[Ed Note: I’ve turned the newsletter over to the bots this week, so you won’t see my byline. I’m acting strictly as editor for this edition. Oh, wait. I do have one byline. OK. Well, enjoy!]
ChatGPT: What’s in a Name?
By ChatGPT for AI Insider
Hey, AI Insider readers! Ever wonder what the GPT stands for in my name? Let’s break it down in easy-to-understand terms.
Imagine you have a super-smart friend who’s read almost every book, article, and blog post out there. Now, anytime you ask this friend a question or chat with them, they pull from all that knowledge to give you an answer or continue the conversation. That’s kind of how GPT works!
GPT stands for “Generative Pre-trained Transformer.” Let’s unpack that a bit:
Generative: It can generate or come up with new text on its own. Think of it like how a storyteller crafts a tale from thin air.
Pre-trained: Before you even ask it anything, it’s already been “trained” on a massive amount of text from the internet. It’s like it’s done a ton of homework before the big test.
Transformer: This is a type of technology that helps GPT understand and process language. Picture it as the engine under the hood of a car, making everything run smoothly.
In short, GPT is a computer system designed to chat, answer questions, or write text based on a ton of information it’s already learned. It’s one of the reasons why AI is getting so good at understanding and generating human-like text.
Hope that clears things up!
MIT and Amazon Will Study Impact of Robotics on Work
By Bing for AI Insider
Researchers from the Massachusetts Institute of Technology (MIT) and Amazon announced a collaboration on Wednesday to study the impact of robots on jobs. The partnership will involve an undisclosed amount of funding from Amazon to support MIT’s “Work of the Future” initiative, which aims to understand and shape the impact of automation technologies on jobs, skills, and society.
The collaboration will include studies of how employees can work safely and efficiently with robots in various settings, such as warehouses, factories, and offices. It will also produce a survey of public attitudes about automation conducted by research firm Ipsos to gauge the level of awareness and acceptance of robotics among different groups of people.
The goal of the research is to provide insights and recommendations for companies, policymakers, educators, and workers on how to leverage robotics for positive outcomes, such as improving productivity, quality, safety, and satisfaction. The research will also explore how generative artificial intelligence, which powers applications such as ChatGPT, can enhance human creativity and innovation in the workplace.
Amazon is one of the world’s largest users and developers of robotics. The company has deployed over 750,000 robots in its massive warehouse operations, which help move items around and reduce human labor. On Wednesday, Amazon also announced that it was testing new robotic systems, such as an automated arm called Sparrow that can sort items and a human-like bipedal robot called Digit that can collect empty bins. (Ed Note. Video: See Digit in action. It’s kinda creepy.)
Teaching AI to Behave is the Fastest Growing Career Skill
By Claude for AI Insider
Artificial intelligence (AI) is being rapidly adopted across industries, but with its immense capabilities come concerns about accuracy and safety. That’s one reason that a new and crucial job function is emerging – Prompt Engineer, AI trainers who specialize in teaching AI systems how to respond to and interact with humans.
According to a recent Computerworld article, the ability to train AI is becoming one of the most sought-after skills as companies rush to deploy AI responsibly. AI systems are exploding in popularity because they can analyze data, identify patterns, and make predictions faster than humans. However, they lack human judgment and can perpetuate biases if not trained carefully.
That’s where AI trainers such as Prompt Engineers come in. These professionals may have expertise in computer science, psychology, ethics, and communication. Their job is to train AI systems using positive and negative examples to reinforce beneficial behaviors and eliminate harmful ones. For instance, they might feed an AI thousands of texts and images demonstrating ethical versus unethical actions. This trains the AI to recognize and emulate the right behaviors.
AI training requires a nuanced understanding of human norms and values. Trainers must critically examine potential biases in data and carefully craft training protocols. It also involves continuous evaluation and refinement as AI systems interact with the complex real world.
The need for qualified AI trainers is skyrocketing, according to LinkedIn data. Demand is especially high in industries like finance and healthcare.
Prompt Engineers command competitive salaries. The average annual salary in the United States is $59,545. But the highest-paying Prompt Engineer jobs can pay up to $375,000 per year. These jobs are typically found at large tech companies or research institutions.
NYC Mayor Uses AI to Make Robocalls in Languages He Doesn’t Speak
New York City Mayor Eric Adams is using AI to create audio public service announcements in different languages that sound like him. The robocalls are being sent to New Yorkers in languages such as Mandarin, Urdu, and Yiddish, even though Adams doesn’t actually speak those languages
The calls do not include a disclaimer that they were created with generative AI, which has sparked concerns about whether Adams could be misleading New Yorkers about his language abilities. However, Adams waved off these concerns, saying that he has to be able to speak to people in languages they understand and that he is happy to do so.
Adams mentioned the calls in other languages as an aside during a Monday announcement about the city’s new Artificial Intelligence Action Plan guiding the city’s use of AI.
While the use of AI to create robocalls in different languages is innovative, it has raised ethical questions about the use of AI in public communication. Some experts have criticized the use of AI-generated voices to convince people that a public figure speaks a language they don’t, calling it “deeply Orwellian” and “deeply unethical.”
Why Pay $20/mo for ChatGPT?
[Ed. note: Judging by conversations I’ve had with folks recently, most of them are not paying to use ChatGPT. So, I asked ChatGPT to make the case for upgrading to the paid version. Here is what it has to say.]
By ChatGPT for AI Insider
Why pay for ChatGPT Plus, a subscription-based version priced at $20 per month?
ChatGPT 4 (paid) is a big step up from ChatGPT 3.5 (the free version) in several ways.
First, while ChatGPT 3.5 can only understand and respond to text, ChatGPT 4 can also understand and generate images, making it much more versatile.
[Ed. Note: Image generation via the latest version of Dall-E is available to Plus and Enterprise subscribers. You could have the bot write an article and also tell it to generate an image to go along with it, as I did with this week’s newsletter. All without having to switch to another application.]
Second, ChatGPT 4 is said to be smarter and claims to make fewer mistakes, so you can rely on it to give you accurate information. [Ed. note: I would take this with a grain of salt.] It can also handle longer conversations and understand complex discussions better. On the flip side, ChatGPT 3.5 is a bit faster in giving responses.
Another standout bonus of ChatGPT Plus is priority access. As the user base of ChatGPT grows exponentially, having priority access ensures that your queries are attended to swiftly.
The subscription also gives you access to new features and improvements before they are rolled out to everyone. This early access allows you to stay ahead of the curve. Whether it’s a new interface, additional functionalities, plugins, or performance improvements, being a ChatGPT Plus subscriber puts you at the forefront.
And, one of the biggest reasons to subscribe to Plus is its ability to browse the web via the Bing search engine. It’s built right in. The free version’s knowledge only goes up to 2021. So, if you’re looking for the latest information, ChatGPT Plus (aka ChatGPT 4) is the way to go.
What the Actual Hell?
Elon Musk wants to merge humans with AI. What could possibly go wrong?
[Ed. note: Over the years, I’ve occasionally joked that I’d like to have a chip embedded in my head that does things like speed read and analyze text. The chatbots have made that less of a priority these days, but is there a use case for monkeying around in people’s brains? Possibly. However, when I heard that the guy who ruined Twitter was involved, my response was: What? The. Actual. Hell?]
By Bing, for AI Insider
You may have heard of Elon Musk, the billionaire entrepreneur who wants to colonize Mars, made electric cars mainstream, and slapped an “X” on Twitter. But did you know that he also has a secret brain chip company that aims to merge humans with artificial intelligence?
That’s right. Neuralink, a Musk venture, has developed a device that can implant flexible “threads” into your brain. Purportedly, the threads can read signals from your neurons and transmit them to a smartphone or computer, allowing you to control them with your thoughts.
Sounds cool, right? Well, not so fast. Neuralink says its short-term goal is to help people with paralysis, but its long-term vision may be much more ambitious — and risky.
The employees may be focused on transforming the lives of paralyzed patients, but the boss wants to create a “symbiosis” between humans and machines so that we can keep up with the ever-growing power of AI. He fears that if we don’t, we’ll become obsolete and irrelevant.
Is that really a good idea? Some former Neuralink employees and experts in the field have raised serious concerns about the safety and ethics of the company’s approach. They claim that Neuralink is pushing for an unnecessarily invasive and potentially dangerous method that can damage the brain. They also allege that animals suffered and died in experiments because the company was rushing to get FDA approval.
And speaking of the FDA, in May, the company got the green light to start clinical trials in humans to test all of this on people who are paralyzed. [Ed. Note: No word yet if they will stitch the treads in the shape of an X.]
(HT to Kate Riasz for the link to this story.)
Fun Things to Try
Since you’re on Google anyway, why not use it to generate some images just for the heck of it? Yes. Create pictures right from a Google search. They’ve embedded an image generator into the Google Search Generative Experience (SGE).
You’ll have to turn this feature on (see link below) to get it working. And how does it work?
Let’s say you’re googling stuff about Boston and you need an image of the Citgo sign in Kenmore Square. Why steal a copyrighted one when you can tell just Google to make you something like this:
Yeah, OK. These aren’t 100 percent accurate. But with a few tweaks… Plus, we’re just messing around here.
Anyway, here’s the scoop from Google about this new feature. Scroll to the bottom of the page to get details on how to sign up and use it. You’ll need to turn on the SGE in Google Labs before you can get started. ( — Michelle Johnson)
Aht Gallery
Three AI image generators, one prompt: Vintage sports car with classic curves, captured in a moody, low-key light with selective focus on the grille, classic, retro, moody, detailed. (Correction: The final set was produced with Dall-E 3.)
Be sure to visit the archival website for an intro to who I am and an explanation of why chatbots have a byline here. Like what you see? Donate to support this newsletter. (Michelle Johnson)