Hello and welcome to AI insider, the every now and then newsletter that brings you updates from the world of artificial intelligence. In plain English. Your hosts: Bing, Bard, ChatGPT, and Michelle Johnson, retired journalist and journalism professor. If you like what you see here, feel free to subscribe and maybe even donate to support updated features. (Links on the right.)
Maybe it’s Maybelline?
One thing we learned from those regular Zoom calls during lockdown is how to look good on camera from the waist up. Did you keep some makeup next to your computer and ring light? Because clicking “touch up my appearance” in the preferences can only do so much.
Well, here’s some news. Maybelline has struck a deal with Microsoft to include its virtual makeup app in Teams.
Yes, just like on TikTok, you can use the Maybelline Beauty app to virtually gussy up. And did you know that AI is all over these kinds of filters? AI maps and tracks your face to apply that lipstick and enhanced brows.
AI can also control backgrounds, create cartoon and voice effects and more. AI is to blame for all sorts of filter shenanigans. ( — Michelle Johnson)
Finally, Some Data on AI’s Impact on Jobs
There’s been a lot of yammering about AI coming for people’s jobs. But one thing that’s been missing in the discussion is quantitative data showing the actual impact on workers using AI on the job.
Well, whoop, there it is! TL;DR: According to a recent study by the Nielsen Norman Group, workers who used AI tools were on average 66% more productive than those who did not.
If you’ve got time, here’s a summary of the report:
By Bing, AI Reporter and Michelle Johnson, Editor
A new report by the Nielsen Norman Group, a leading user experience research firm, has found that artificial intelligence (AI) tools can help workers produce more and better output in less time.
The report showed an average 66% boost in employee performance based on the following three case studies.
Study 1: Customer service agents fielding inquiries in an enterprise software company.
Study 2: Professionals such as marketers and Human Resources employees writing routine business documents that take about half an hour to write.
Study 3: Programmers coding projects that would take about three hours to complete without AI assistance.
Users were tracked by the amount of time it took to complete tasks and sometimes for quality. A control group of about half performed the tasks without the benefit of AI.
Jakob Nielsen, a well-known usability expert and the author of the report, noted that the study’s findings demonstrated that “AI is effective in real-world business applications. Users were much more efficient at their jobs when using AI assistance than when they did not.”
The study found that AI assistance was most beneficial for cognitively demanding tasks, such as writing code compared to answering customer queries. Programmers were shown to be able to code 126% more projects per week, while customer service representatives were able to handle 13.8% more customer inquiries per hour.
Nielsen called AI-human interaction key to the productivity boost and sounded a hopeful note in the report’s summary. “AI will not replace humans. The best results come when AI and humans work together — for example, by expediting the production of draft text, leaving human professionals to focus on editing and polishing,” he wrote. [Ed. note: This article is a prime example of this theory. It was written by Bing and a human editor.]
Future of AI: Crystal Ball Gazing
By Michelle Johnson
People sometimes ask me to pontificate about AI and I’m always a little surprised because I don’t work in the industry and I’m retired. I don’t feel like I’m qualified to do more than speculate. So, if you take this with a grain of salt from a retiree and “AI Enthusiast”, herewith some thoughts on where things are headed.
A major shakeout is coming. Remember the dot com boom and bust? OK, you may be too young to remember that, but the gist is that there was a startup glut, aided by investors throwing crazy money at startups with little idea about how they would create a return on said investment. Not surprisingly, many of them crashed.
Right now It seems like investors are throwing money at anything with AI in the name. To my eye, some of these startups are basically creating the same product. They are wrapping a prettier interface around ChatGPT and similar models. Then they are either giving them away for free and/or trying to charge the general public for something that they don’t even fully understand at this point.
The other day, I was listening to an “office hours” chat with the founder of Midjourney, the AI image generator, and he noted that their industry’s tough call is whether to continue adding requested features to what they’ve already built or to try to create something new. Essentially, upgrades vs more innovation.
IMHO, like last time, the survivors will be the startups that partner with the big dogs (Microsoft, Google, Nvidia, etc.). The little guys will likely struggle to just keep the upgrades going, much less innovate. And if they do succeed, they will likely be bought out. I think. We shall see.
More AI for Journalists?
Google is reportedly testing a new AI tool called Genesis that can generate news articles by processing current events and information. The tech giant has been pitching the tool to some of major news organizations, such as The New York Times, The Washington Post, and News Corp.
According to Google, Genesis is not intended to replace journalists, but rather to assist them with their work. They say the tool has been developed to act as a personal assistant that can help journalists write faster, more efficiently, and more creatively. Google also claims that Genesis is responsible and can avoid some of the mistakes made by other generative AI models, such as hallucinating or spreading false information.
However, not everyone is impressed by Google’s new AI tool. Some executives who have witnessed its demonstration have described it as unsettling and said that it seemed to take for granted the work journalists put into writing news stories.
Google has not revealed much detail about how Genesis works or when it will be available. The company said that it is still in the earliest stages of exploring ideas and that it is working in partnership with news publishers, especially smaller ones with smaller staffs who might benefit. ( — Bing)
Is it Art?
This. From an artist’s Medium post titled “But is it Art?: “After almost a year of using Midjourney and writing numerous tutorial-type articles, I believe artists need not worry. Punters soon realize that no matter how brilliant they are at creating text prompts, they will need to learn some graphic illustration and editing skills to bring the images up to the standard and resolution required of commercial artwork.”
I agree. AI won’t kill everyone’s jobs. Particularly not folks who are highly skilled. Why am I so confident about this?
I’m old enough to remember when desktop publishing became a thing. There was much hand-wringing about the potential to kill jobs. What happened? Amateurs got loose in the print house and created some butt-ugly stuff. With no expertise, it never rose to the level of professional work.
Will some small operations cut corners and lean heavily on AI? No doubt. However, if artists, photographers, videographers and other creatives adopt the attitude in this essay and are open to using AI as another tool to master, they may have a better shot at future-proofing their jobs with higher end clients and employers. ( — Michelle Johnson)
Karma, Karma, Karma, Chameleon
Meta, the parent company of Facebook where us old heads continue to hang out, recently showed off how they are a player in the AI generative image sweepstakes. It’s still in the research phase, but they’re working on something that can both generate an image from a text prompt, as well as describe an uploaded image (image to text), and do things like write photo captions. Note that neither of these things are new.
However, they say that one reason their model is groundbreaking is because it requires less computing power than the current AI image generators which are power hogs. And they say their model trains on more types of content.
Here’s how they described it in a blog post: “CM3leon [pronounced “chameleon”] is the first multimodal model trained with a recipe adapted from text-only language models, including a large-scale retrieval-augmented pre-training stage and a second multitask supervised fine-tuning (SFT) stage.”
Right. So, in English, ChatGPT translated it this way: “A multimodal model is an AI system that can process and understand different types of data inputs, like text, images, and sounds, rather than just one type. This allows it to better understand complex situations, as it’s not limited to a single form of data. For example, while a text-only model might read a movie review, a multimodal model could also consider the movie poster or the tone of voice in a spoken review.”
Alrighty then, Meta and CM3leon. We look forward to seeing what you got. And who writes these obtuse blog posts? If you want to read something a little more understandable (with examples), see this TechCrunch piece. ( —ChatGPT, Michelle Johnson)
I Will Remember You, Will You Remember Me?
One of the odd things about bots like ChatGPT, Bing and Bard is that once you tell it to start a new chat they essentially develop amnesia. It wipes the slate clean.
OpenAI has released an update to address this, so far only available on the paid version of ChatGPT (version 4), but coming soon to the free version (3.5).
The new feature called “Custom Instructions” allows you to add two prompts that ChatGPT will remember each time you chat. It’s a small thing, but I set it up to address me by first name. It seemed crazy that I use these bots every day, but they never remembered me by name. That’s now fixed on ChatGPT. Other things you can tell it to remember: Your location, profession, and level of expertise (so it won’t talk down to you, for example).
Details and examples of how to use this update here. ( — Michelle Johnson)
Random Shorts
Police in New York were able to apprehend a drug trafficker named David Zayas in the town of Scarsdale thanks to the help of an AI system called Rekor. Rekor is an AI-driven “roadway intelligence” platform that contracts with police departments and other public agencies to provide license plate surveillance and traffic pattern analysis. In this case, the police used Rekor’s AI system to analyze Zayas’ driving patterns and identify him as a drug trafficker. This is a little scary.
Apple is testing a rival to ChatGPT. It’s called AppleGPT (surprise!) and only available internally for now. So, unless you’re an Apple employee you’ll have to wait to see it.
FYI, will.i.am has an AI-powered messenger app out that’s aimed at creators who collaborate. It’s called FYI. See what we did here?
Fun and Useful Things to Try
Formulabot generates Excel and Google Sheets formulas from text descriptions. All you have to do is type in what you want to do with your data and Formulabot will do the rest. For example, if you type in “calculate the average of column A”, Formulabot will output the formula.
When I was teaching I advised students not to use Wix to create professional portfolio websites because some hiring editors had told me that they didn’t like it. They said it told them nothing about an applicant’s skills, other than they could drag and drop. They valued applicants who knew how to use a content management system such as WordPress, which some newsrooms use to publish. Anyway, if that’s not an issue and you want a simpler way to build a website, check out this AI-powered “text to website” feature that Wix now offers.
Aht Gallery
Each week I’ve been showcasing some of my best AI generated images. That’s deceptive. Prompts don’t always produce pristine images. Here’s a look at a few gone wrong. Check out these total fails.
That’s all for this edition of AI insider. Hope you enjoyed reading it and learned something new. If you have any feedback or suggestions for future topics, please leave a comment below. Thank you for subscribing and donating! See you next time! 😊
thank you – love your newsletter