AI prompt upgrade: Imagine you want to market your company or pet project. Ask ChatGPT (or your favorite bot) to generate a blog or social media post to share. You’ll get a better answer if you simply preface it with, “You’re a marketing expert.“
Don’t trust AI for these 4 things
I used AI recently to write a birthday note, plan a trip to Hong Kong and program my robot. But this tech is still in the infant stages and it has some major blind spots you need to know about.
Ask things like “How many basketballs would you have to stack up to reach the moon?” all day. (It’s 1,595,524,852 by the way.) But when it comes to these four Ps, you’re better off skipping the chatbot.
1. Protection
If you own a small business it’s tempting to save money by skipping the lawyer fees and drafting contracts via ChatGPT or any other chatbot. In almost all cases, the legal risk isn’t worth it.
If you ask the wrong question or use the wrong legal jargon in your prompt, you might get an answer that’s totally off-base or out-of-date with current updates.
Lawyers are experts at navigating legal “gray area” and the specific legal language that applies to your personal situation. They’re also accountable for getting it right (i.e. you can sue them if they mess up).
When it’s OK: Need help reviewing a legal doc or getting your feet under you with how existing laws and regulations work? Sure, chatbot away.
2. Precision
Don’t trust AI to do your taxes. Even the AI assistants built into platforms like H&R Block and TurboTax are wrong more than 30% percent of the time (paywall link).
Tax laws for things like deductions and retirement contributions change every year. Being off by even few hundred dollars could cost you dearly down the road.
When it’s OK: Don’t know what you don’t know? Use AI to figure out what questions to ask a tax pro (or how to find a great CPA).
3. Prognosis
AI can now tell if your political leanings are conservative or liberal using just your face. Algorithm VGGFace2 can suss you out with over 70% accuracy. Apparently, liberals have smaller lower faces, while conservatives’ faces are typically larger and wider. And independent voters? Theirs go straight down the middle. OK, I made that one up.
$9 an hour is the pay for AI nurses
Over 40 health care companies are already using NVIDIA and Hippocratic AI nurse bots to perform risk assessments, remote patient monitoring, new medication onboarding and more. The median pay for human nurses in the U.S. is $39 an hour.
It’s happening: The Senate passed the bill forcing TikTok’s Communist China parent company, ByteDance, to sell or face a ban. And while President Biden has signed off on it, there’s a catch: The ban isn’t rolling out until after the 2024 election. The legislation gives ByteDance nine months to sell with a possible three-month extension. Bet they’ll go down to the wire.
In northern Virginia (paywall link), internet data and AI centers are using so much electricity old coal plants are being fired back up. Yup, those very same plants we closed because of air pollutants, such as sulfur dioxide, nitrogen oxide and particulate matter. Cue a $5.2 billion project for new power lines stretching across state lines.
Rappin’ Lisa: Microsoft’s new AI tool, VASA-1, animates still images into videos loaded with facial expressions, head motions and lip movements that match a speech or song. Here’s a video showing the Mona Lisa rapping. Microsoft is delaying the release of this tool, though, because it makes creating deepfakes too easy.
Don’t believe everything you see on Netflix: The latest doc, “What Jennifer Did,” uses AI-generated images to depict killer Jennifer Pan as a fun, bubbly girl. In reality, she paid hitmen $10,000 to murder her parents in 2010. With no AI disclosure during the film, it’s clear the documentary isn’t about historical accuracy — just bringing in more viewers.
Try AI in Google Chrome: Turn it on by going to Chrome > Settings > Experimental AI. Now, you’ll get writing help and tab organization. Nice.
Taylor Swift’s new album, “The Tortured Poets Department,” was leaked to Google Drive two days early. Swifties were quick to call it an AI deepfake (paywall link). Surprise, the leak was real. Swifties did get a consolation prize: 15 additional songs. Share this with your favorite Swiftie.
2.3 million views on a fake James Bond movie trailer
Now folks are begging for a real version. The phony “Bond 26” vid “stars” AI versions of Henry Cavill and Margot Robbie. The creator used a mix of AI and existing film clips to create the trailer. Fun fact: Remixing film clips generally falls under fair use; otherwise, this trailer would be banned, James banned.
It started off with a gift: Kids are renting out their spare computer power to AI companies. Not a hard sell, when they’re offering Fortnite skins and Roblox gift cards. Here’s what your kid won’t realize: Their machine is being used to create AI porn, and they’re automatically opted in. Make sure none of the computers in your home are signed up to sites like Salad.
Ask Me(ta) anything: Meta just added its AI chatbot to the search bar on Facebook, Messenger, Instagram and WhatsApp. Ask it anything you’d ask a search engine. Google Search is so dead. Access Meta’s chatbot without logging in at meta.ai.
🎬 AI love story? TV manufacturer TCL is cooking up the “first-ever fully AI-generated rom-com” called “Next Stop Paris.” The trailer is, um, interesting. Despite the AI claim, they’re using human writers, actors and animators. Sounds like someone really wants investor money.
🤖 Work-from-clone: Influencers and business coaches are making “digital clones” to go to Zoom meetings for them. AI clone platforms use clips from podcasts, voice memos and Slack messages to mimic your responses and speech patterns in email or video. Deepak Chopra made one he calls “Digital Deepak.” Spoiler: It’s not very convincing, yet.
Love making AI images in Midjourney? This one’s for you. Use the Niji mode to create some seriously cool Japanese-style animation. On the web version of Midjourney, click Settings and then select Niji V6.Pro tip: Include “anime” or “manga” in your prompt.
20M student papers included AI writing last year
That’s about 10% of the 200 million assignments analyzed. Oh, and about 6 million of those submissions were at least 80% AI-generated. Might be time to bring back the blue book.
Meet your new CAIO: LinkedIn says the number of companies with a designated head of AI has tripled globally (paywall link) in the last five years. The Chief Artificial Intelligence Officer oversees AI’s role in order to improve workforce efficiency (aka fire people), identify new revenue (sell more things) and mitigate ethical and security risks (CYA). You know it pays really well, too.
Level up: OpenAI just launched GPT-4 Turbo, making ChatGPT sharper in writing, math, logical reasoning and coding. It’s now trained on public information up to December 2023. It’s for paying customers only; ChatGPT Plus, Team and Enterprise subscribers will get this upgrade.
📸 Don’t buy photo-editing software now: Starting May 15, Google Photos will open access to everyone for its free AI editing tools, including Magic Eraser, Photo Unblur, Portrait Light and Magic Editor. Android phones need to be running Android 8 or later to use these features, and iPhones must have iOS 15 or later.