Anti-universes, Microsoft's gone quantum, and how to get rid of all the deficits...
TNW View in browser →

It's time for robots to start paying taxes

Howdy folks,

I'm here today to announce my candidacy for supreme ruler of the Andromeda galaxy.

As far as I know, Andromeda isn't currently holding elections. But I figured if I was on record as being the first to throw my hat in the ring, I'd have a pretty good shot at capturing the early vote.

In the meantime, I've been working on my political platform just in case you Earthlings want to co-opt it. I call it: "The stupid AI tax."

Here's how it works:

Nobody has to pay taxes... unless they're involved in the creation of any artificial intelligence system purported to perform at or exceed "human-level" capabilities.

That's it. That's the whole plan. And the best part is that it'll rely 100% on self-reporting.

Every time a corporation, startup, research paper, marketing team, PR agency, or rando with a Github account claims they, their company, or their client has built an AI system that's as good as a human, everyone involved gets a letter from the IRS (or your country's applicable equivalent).

Maybe this is a brilliant political strategy designed from inception to limit disingenuous marketing hyperbole in hopes of normalizing the technology sector, with the ultimate goal of de-stratifying access to the future.

And, maybe it's just a completely self-serving plot to stop the never-ending glut of hackneyed pitches I get.

Either way, the future is now and my inbox is overflowing. Let's make it happen.

#NeuralforSupremeLeader

And now for something completely unrelated: Physicists suggest there’s an ‘anti-universe’ behind ours 

Tales from the arXiv archives

Consistency is key

Today's Tale from the arXiv Archives is brought to you by the letter C

As in:

Common sense

Certainty factor

and

Consistency

A team of researchers from the Google Brain team recently published a pre-print paper titled "Self-Consistency Improves Chain of Thought Reasoning in Language Models." 

Simply put, the team's work demonstrates that generating multiple outputs and then parsing them for consistency produces stronger results than just spitting out a single answer.

That sounds like a no-brainer, but the Google Brain team managed to train a neural network to be self-consistent in an unsupervised scheme. 

In other words: Google taught an AI how to achieve a high certainty factor through internal consistency. That sounds a lot like common sense to me. 

Wow!

You can check out the whole paper here on arXiv.

What we’re reading

💡

Brain Implant Allows Fully Paralyzed Patient to Communicate (HeroMag)

⚛️

Researchers store a quantum bit for a record-breaking 20 milliseconds (Phys.Org)

🤖

Andrew Ng predicts the next 10 years in AI (Venture Beat)

Something profound from the internet

Lol. This is pretty much what my whole Twitter feed looks like. 
boredape
 

 

Our favorite AI video of the week

Don't try to BS me. I know Zuul the Gatekeeper when I see it. (Click to watch on Twitter)
 
zuul

 

Well, bye

A few months back we rescued a dog named Lucy who'd just recovered from a massive trauma after an accident (we think she was hit by a car). 

We'd just gotten Bella, the puppy, and we have a small child who can be a bit rambunctious. We weren't sure that Lucy would be a great fit, but she was willing to give us a shot so we gave it our best effort.

I'm proud to say that Lucy's not only found her place in our home and hearts, but that she's also living her best life. 

May you find as much happiness as Lucy and her new family have:

Lucy the dog

 

Any good?

How was today’s newsletter? Amazing? Awful?! Help us make it better by sharing your brutally honest emoji feedback 👇

Positive Meh Negative

Feedback

Email tristan@thenextweb.com with any suggestions, complaints, or compliments. If you don't want to talk to him, email his boss: editor@thenextweb.com.

Facebook Twitter LinkedIn Instagram

From Amsterdam with <3