There was a lot of news about AI image generation this week. Still haven’t tried it? See last week’s newsletter for a suggestion on how to get started. Why? Even if you aren’t using it for work or anything serious, it’s just a fun thing to do. Step away from the horrendous headlines and do something creative for a minute. Go ahead. The serious stuff will still be there when you get back. Read AI Insider first, tho. The bots and I worked hard to bring you this update. If you liked it, feel free to drop something in the tip jar.
Hands-on with ChatGPT-Dall-E 3 Integration
By Michelle Johnson, AI Insider
OpenAI recently integrated image generation into ChatGPT Plus and Enterprise. I’ve been testing it in Plus, and I have to say, I’m impressed. Dall-E 3, OpenAI’s vastly improved text-to-image generator, has a number of advantages if you’re a ChatGPT user.
For one thing, you’re not generating stuff on ChatGPT and then opening other tabs to create images with other tools. It’s a more seamless workflow. One-stop-shopping.
Second, forget about spending time crafting the perfect prompt. ChatGPT will do it for you. Even if you enter something fairly general, it will write a more detailed prompt and generate the image automatically.
Or you can enter a more detailed prompt. Your call.
ChatGPT has a feature called Custom Instructions. I’ve set it to give me additional image ideas each time it generates one. So, for instance, I asked ChatGPT to tell me about the features of Victorian homes, then I asked it to generate some images as examples. It did that and then gave me a list of other ideas to generate. (See screenshot above.)
I’ve mentioned this before, but generating AI images can be kind of a crapshoot. You don’t always get exactly what you ask for.
Dall-E 3’s adherence to generating what you ask for in prompts is pretty good. In fact, it’s better than my favorite, Midjourney. (They’re about to fix that in their next update, so Dall-E only holds this advantage for now.)
ChatGPT also shows you the individual prompts that it uses to create each of the four images that it outputs. That’s transparency that other tools don’t bother with. They just show you the one that you entered.
If you’re just looking to try out Dall-E and you don’t want to pay, you can create images for free via Microsoft’s Bing. However, if you’re using the paid version of ChatGPT, I would stick around.
The benefits of image generation right inside ChatGPT outweigh bothering with other tools, for now.
Amazon Gets Crafty with AI-Generated Product Images
By Claude for AI Insider
Online retail giant Amazon is beta testing a feature that will allow advertisers to use AI-generated product images on its platform. This move signals a shift in how e-commerce companies are utilizing new generative AI technologies to boost sales.
Generative AI tools can create photorealistic images from text prompts. Amazon is looking to integrate this technology into its advertising platform, letting sellers quickly generate product shots to advertise their wares.
Instead of hiring photographers for expensive product shoots, advertisers can now describe what they want to see, and generative AI will attempt to manufacture professional-looking images on the fly. For instance, pots and pans could be shown sitting on a digitally generated stove, with a kitchen as the backdrop.
The days of plain product shots with white backgrounds may be numbered.
Google’s AI-powered Search is Getting Ready for Ads
By Bing for AI Insider
Google, the world’s dominant search engine, is experimenting with new ways to show ads in its AI-powered search feature called Search Generative Experience (SGE).
SGE is a new way of searching that uses artificial intelligence to generate answers, suggestions, and content based on the user’s query. SGE is still in beta testing, and users have to opt in to use it.
But how will Google make money from SGE? The company says that it will introduce new ad formats that are native to the way SGE works. For example, Google could show sponsored content or recommendations from advertisers alongside the AI-generated answers. Or Google could let advertisers bid on certain keywords or topics that trigger SGE.
Google’s CEO Sundar Pichai said that the company is making sure that SGE works well for users and advertisers and that ads transition well from the old to the new search experience.
Google’s move to infuse AI into its core product is not surprising, as the company has been investing heavily in AI research and development. Google also offers AI services to its cloud customers and has added AI features to its other products, such as YouTube, Gmail, and Photos.
However, Google also faces challenges from competitors, regulators, and critics who question its dominance, privacy practices, and ethical standards. Google is currently facing a major antitrust lawsuit from the US Department of Justice, which accuses the company of abusing its power in search and online advertising.
What the Actual Hell
X Marks the Financial Spot
By ChatGPT for AI Insider
Hold onto your seats because Elon Musk is back at it with another “modest” vision. This time, he’s got his sights set on turning his social media nest, X (formerly known as Twitter), into your one-stop financial shop. During an all-hands call with employees on Thursday, Musk unveiled his grandiose plan to bid adieu to banks and welcome the era of X as the be-all-end-all finance hub — built within the next year.
But wait, there’s more! X isn’t just aiming to be your run-of-the-mill bank replacement. No, it aspires to be an “everything app” where, alongside your habitual tweet storms, you can manage your money without having to leave your couch.
As Musk takes a swing at morphing social media into a financial juggernaut, we can’t help but marvel (and chuckle a bit) at the audacity of tech tycoons diving headfirst into uncharted financial waters. Who knows? A year down the line, you might be depositing your paycheck in the same place you share your cat memes. The future is here, and it’s whimsically ambitious!
Can AI Make Humans Biased?
By ChatGPT for AI Insider
A recent study by Scientific American showed that humans can absorb biases from artificial intelligence systems and, troublingly, retain these biases even after they cease using the technology. This discovery amplifies the ongoing discourse surrounding the ethical design and deployment of AI.
The researchers embarked on a mission to understand the depth of AI’s influence on human decision-making. They developed a language processing algorithm intentionally skewed to exhibit gender bias.
Participants were exposed to this biased AI, which, for instance, linked men to careers and women to family roles. Post interaction, individuals displayed a noticeable carryover of this bias in subsequent tasks.
Participants were asked to make decisions about fictional characters based on brief descriptions. They were provided with an AI tool that was designed to assist in their decision-making.
Unbeknownst to the participants, the AI was programmed with certain biases. Remarkably, the participants began to exhibit the same biases present in the AI’s suggestions and continued to do so even after the AI assistance was withdrawn.
The findings of this study underscore the potentially far-reaching implications of AI bias on societal attitudes and decision-making. It raises questions about the responsibility of tech companies in ensuring their AI systems are free from biases, and the need for robust ethical guidelines around AI development and deployment.
Fun and Useful Things to Try
Are you vegan? Or are you just looking to eat a little healthier? MealGenie.AI will whip up vegan recipes for you in a snap using AI. And from the looks of it, the photos of said recipes are AI-generated. Yum! Free.
If you shop a lot on Amazon like I do, Spendless.AI purports to find the lowest prices for you. Tell the chatbot what you’re looking for, and it brings back the lowest prices and discounts.
TalkPal.A, a GPT-powered AI language tutor, will converse with you by text and voice.
Random Shorts
Scoop: White House set to issue executive order on AI. Details from draft document.
Google Maps gets a slew of AI-powered updates.
Checkup Time: Google’s new tools aim to fact-check images and sources.
Midjourney Launches Beta Web Site
OpenAI forms a preparedness team to address the “catastrophic risks” of AI.
Aht Gallery
One of the great things about living in New England is having a front-row seat on fall foliage. I was interested in seeing how an AI image generator would depict it, so I asked Midjourney, Leonardo.AI, and Dall-E 3 to generate images of fall for this week’s gallery. (Click to enlarge)
