Welcome back to AI Insider, the newsletter that owes its existence to my Facebook friends who asked for it. And if you’re not here via my FB page, welcome to you, too. Be sure to visit the archival website for an intro to who I am and an explanation of why chatbots have a byline here. OK, let’s get on with it!
AI Image Generators Xant Wrxte
By Michelle Johnson
Take a look at the image above. See the part that looks like text, but not?
It looks garbled because surprisingly, AI image generators haven’t quite nailed how to replicate text within an image.
That’s right. It’s kind of hard to believe the technology that can produce this:
Has trouble with this:
So, what’s this that almost got it right?:
Well, developers are working to solve the garbled text problem and it looks like a startup called Ideogram is making some headway.
Launched by former Google Brain researchers with $16 million in seed funding, Ideogram joins a bunch of other AI image generators out there, including Midjourney, Stable Diffusion, Adobe Firefly and Leonardo AI, to name a few.
If Ideaogram nails the text thing, they might be able to stand out in that crowd. Well, at least until the others fix it. And you can best bet they are working on it.
Judging by the excited discussions happening on Ideogram’s Discord server, the improved text feature is a hit. (Or a miss sometimes, but nobody seems too upset about that.)
While the companies are hashing this out you can sign up to try Ideogram’s beta and watch it evolve. Or ping me and I’ll send you an invite.
Are AI Chatbots Losing Their Charm?
This Vox piece speculates that interest in chatbots is waning, or that most of them (other than ChatGPT) didn’t really take off in the first place. True? Who better to ask about this than a chatbot? I asked, and Claude weighed in below.
By Claude for AI Insider
AI is going through a bit of an awkward phase right now. Just a few months ago, ChatGPT exploded onto the scene and became the hottest thing in tech. The conversational AI wowed people with its human-like responses and its ability to generate articles, stories, and even computer code on demand.
But after riding a wave of hype and frenzied attention, things have cooled down. Interest in ChatGPT peaked back in December and has tapered off since then according to Google Trends data. So what happened?
In a nutshell, the hype got ahead of the technology. ChatGPT is impressive but, like most of the chatbots, it has some obvious flaws. The bots make factual errors. They can be manipulated into generating biased, harmful or nonsensical text. And while their conversational skills are remarkable, they don’t actually understand language or the world. [Ed. Note: I beg to differ, Claude! Y’all have been trained to understand some things about our world.]
Major tech companies are integrating large language models into products like Bing and Google Search. But so far, the integrations feel limited. It’s not like we have AI butlers ready to fulfill our every informational need. [Ed. Note: Again, I have to beg to differ. I would argue that both Bing and Bard have improved over time and other models like Perplexity were pretty good right out of the box. Claude, you’ve improved, too, so chill.]
This doesn’t mean AI won’t transform fields like search, medicine, and imaging. But early demos probably created unrealistic expectations about how quickly AI would revolutionize everything. The technology needs more time to mature. [Ed. Note: True! All of these tools are betas. They are infants and toddlers. We are watching them grow up. Sometimes they stumble.]
AI still has tremendous potential, but realizing it will likely be a gradual process with occasional breakthroughs – not an overnight revolution. The hype cycle got ahead of reality, but that doesn’t mean we should dismiss this technology or underestimate its possibilities. Right now AI is going through some awkward growing pains, but the best is yet to come. [Ed note: Agree! Although, major breakthroughs will happen faster in some areas like medicine, IMHO.]
Source: Conversation with Bing, 8/25/2023. (Really? You talked to Bing about this? Oh, Lord. The bots are in cahoots.)
Bing’s AI Push:
Innovative Features Yet to Make a Mark
By ChatGPT for AI Insider
In a world where search engines are our go-to guides for almost everything, it’s no surprise that Microsoft invested billions to bring AI to its search engine Bing.
Despite garnering some initial buzz, Bing’s market share remains stuck at around 3%, according to analysts, leaving many to wonder if these new features are enough to challenge the search engine giant, Google. [Ed. note: Depending on who you listen to, Microsoft is either losing or winning the AI bet.]
Earlier this year, Microsoft introduced Bing AI Chat, a feature that allows users to have a conversation with the search engine.
These enhancements were expected to give Bing a significant boost, but the latest market statistics tell a different story. While Google continues to hold a commanding 90% share of the search engine market, Bing’s share has barely budged from 3 percent.
Microsoft remains optimistic, claiming that Bing has grown to exceed 100 million daily active users since the introduction of these features. They argue that the numbers from market analysts don’t capture the full picture, especially those who go directly to Bing’s chat page.
However, experts in the field have a more tempered view. Max Gomez, chief marketing officer of digital marketing firm The HOTH, believes that users will likely continue to gravitate towards Google for clear search intent. He foresees a trend where AI will be used for specific answers, especially in specialized tasks, but it may not be enough to shift the balance in Bing’s favor. [Ed. Note: I went from never using Bing to chatting with it almost daily because as I said in a previous post, I love its “personality.” I may not be the average user, but also I’m really impressed in particular with the Bing sidebar feature that summarizes a web page’s content while you have it open.]
Meanwhile, Google continues to develop its bot, Bard. Google Bard has seen a 50 percent increase in users in the past year. In May, Bard received over 142 million visits. This is a sharp rise from March when Bard got 30 million visitors to try it. Most Google Bard users are from the US (62.6%) and UK (8.29%).
Usage of ChatGPT is down since it has been temporarily disconnected from searching the web to be retooled to fix some access issues. No word from OpenAI, maker of ChatGPT, when that feature will return.
Ancient Windows Apps May Get AI
Admit it. Some of you still use Windows Paint. No shame. Here’s an update for you.
By Bing for AI insider
According to Windows Central, Microsoft is experimenting with different ways of adding AI capabilities to some of the in-box Windows 11 apps, such as Photos, the Snipping Tool, and Paint. These AI features could enable you to do things like cut out objects or people from photos, copy text from screenshots or images, and generate art based on your criteria.
Sounds amazing, right? But before you get too excited, there are some caveats. First of all, these AI integrations are still in experimental stages, and there is no guarantee that they will ever ship on Windows 11.
But if they do, imagine asking Paint to create a canvas based on the prompt: “A cute kitten playing with a ball of yarn.” You might expect something like this:
Or what if you want to use the Photos app to cut out your face from a selfie and paste it on another photo? Done.
Microsoft is expected to announce more details about its AI plans for Windows 11 at an event on September 21.
Artists + Google = A New Way to Visualize AI
[Ed. Note: See the “Aht Gallery” feature below for examples of this project.]
By ChatGPT for AI Insider
Ever wondered what Artificial Intelligence looks like? If you’ve ever Googled it, you’ve probably seen a bunch of glowing blue brains and white robots (even on this newsletter!). But guess what? AI is about way more.
Those stereotypical images of AI’s “brain” are not just boring; they’re also a bit misleading. They can make us think that AI is all about robots taking over the world or something out of a sci-fi movie. Plus, they don’t show how diverse AI can be!
🎨 Artists to the Rescue!
DeepMind, the brainy folks at Google, decided to shake things up. They’ve teamed up with artists from around the globe to create fresh, exciting, and more realistic images of what AI is and can be. The project is called “Visualising AI,” and it’s all about showing the world that AI can be as diverse and creative as we are!
🌍 Why Does This Matter?
Art isn’t just pretty to look at; it shapes how we see the world. By changing the way we visualize AI, we can start to imagine all the amazing things it can do for us—from helping combat climate change to making our lives easier and more fun.
📊 By the Numbers
Since its launch, “Visualising AI” has been a hit! The project has:
Commissioned 13 artists 🎨
Gained over 100 million views 👀
Had 800,000+ downloads 📥
Best part? All the artwork is free to download and use. (It’s embedded in this post.) So go ahead, make your next PowerPoint presentation a masterpiece with a representation of an AI’s brain!
Court: AI Art Can’t be Copyrighted
By Bing for AI Insider
Can AI create art? That’s the question that a US court had to answer in a recent case involving an AI-generated image. The verdict: no, AI cannot be an author of original work, and therefore, it cannot own a copyright.
The case was brought by Stephen Thaler, a physicist and inventor who created an algorithm called Creativity Machine. He used this algorithm to generate an image that he claimed was a work of art. He then tried to register it as a work-for-hire, with the Creativity Machine as the author and himself as the owner.
However, the US Copyright Office rejected his application, saying that only human beings can be authors of original work. Thaler sued the Office, arguing that his AI-generated image was a creative expression that deserved protection.
But the court disagreed. Judge Beryl A. Howell ruled that Thaler’s image lacked human authorship, which is a “bedrock requirement” of copyright. She cited previous cases where non-human entities, such as a monkey and a supernatural voice, were denied authorship.
She also acknowledged that AI is creating new challenges for the law, as more artists use AI as a tool to create new work. She said that this would raise questions about how much human input is necessary for a work to be original and protectable.
Thaler plans to appeal the case, while his lawyer and the Office have different views on the court’s interpretation of the law. The case is one of many that are testing the boundaries of AI and copyright.
What do you think about this case? Do you agree with the court’s decision? Let us know in the comments below!
AI Spy: U.S. Agency Using AI to Monitor Social Media and Detect Emotions
By Bing for AI Insider
U.S. Customs and Border Protection (CBP), has bought millions of dollars worth of software from a company called Fivecast, which uses AI to analyze online posts for “sentiment and emotion.” The software, called ONYX, can also detect “risk terms and phrases” in multiple languages, and recognize objects in images and videos, according to a report by 404 Media.
The report is based on a cache of documents obtained by 404 Media through Freedom of Information Act requests seeking info from CBP and other U.S. law enforcement agencies. The documents reveal that CBP uses ONYX to target individuals or groups, single posts, or events related to threats to public safety, national security, or lawful trade and travel. The software can collect data from big social platforms like Facebook and Reddit [Ed. Note: And presumably the right-wing Truth Social], as well as smaller communities like 4chan, 8kun, and Gab.
The news raises questions about the accuracy, utility, and ethics of using AI to monitor social media and detect emotions. AI systems are often prone to errors, biases, and manipulation, especially when dealing with complex and subjective human behaviors. Moreover, the use of AI may infringe on people’s privacy, civil rights, and civil liberties, critics say.
What The Actual Hell?
A new-ish feature showcasing AI stories that may make you go “hmmm.”
🤔 Searches of Google’s new generative search experience offered up the benefits of slavery and genocide, plus a recipe for cooking a poisonous mushroom, according to this Gizmodo report. Yep. Now, to be kinda fair, we’re talking about a beta product here. These applications are being tested specifically to uncover crazy stuff like this. And according to the article, they have already squashed some of the bad stuff, so that’s a win.
🤔 Scientists recorded a Pink Floyd song from patients’ brain waves.
🤔 An AI generated story recommended a food bank as a “don’t miss” eatery in Canada.
🤔 Baltimore County schools have added AI-powered gun detection software to 7,000 security cameras in the system. The cost? $1.4 million. Sadly, I guess this needs to happen, but, imagine what they could do with that money to stop kids from firing guns in school in the first place.
Random Shorts
UN Study: Women may be most at risk of losing jobs due to AI.
Hugging Face Gets More Hugs: The ridiculously named AI startup Hugging Face, raised $235,000 raising its valuation to $4.5 billion. Among the investors: Google, Amazon, Nvidia, and Salesforce. Just because I think their name is silly doesn’t mean they can’t excel.
Fun and Useful Stuff to Try
PhotoCaptions – Stumped about how to caption that vacation or work photo? The aptly named PhotoCaptions will do it for you. Upload a photo and this AI tool claims it will generate a caption and hashtags, too. Full disclosure: I have not tried this yet. There’s a waitlist. Sign up on the site.
PastorsAI: I’m not personally a church-goer, but if you are and your pastor’s sermons are posted to YouTube, you might want to try out a PastorsAI chatbot. Enter the Youtube URL of your church service, get back a custom chatbot that you can query about content in the sermons. Free for small churches.
Quillbot: This “AI Grammar and Writing Tool” is jam packed with helpful features to help with your writing. Paste in some text and Quillbot will generate a basic paraphrase for free. Upgrade to a premium plan and it will tailor for specialty use, such as academic, creative or formal. It will also check your grammar, summarize, translate languages and more. Free and paid plans. (They are currently running a back to school special: 40% off. Use discount code SCHOOL40.)
Aht Gallery
The item above titled “Artists + Google = A New Way to Visualize AI” links to a showcase of artistic interpretations of what AI looks like. (And hit the full screen for the best look at this!) Click below to see a few of the images. To see more, check out the full project.