List Posts

Our AI work on Story Telling

In September, 2022 we will use a version of the ‘Transformer’ Model to initiate our attempt at Story Telling with AI. We will explain and employ Transfer Learning and Fine Tuning to accomplish this. What is uncertain now is; 1) What ‘Stories’ to use for Fine Tuning, and 2) What ‘Transformer’ Model we will use.

Read More

Executing NLP

On the right is sentiment analysis using BERT, a Transformer Model which we will encounter in a future post. Below we use N-Grams to Generate Shakespeare. This approach preceded the use of LSTMs (Long-Short-Term-Memory) used for Artificial Shakespeare. Currently, ‘Transformers’ are the preferred Model for NLP and vision. Following is an introduction to the Mechanics of Natural Language Processing. https://vimeo.com/364079108 Forward and Back Propagation. Back Propagation is the bain of Data Scientists!

Read More

Story Generation is Hard

We have seen the progress in Generating Art, Music, and Images using State of Art AI. How did you fare with the quiz on Artificial Shakespeare? The Beauty of Language demonstrated the elements of magnificent poetry from Dante’s Inferno and narrative development from Jane Austin’s Pride and Prejudice. Your experimentation with Experiment with Story Telling convinced you of the inability of AI to Generate Stories. Prehistoric Cave Drawings https://vimeo.com/720328798 This is a nice state-of-art AI challenge. Summarize a Shakespeare play and represent it in 3 Panels. What is missing in the current AI approaches? A Midsummer Night’s Dream https://vimeo.com/720341855 Do…

Read More

State of Art AI

Human vs AI Test: Can We Tell the Difference Anymore? A survey from March 2022 on the power of AI generated art to appear real. Take the test! Hit the back button to return to this site. The study encompasses AI generated; photos, artwork, music, text, and translations. Included are brief definitions of Artificial Intelligence, Machine Learning and Deep Learning. The examples were generated by Generative Adversarial Networks (GANs).  GANs are two networks (a generator and a discriminator) that attempt to outsmart each other. In the case of images, the generator creates an image and the discriminator evaluates if it ‘looks’…

Read More

Artificial Shakespeare

In The Beaty of Language Post we discussed the difficulties of AI Generated Storytelling. A story has a developing plot, modulating moods and relationships, and creates tension and resolution. All these pieces somehow ‘fit together’. The complexities are beyond the capabilities of current AI. If story telling is too difficult, what about generating a few lines of Shakespeare? See if you can distinguish the real Shakespeare from AI Generated Shakespeare here: Take the quiz! Please hit the back button to return to this page. How did you do? OK, let’s step it up to ‘The AI Poet’. This describes the…

Read More

Experiment with Story Telling

Now its time for you to experiment with the OpenAI model GPT-2. This is available at: https://app.inferkit.com/demo Unfortunately, until recently, there is not a constraint on model size or training time when evaluating “best models”. This has led to an arms race in size represented by the graph. The following graph lists “best model” by year against the number of parameters (in billions). We use the GPT-2 Model through an interface created by InferKit. The GPT-3 Model is currently restricted because of size. Exponential increase in Model Size I will address some of the models in the graph below in…

Read More

The Beauty of Language

Imagine the expanded view and richness a child derives from being read too. What describes a great novel? Sandow Birk’s early 2000s illustrations of the Divine Comedy A developing plot, modulating moods and relationships, and creating tension and resolution. All within a unified whole of 350 pages in the case of Jane Austin’s Pride and Prejudice. Story Generation is a difficult problem. The complexities are beyond the capabilities of current AI. We will chart the progress in AI generated stories in future Posts. The Beauty of Language Video: https://vimeo.com/360606506 Now for a bit of Fun. Jack’s ‘Speed Dating’ Video: https://vimeo.com/264668962

Read More

Jackson Story Generation

We use props to ‘move the story along’ and got unexpected responses. Our ‘Capture Memories’ videos will include the parent or grandparent. The video link is below. https://vimeo.com/355600063 Jackson Story Generation A collection of treasured and unexpected moments. https://vimeo.com/360970768 Dark Clouds https://vimeo.com/360971062 DONE! https://vimeo.com/360971607 I did it! https://vimeo.com/360971935 A bop pue https://vimeo.com/360972198 connect connect connect https://vimeo.com/360971232 I don’t want to! Perhaps we could do the same for you. For information Contact Us

Read More

Voyage of Discovery

We extend an offer to join our Voyage of Discovery. Like Columbus we are unsure of our destination and what, if anything, will be achieved. Our objective is to use technology for good. How so? We wish to use books to develop pre and early readers through story generation. Story Generation through Artificial Intelligence (AI). The three essential components for children are 1) Books, 2) Dedicated Time, and 3) Focused Interaction. We will discuss how we intend to achieve Focused Interaction with the child in succeeding Blogs. The Voyage of Discovery Video: https://vimeo.com/360633647

Read More

Creating Images (Outpainting)

OpenAI has created DALL.E an AI system which generates images and artwork from a natural language description or extends original images to a larger canvas. Here is an example of Outpainting Johannes Vermeer’s Girl with a Pearl Earring. Original Vermeer AHA, After Outpaining Notice that the image was created with the same shadows, reflections, and textures. The capability is available for you to try: https://openai.com/blog/dall-e-introducing-outpainting/

Read More

Leave a Reply

Your email address will not be published. Required fields are marked *