Apogee Suite: AI-Powered Legal Document Research Platform
By VICTOR ANJOS
Welcome to my multi-part series on AI as a creator. In this series we aim to look at whether today’s artificial intelligence is capable of devising new works of arts.
We realize that Art is completely in the eye of the beholder, so what may be considered art to some is different than others. Our definition of art will be loose and we will use it to mean the following:
In today’s part of the series we are examining:
Photos, Memes, Emails and Code
Creativity may be the ultimate moonshot for artificial intelligence. Already AI has helped write pop ballads, mimicked the styles of great painters and informed creative decisions in filmmaking. Experts wonder, however, how far AI can or should go in the creative process.
When we recently spoke to AI experts and thought leaders, their opinions varied as to whether AI has the potential to become a true creative partner or even the creator of solo works of art. While this debate will likely continue for some time, it’s clear that as digital content and delivery platforms continue infiltrating all forms of media and expression, the role of AI will undoubtedly expand.
AI can make “dank” memes
When it comes to artificial intelligence, some people are worried that they might attack us or take our jobs. However, it turns out that they might also supersede us at making memes. A new deep neural network approach has demonstrated that AI can produce funny and relevant captions when given a meme image. What a time to be alive.
The algorithm is called “dank learning” after dank memes, a slang term that refers ironically to the most bizarre memes or memes that are now so overused that they have lost their original comedic value. That said, the program is capable of breathing new life into the memes of bygone days. The project was created by Stanford students Abel L. Peirson V and E. Meltem Tolunay.
This AI approach is known as machine learning. The software was not explicitly programmed to create memes, but to perform certain actions and create new outputs based on what it has learned. In this case, the team used about 2,600 unique memes, with up to 160 different captions. Based on this extensive library, the algorithm began generating memes.
And the memes created by the AI seem real. You can’t tell these and the non-AI generated ones apart at all. When 20 different memes were shown to five individuals from diverse backgrounds, they could hardly distinguish the real memes from the AI-generated ones and gave them similar ratings of hilarity.
“Obtained results indicate that memes can be generated that in general cannot be easily distinguished from naturally produced ones, if at all, using human evaluations,” the researchers wrote in their paper. “We acknowledge that one of the greatest challenges in our project and other language modeling tasks is to capture humor, which varies across people and cultures.”
The algorithm is not perfect. One issue the team would like to solve is teaching the AI to realize where the breakpoint in the text should be. In this project, those were chosen by the researchers, but they believe that once a program can do that, it will be capable of autonomously creating memes.
The researchers also noted that in the process of trying to learn from the dataset, the AI began to use expletives as well as racist and sexist terms. This has been seen before (read more here) and shows how learning algorithms, just like people, are strongly influenced by what they are exposed to.
AI Generated Photos used instead of stock photos
It’s getting easier and easier to use AI to generate convincing-looking, yet entirely fake, pictures of people. Now, one company wants to find a use for these photos, by offering a resource of 100,000 AI-generated faces to anyone that can use them — royalty free. Many of the images look fake but others are difficult to distinguish from images licensed by stock photo companies.
The project’s Product Hunt page lists the team at Icons8, a designer marketplace for icons and photographs, as the creator of the project. The AI-produced images are intended to be used as design elements in anything from presentations to websites and mobile apps. Everything is free to use with link attribution back to generated.photos.
Over the course of the year, we’ve seen a number of AI projects generate fake AI faces, most notably ThisPersonDoesNotExist.com, a website capable of producing an infinite series of mostly-believable headshots. The faces found on generated.photos cover a variety of ages, shapes, and ethnicities, and they’re all consistently lit and consistently sized to make them useful for designers.
The project is currently in its early stages, and Icons8 product designer Konstantin Zhabinskiy notes that some of the faces might look a bit “off.” You can see some examples of this in the images above where the AI has produced a mangled hand and background in one image, and given another fake model a wound to their forehead. However, the team’s hope is to eventually produce a simple API that can easily generate new photographs based on a variety of inputs, allowing designers to quickly produce photorealistic images to illustrate their work without having to worry about copyright or model releases.
Zhabinskiy is keen to emphasize that the AI used to generate these images was trained using data shot in-house, rather than using stock media or scraping photographs from the internet. “Such an approach requires thousands of hours of labor, but in the end, it will certainly be worth it!” exclaims an Icons8 blog post. Ivan Braun, the founder of Icons8, says that in total the team took 29,000 pictures of 69 models over the course of three years which it used to train its algorithm.
There are valid concerns about technology that’s able to generate convincing-looking fakes like these at scale. This project is trying to create images that make life easier for designers, but the software could one day be used for all sorts of malicious activity.
Can AI do better than generated language and text?
A new AI from Microsoft automatically generates to-do lists based on sent emails. Researchers from the University of Washington and Microsoft’s AI team announced the tool in a pre-print research paper.
The Smart To-Do tool uses AI to scan all outgoing emails for usable text, then converts these tasks into an automatically generated to-do list. Suppose you send an email to a customer saying that you will send a draft on Tuesday afternoon. The AI scans the email, registers the task and puts this task on the list with, for example, the title ‘send a concept to customer’, together with the time, in this case, Tuesday afternoon. In this way, the tool scans all emails and updates the list continuously.
The researchers had to build the system from scratch as there is no comparable AI system. This meant that Microsoft had to develop everything from the task parameters to the datasets on which the deep neural networks were trained. If the new tool were to work well in practice, this would save a lot of work for employees. They will then be able to keep answering emails instead of making to-do lists.
Summing it up
We’ve seen that AI currently is able to create many things. This month we’ve examined AI as a creator or Art, Food, Music, Novels, Emails, Recipes and more, but what about AI creating code and AI. Is that even possible? In 2018, Bayou launched and is a deep-learning system that can write code for programmers and generate API idioms for complex databases. It teaches itself how to code through GitHub, training with millions of human programmers using Java. It can interpret and recognize high-level patterns in hundreds of thousands of Java programs through an artificial neural network method called Neural Sketch Learning. Developers can initialize variables that are intended to be used in the programming task or submit a query that includes names of API methods or the type of variables for the programming task.
As you can see, even AI generating AI is in the future. This is all at the start of “creation” by AI but expect that as more people become data and AI literate and as we see a shift from Computer Science and Software Engineering to AI Engineering, we’ll get there faster and faster.
So what do you do when all signs point to having to go to University to gain any sort of advantage? Unfortunately it’s the current state of affairs that most employers will not hire you unless you have a degree for even junior or starting jobs. Once you have that degree, coming to my Modular Lab Program, with 1000ml with our Patent Pending training system, the only such system in the world; is the only way to gain the practical knowledge and experience that will jump start your career.
Check out our next dates below for our upcoming seminars, labs and programs, we’d love to have you there.