China Releases OpenAI Rival

This is Synthetic Mind, the A.I newsletter thats like the friend you can always trust to scratch the spot on your back you just cant.. quite色 reach..

Its Wednesday. Lets get to it:

  • 劾 China Releases OpenAI Rival

  • A.I Robot Stars In Commerical

  • 蛛儭 A.I Detectors Threaten EU Plans

Read Time: 4 min 16 sec

劾 China Releases OpenAI Rival

Late last week Alibaba (a major Chinese company) released Chinas first major text-to-image tool named Tongyi Wanxiang.

(Wanxiang means tens of thousands of images)

As an A.I image generating addict.I had to check it out.

Heres a quick side-by-side for ya:

I used the same prompt given on Alibabas page for DALL-E and Midjourney

IMO, Tongyi Wanxiang generated a better overall image than DALL-E, but Midjourney takes the cake with more detail and the best lighting.

Either way, its a big step for China whos been trailing behind the U.S. in generative A.I (A.I that creates content like images, text, videos, etc).

Why is China so far behind? Alibaba and other Chinese tech giants are scared to create an all-encompassing A.I tool in fear of breaking Chinas strict censorship laws.

Simply put, the more content the A.I tool can create, the higher the chances of breaking censorship regulations. Something NO Chinese company wants to gamble with.

But this puts China in a tough spot

China wants to keep A.I content heavily monitored, but in doing so they are significantly slowing their A.I innovation - keeping them behind the U.S.

Right now the EU is struggling with the same problem as they try to balance A.I regulation and encouraging A.I innovation.

If you ask me, Id say this is the #1 reason the U.S. has not enforced any strict A.I regulation.

Get an Edge on the A.I

To get an edge in A.I careers of tomorrow, learn the concepts that drive it today. Brilliant makes it easy.

In just 15 minutes a day, you can level up in the core building blocks of A.I, as well as data science, programming, logic, and beyond.

Thousands of bite-size, interactive lessons help you master the key concepts and make it insanely easy to build a daily learning habit. Join over 10 million people and start your 30-day free trial today.

Plus, Synthetic Mind readers get 20% off an annual premium membership.

A.I Robot Stars In Commerical

With all the new A.I updates going on, weve missed some big events in the robotics world

Most of you have heard of Ameca - the closest robot ever to match human behavior.

Ameca can:

  • Talk with humans and other robots in many different languages

  • Make facial expressions based on your conversation

  • Give public speeches

In the last few days, Engineered Arts (Amecas creator) has combined Amecas abilities with an A.I image generator - Stability A.I.

Now, you can ask Ameca to draw a picture and the robot will draw it + even sign its name.

Most robotic companies like Figure build their humanoid robots for physical labor or even as security guards. Yet, Engineered Arts builds theirs for entertainment purposes only.

Youve probably seen their bots in a Jose Cuervo commercial called Date More Human

If not, you can watch the full commercial HERE

Ameca is yet another example of the growing presence of A.I and Robots in media and politics. Weve already seen:

  • Robots speak at an international conference (more on this tomorrow!)

  • Celebs contracting out their A.I clones for commercials and movies

  • Musicians selling their voices for A.I music

  • ChatGPT write and direct its first movie, The Safezone.

PS - If you want your own drawing by Ameca, contact [email protected]

蛛儭 A.I Detectors Threaten EU Plans

A new study by Standford University showed that A.I detection systems are wrong over 50% of the time.

And that my friends, is a BIG problem. Why?

A.I detection is how we know if something is written or made by A.I. We use it to catch things like:

  • Scams/ identity fraud. A new trend popping up where scammers clone childrens voices and call their parents to ask for ransom money. On average, families lose $11,000 in each fake-kidnapping scam in the U.S.

  • To label deepfakes. Imagine if someone used A.I to create a video of you stealing something. Youd want to be able to prove that fake.

  • Students cheating with ChatGPT on their homework. In the same study, a researcher ran 91 human-written essays through 7 A.I detectors. A.I incorrectly flagged 98% of essays as ChatGPT generated.

Yikes.

This will start to make waves in the next few months as the EU tries to pass its Digital Markets Act - an act requiring social media platforms to label all A.I generated content.

But now that platforms cant determine what is A.I generated and whats not the EU will have to halt their plans.

Soon well all have to decide who is responsible for auditing A.I generated content. The government? Social media platforms? The A.I companies?

Decisions. Decisions.

Readers Spotlight

We sure do talk a lot in these newsletters, but now we want to hear about YOU.

If you have a cool A.I story, tool, business, or you just love A.I as much as we do.

(seriously ANYTHING goes)

Were all ears.

If we love it, well put you in our newsletter in this very spot.

Its a crazy world in the wild west of A.I, and getting to know a friend or two goes a long way.

Reply to this email with your story!

Join the conversation

or to participate.