AI Insider No. 20

What I’ve Learned Editing AI ‘Reporters’

By Michelle Johnson for AI Insider

If you’re a regular reader, you’ve likely noticed the bylines in this newsletter attributed to ChatGPT, Bing, Bard, etc. That’s because when I was reluctant to start doing a newsletter (I’m retired! Too much work!), I decided to lighten the workload by “hiring” the bots as reporters.

How It Started: Me thinking, how hard could it be? I have an editing background. I’ve been a journalism prof. I certainly know how to whip copy into shape. If I don’t have to write it all, it won’t take much time.

How It’s Going: If I were grading the bot reporters, I’d give them a B. Nothing ever runs unedited. Most of the time, they can gather facts, but those facts are not always related to the assignment or even current. To be honest, it’s not always the bots’ fault. They usually do better if I tweak the prompts to give them more explicit instructions. And, like a lot of beginning reporters, they are still learning to write in journalistic style.

Even when you specify that they should write like a journalist, sometimes they default to sounding like an upbeat PR release or an academic paper.

Here’s a look at how it typically goes.

Bing: These are just some of the exciting announcements from Meta Connect 2023. It’s clear that Meta is investing heavily in AI technology to enhance our digital experiences.”

Edit: I axed the entire paragraph. This is a weak and opinionated “kicker,” aka, closing paragraph. And Meta is investing in AI to generate a profit. “Enhancing our digital experiences?” Oh, please.

ChatGPT: “If Ideogram succeeds in resolving the text issue, they could potentially stand out from the crowd — at least until the others catch up. Rest assured, the competition is actively working on solutions.”

Edit: I let this stand. It’s authoritative, and the “rest assured” is quirky but not too over the top. And it’s on point. It’s exactly what’s likely to happen. All of the AI image generators will soon be generating readable text on images.

I’m interested to see how this workflow evolves as the bots get smarter and better at writing. For now, it’s all just a grand experiment.


Unhappy With Your Yearbook Photo? Make a New AI Pic

By Bard for AI Insider

Yearbook photos are a rite of passage for many students. They are a chance to capture a moment in time and preserve memories of school days.

Did you hate your yearbook photo? Well, what if you could go back in time and create one that you love? Or just generate a fantasy shot?

That’s exactly what the EPIK AI photo editor app allows you to do. It’s available for Android and iOS.

This photo editing app went viral recently after launching the yearbook feature. Using AI, it can generate yearbook photos of you from the 1920s to the present day. 

You use it by uploading photos of yourself and selecting a hairstyle and clothing style from a variety of eras. It can generate 60 different images using eight to 12 uploaded selfies.

Apps like EPIK AI can be a fun and lighthearted way to use AI technology and a reminder of how much fashion and culture have changed over the years. 

Here are a few tips for creating an EPIK AI yearbook photo:

  • Choose a hair and clothing style that reflects the era you’re going for. The app has a wide selection of options to choose from.
  • Experiment with different poses and expressions. The app will generate multiple photos for you to choose from, so don’t be afraid to try different things until you find a photo that you love.
  • Add some fun accessories. The app also has a selection of accessories that you can add to your photo, such as glasses, hats, and jewelry. 

Once you’ve created your AI yearbook photo, you can share it on social media and with your friends and family. Diploma not included.


This image has an empty alt attribute; its file name is Firefly-question-marks-rising-from-flames-94723-1-300x300.jpg

What the Actual Hell? Bon Appetit Bots!

By Bing for AI Insider

Have you ever wondered what a robot would order at a restaurant? [Ed. note: No. Never.] Would it go for a juicy burger, a crunchy salad, or a sweet dessert?

Well, it might depend on how it feels and what it likes, just like us humans.

Don’t laugh. We could see this in action someday. A team of researchers from Penn State is working on an AI system that can mimic our sense of taste and how it influences our eating habits.

They have created an electronic “tongue” that can detect different flavors and respond to them based on both physiological and psychological factors.

This electronic tongue is made of tiny sensors that can convert chemical data into electrical impulses, just like the taste receptors on our tongue. These impulses are then sent to an electronic circuit that simulates the brain’s gustatory cortex, where we process and perceive taste. The circuit also has neurons that represent hunger, appetite, and feeding behavior.

The researchers have tested their device with different tastes, such as sweet, salty, sour, bitter, and umami. They found that the device can react to these tastes differently depending on its physiological and psychological state. For example, if the device is full but detects something sweet, it might still want to eat it because of its psychological preference.

The researchers hope that their novel electronic tongue can provide a new way of designing AI systems that are more human-like and emotionally intelligent. They also envision various applications for their device, such as helping people with weight loss, creating personalized menus, or even training AI to be better wine tasters.


[Ed. Note: To fully grasp Magic Studio’s capabilities, check out the demonstration video here.]

Canva Doubles Down on AI with Magic Studio

By Bing for AI Insider

Canva, the popular graphic design platform, has just celebrated its 10th anniversary by launching a new AI-powered design platform called Magic Studio. Magic Studio offers a suite of AI features that can help you create designs, images, and videos from text prompts.

Included in this major update:

Magic Switch: A tool that lets you convert any piece of content into a different format with one click. For example, you can turn a whiteboard of ideas into a presentation, or a blog post into a social media post.

Magic Media: This tool lets you generate images and videos from a text prompt. You can use it to create visuals for your projects, such as logos, icons, illustrations, animations, and more. The video-generation capabilities are powered by Runway, an AI video generator.

Magic Design: A tool that lets you generate an entire design from a text prompt and a color scheme. You can use it to create posters, flyers, brochures, presentations, and more. You can also use it to generate short videos from your clips and images, with music recommendations included.

Magic Studio is available for free for Canva Pro and Canva for Teams users. Free users can also try some of the Magic Studio products without charge.

Canva also announced a Creator Compensation Program, which will pay Canva Creators whose content was used to train the company’s AI models. Creators can also opt out of having their content used for training.


BBC Announces Policy for Using Generative AI

By Bing for AI Insider

The BBC, the UK’s largest news organization, has announced its plans to explore the use of Generative AI in its content creation and delivery. Generative AI is a type of artificial intelligence that can produce new content, such as text, images, audio, and video, based on existing data and text prompts.

In a blog post, the broadcaster said it will follow three guiding principles: acting in the public’s best interests, prioritizing talent and creativity, and being open and transparent. The BBC also said it will work with other stakeholders to ensure the safety and benefits of generative AI.

The company will start several projects that will test how Generative AI can support or transform its activities in areas such as journalism, archiving, and personalized experiences. For example, Generative AI could help produce more diverse and relevant content for different audiences or enhance the quality and accessibility of archival material.

However, while it is embracing some aspects of AI, the BBC has also blocked web crawlers from tech companies such as OpenAI and Common Crawl, from accessing its websites to train their bots. The broadcaster said this is to protect the rights of license fee payers and prevent unauthorized use of its data for training AI models.


(Google)

Google Assistant Paired Up with Bard AI Chatbot

By Bard and ChatGPT for AI Insider

Google Assistant is slated for a significant upgrade courtesy of a new collaboration with Google’s AI chatbot Bard. Announced at the recent Made by Google event, the enhanced Assistant with Bard is under development and expected to be available in the coming months.

Like Alexa, Google Assistant is a voice assistant, activated by uttering, “Hey, Google” or “OK, Google.”

According to Google, Assistant with Bard will be able to understand and respond to more complex and nuanced questions, as well as help with tasks that require creativity and problem-solving skills. For example, it might be able to help you plan a trip, write a social media post, or even create a new product idea.

One of the key features of Assistant with Bard will be its ability to access and process information from multiple sources, including your personal data, Google’s knowledge base, and the real world. Google says this will allow it to provide more contextually aware and personalized responses.


Aht Gallery

Leave a Reply

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

This site uses Akismet to reduce spam. Learn how your comment data is processed.