Welcome to our roundup of this week’s artisanal hand-typed AI news.
This week AI brought back a Beatle.
Deepfakes made vegans eat meat.
And China wants to build a gazillion humanoid robots.
Let’s dig in.
Pass the snacks
Are your gaming teammates complaining about the noise your Doritos addiction makes? Pepsico has used AI to silence the crunch so gamers can snack quietly.
A steak sandwich company thought they’d have some fun messing with vegans. They gaslit them with video evidence that showed them eating meat and then questioned their vegan credentials. Steak-umm said this was all for a good cause.
AI is tricky
It’s hard to believe but in many places, it’s not illegal to create a fake nude of someone and then pass it around. AI-generated fake nudes made things a little awkward in a New Jersey school.
Think you could tell a real face from an AI-generated one? You’ll be surprised at just how easily humans are tricked by AI-generated faces according to this study.
Seeing as we’re so easily fooled by AI it’s probably a good thing that Meta has updated policies to require labeling of AI-generated ads.
AI safety – Should we chill or panic?
After a flurry of executive orders and summits, there’s been self-congratulatory backslapping among the people trying to regulate AI for safety. This summary of the state of affairs explains what (little) has been achieved so far and where we’re headed.
While regulators work hard to avert the imminent extinction of the human race at the hands of AI, some experts are wondering what the fuss is all about. The ‘AI godfathers’ are having a very entertaining back-and-forth on X about whether the risks of AI are real or imagined.
The ongoing conflict in the Middle East shows that humans are accelerating our demise just fine without the help of AI. The UN hired a marketing intelligence company to create a kind of SIM City AI model to help with the Israeli-Palestinian war.
— Elon Musk (@elonmusk) November 2, 2023
In the lead-up to Elon Musk’s xAI releasing its first products to a select group, we took a closer look at what to expect and what xAI is actually trying to achieve.
Some X Premium subscribers got early access to xAI’s chatbot ‘Grok’ and it seems like it was trained on Musk’s peculiar sense of humor. Grok’s responses can also be pretty spicy.
Want to know how Grok compares with ChatGPT? These stats make for interesting reading but we’re guessing xAI is working in something a lot more powerful. Let’s hope they consider a name change.
AI chatbots are really good at lying and some people found that out the hard way. Australian academics had to apologize for submitting false AI-generated claims to parliament.
Some people say that crime doesn’t pay, but GPT-4 may disagree. The model was quite happy to perform illegal insider trades and then lie to cover it up.
If you were hoping to get in on the AI cheating action you may want to reconsider. A new ‘ChatGPT detector’ can spot AI-written academic papers.
All you need is AI
If your grandparents were fans of the Beatles when they were still touring you may have had some tricky tech conversations this week. ‘What do you mean the AI brought John Lennon back from the dead, dear?’ This is such a cool application of AI and we want more.
Scarlett Johansson is less enthralled with AI. She’s suing a company that used her likeness to market its image generation app. ‘Hey, we should use an AI fake of Scarlett in our ad. I’m sure she’ll be cool with that.’
Johannson will have plenty of time to focus on her lawsuit because she won’t be acting any time soon. The SAG-AFTRA actors’ labor union still hasn’t come to an agreement with Hollywood studios over AI concerns so the strike continues.
ChatGPT got a big boost
Some AI companies like to use GPT-4 as a benchmark saying, ‘Our model is almost as good as GPT-4.’ This week, OpenAI moved the goalposts further out. At its OpenAI DevDay the company unveiled GPT-4 Turbo with knowledge up to April 2023.
Sam Altman also announced that ChatGPT now lets users create custom agents called GPTs. The level of integration with other apps makes this a huge deal.
From Altman’s DevDay closing remarks it looks like we can expect some more exciting announcements soon.
Sam Altman: “We hope that you’ll come back next year. What we launch today is going to look very quaint relative to what we’re busy creating for you now.” pic.twitter.com/nUAQH1De5i
— Smoke-away (@SmokeAwayyy) November 6, 2023
Who ordered the army of robots?
China wants to start mass-producing humanoid robots. You’d think this was part of a 10-year plan but, in true Chinese style, they say the robots will roll off the production line in 2025.
The US tech export restrictions may just have been a speed bump rather than a barrier to China’s AI ambitions. The country is becoming less dependent on American tech. This week we saw more details of a new AI chip from Huawei that could rival Nvidia’s A100.
The UK is also trying to become more AI independent as it invests $273m to build its most powerful AI supercomputer. It’s going to be a beast.
Microsoft in the money
Want to try out Office 365 Copilot? You’ll need $9,000 a month to cover the minimum 300 licenses Microsoft wants you to commit to.
It may be worth it though. A Microsoft study showed companies get a surprising return for every $1 they spend on AI.
And in other news…
There were a lot of other cool things happening in AI this week. Here are some of our favorites.
- Elon Musk’s brain-computer interface got the go-ahead for human trials and they’re looking for volunteers.
- An unemployed programmer used AI to apply to 5,000 jobs and got 20 interviews.
- Instagram is working on giving you a customizable AI friend.
- This Chinese analog chip could be almost 4 times faster than an Nvidia A100.
- If you’re unconvinced that open source AI is a good idea then this long read is worth checking out.
And that’s a wrap.
Will you be trying out Dorito’s Silent during your next gaming session? What do you think about Grok’s sense of humor? I’m not sure I’m a fan.
When you realize that the guy behind Grok’s humor also wants to insert a chip into your brain it makes me hesitant to hit the “Apply to be a test subject” button.
Let us know what you thought of this week’s crop of AI news and give us a heads-up if we missed a juicy story.