathering, and identity theft are among the top ten issues. The survey also shows that nearly one in three respondents fear losing their jobs to robots, and one in four doubts that artificial intelligence can be trusted. This widespread subliminal fear is quite understandable. And that’s why it’s so important to make Ethical Principles binding on the use of these new technologies and methods.
From my point of view, Microsoft and its CEO Satya Nadella have come up with a very good proposal here would like to share here: Six ethical principles
A.I. must be designed to support humanity.
A.I. must be transparent
A.I. must maximize efficiency without violating the dignity of man
A.I. must be designed for smart privacy
A.I. must have algorithmic accountability so that inadvertent damage can be corrected.
A.I. must ensure impartiality and representative research so that false heuristics can not be used to discriminate
(Satya Nadella / Microsoft – 6 ethical principles for AI)
And how was that with the cows now?
Yes I already know hard cut of 6 ethical principles to cows but does not our meat consumption have anything to do with ethics? I think meat is way too cheap nowadays. If REWE with pork tenderloin (ok no cows …) for 6,90 € / kg advertises something is wrong and here animals are slaughtered that actually deserve to live. Do not get me wrong: I’m an avowed meat eater – preferably cows – but that goes too far and it just can not be right.
But let’s get to the subject: In 2013 there was an IoT project in Japan with the help of which Japanese farmers have managed the desire to breed cow – or better wish calf – to breed. Not from the DNA but from the sex. In industrial cow production, the female calf is more interesting than the male, so each farm tries to breed as many female calves as possible. Yes, even with Wagyu. It has now been determined on the basis of many data (Big Data) and ML that a female cow 16h before the optimal time for fertilization by a male cow moves extremely much and is very active. In addition, it has been found that the probability of a female calf is extremely high at the time up to 2 hours before the optimal time and that a male calf is raised up to 2 hours after the time. Usually, this start time is always between 22.00 o’clock in the evening and 8.00 o’clock in the morning. To meet exactly this point in time, the farmer would either have to be awake all night, or provide the cows with a smartwatch that records these movements and informs the farmer the best time for insemination by a bull the next morning, and the farmer has a close 100% probability for a female or male calf. And whether this is the optimal use of AI and ML and whether this method of procreation is now ethically acceptable, we can now discuss here ?