Last Week in AI #98

AI to create images from text, how to build AI for good, and more!

Mini Briefs

This avocado armchair could be the future of AI

OpenAI showcased a new AI system, DALL'E, that is able to generate images from image captions. The system is a smaller version of GPT-3 but trained on image caption data - the input is a paragraph describing the image, and the output are a low-res version of the target image that is then upsampled with another neural network. What's novel about this system is its ability to generate images of combinations of concepts that were not seen in the training data. For example, a drawing prompt of "an avocado chair" generates chairs plausibly shaped like avocados, with the pit as a seat cushion. Similar creative generations can be seen with wilder prompts like "a baby daikon radish in a tutu walking a dog" and "a snail made of harp." While experts caution to take these results with a grain of salt, since details of the system have not been released via a research paper, the results are impressive nonetheless and point to exciting progress in the world of AI generative models.

Why we must democratize AI to invest in human prosperity, with Frank Pasquale

In an interview, Frank Pasquale, Professor of Law at Brooklyn Law School, proposes and elaborates on what he thinks ought to be the new "laws" for autonomous robots and AI. While many sci-fi enthusiasts may be familiar with Isaac Asimov's three laws of robots, these laws assume that "the development of robots that are indistinguishable from persons" and "try to make sure those robots can’t hurt anyone or hurt them too badly." However, it is clear now that autonomous robots that are being deployed now and in the near future are quite different from humans, so they require a different treatment.

Professor Pasquale's four laws are:

First, robotic systems and AI should complement professionals, not replace them. Second, they should not counterfeit humanity. Third, they should not intensify zero-sum arms races. And fourth, they must always indicate the identity of their creators, controllers, and owners.

Podcast

Check out our weekly podcast covering these stories! 
Website | RSS | iTunes | Spotify | YouTube

News

Advances & Business

Concerns & Hype

Analysis & Policy

Expert Opinions & Discussion within the field

  • What Buddhism can do for AI ethics - "Buddhism teaches us to focus our energy on eliminating suffering in the world. The explosive growth of artificial intelligence has fostered hope that it will help us solve many of the world's most intractable problems."


That's all for this week! If you are not subscribed and liked this, feel free to subscribe below!

Recent Articles:

Copyright © 2021 Skynet Today, All rights reserved.