30.11.17

Bye bye Charlie

Charlie Hebdo Deutschland 1.Dezember 2016 bis 30.November 2017
(download große Bilddatei (2.91 MB))

15.11.17

Bezirksrat Südstadt-Bult November 2017

Der Piratenvertreter hatte 2 Anfragen zu Luftschadstoffimmissionen in der Marienstraße sowie zur Umsetzung von Elektromobilität im Stadtbezirk, und 2 Anträge zu Einrichtungen für Jugendliche auf der Hoppenstedtwiese sowie zur Aufstellung von Fahrradbügeln, die beide Zustimmung der anderen Bezirksratsmitglieder fanden. Daneben quälte er die CDU mit 3 Änderungsanträgen. Von deren 10 Anträgen hatten nur 3 Erfolg (ein Sachstandsbericht zur Verbesserung des Verkehrsflusses, die Aufstellung einer Tischtennisplatte sowie fehlende Radwegemarkierungen), einer wurde zurückgezogen und der Rest – die meisten gingen um die Schaffung zusätzlicher Parkplätze - fiel durch. Der AfD-Vertreter machte sich während der Sitzung kaum bemerkbar außer dass er sich bei den Beschlüssen zum Bezirksrathaushalt enthielt.

Sitzungsunterlagen | Protokoll

8.11.17

Mongolei

"Berlin (im Botanischer Garten) 12.03.2004"
... Foto gefunden in öffentlichen Bücherschrank...

1.11.17

A New Way to Create Human-Like Artificial Intelligence

I have an idea that can be used to create a smart AI. The AI will have very good linguistic intelligence.
It will be able to read and extract ideas from books. It will also be able to have conversations with people.

The idea is to create a website that is used to gather data from at least one million users. Each user will see a random word and will have to say what they feel when they read the word. The users will also have to rate how intense their feeling is, on a scale from 1 to 100. The user will also be shown groups of two, three and four words, and they will be asked what emotions they feel when they see them.
In addition, the users will be shown sentences. The users should be allowed to say what these sentences mean to them, and what emotions they feel when they read one of the sentences. The same thing should be done for groups of two and three sentences.

Furthermore, for every word from a list of the 5000 most used words in the English language, people will be asked to add another 5-10 words that come to mind when they read the word, and then rate how close each is to the initial word.

Another kind of data we want to gather is to identify which sentence someone would say next in a conversation. So the user will see a sentence and then have to add another 1-3 sentences that are appropriate to continue the conversation. We should also give the user a context together with each sentence or group of sentences. Each sentence will be shown to groups of 1,000 people. We should only use an answer if it was given by at least ten people. That way we know that it wasn't written by mistake and that people usually talk like that. This will allow us to know build a very smart AI that knows exactly what to say next when it talks to people.

We will also need to gather data about how important a sentence is in many different contexts.

Moreover, the user should be  asked which words in a sentence are more important. Even better would be to ask a user to give each word in a sentence a rating of importance from 1 to 100. Another thing that can be done is to allow the user to rate the words for groups of 2-5 sentences. Using this data, the AI should be able to compare two sentences and decide which one is more important. In that way it will be able to summarize a text.

Because the AI should also be able to tell what data resembles other data, we should gather that information from the users. It should be able to detect even slight similarities between sentences/ideas. To be able to do this, the AI needs to know what is similar between every two sentences. It should know which ideas are similar, and how similar they are.

Create a reward system on the website to motivate people to contribute more. Give points for every answer people supply, and show these points to everyone else on the website. Because there may be users who don't want to create an account on the website, allow anonymous users to contribute. Also, allow people to know how much time they are spending on the website to increase their motivation. We should allow the users to chose if they want to add input only for words, groups of words, sentences, or groups of sentences.

I believe that people will want to contribute because in the long run they have a lot to gain if the project succeeds. Many people will want to advertise the existence and purpose of this website on their websites.
The data that is gathered should be available for free to everyone who wants to build an AI.

The sentences and groups of sentences that are presented to people will be taken from books, Wikipedia articles, and online conversations.

The gathered data will enable the AI to know what is important in a book it reads and what it is not.

We have to teach the AI everything that a person sees as important, and everything that he or she sees as less important.
A human being has in his or her brain all the possible combinations of words, what each word means, and what feelings a word or groups of words generate. If we give software all this knowledge, we will have an AI that can make decisions just like a human being does. For this we need hardware that is extremely fast. The hardware may have to be optimized for searching and sorting algorithms. We may even need another kind of hardware to efficiently store all this data and then be able to access it very quickly. In this way, we can build an AI that can read books and write texts as well as a human being.

If you want to create a AI that is a psychologist ask 50,000 psychologists to create accounts on your website. Then ask them to give you the data you need. The same can be done if you want to create a historian, a medic that can diagnose patients by talking to them, or a literature teacher.

I believe that the only way to build a very smart human like AI is by gathering a lot of data from millions of people.

Kind regards,
***