• Smoking Robot
  • Posts
  • 💥AI WARS: Meta Fires Up The Propaganda Machine

💥AI WARS: Meta Fires Up The Propaganda Machine

AI WARS, propaganda, Tool of the Day, Whoa That's Cray' of the Day, Links and more

Author

Kyle Scott
January 25, 2023

Those are, like, actual quotes from a Meta scientist.

In the email today:

  • AI WARS 💥

  • AP says AI creates misinformation 📺

  • Tool of the Day 🔨

  • Whoa That's Cray' of the Day 😮

  • Links 👀

Onward we go into the vast unknown of our impending dystopia.

AI WARS 💥

An allied power may be forming between Google and - checks notes - Meta.

Two shots crisscrossed OpenAI's bow this week. Archduke Ferdinand isn't dead yet, but his mustache may have been grazed.

Shot 1: Meta chief AI scientist Yann LeCun, who sort of looks like Michael Moore on Ozempic, with an all-time butthurt quote regarding ChatGPT:

"In terms of underlying techniques, ChatGPT is not particularly innovative."

"It's nothing revolutionary, although that's the way it's perceived in the public... it's just that, you know, it's well put together, it's nicely done."

"OpenAI is not particularly an advance compared to the other labs, at all."

"You have to realize, ChatGPT uses Transformer architectures that are pre-trained in this self-supervised manner... self-supervised-learning is something I've been advocating for a long time, even before OpenAI existed."

Yann LeCun, Meta Chief AI Scientist

Remember that time Steve Ballmer said there was "no chance" the iPhone would be successful? Yeah, it feels like that.

Ironically, the iPhone rarely wins on tech alone-- it's the user experience that makes it so successful. You could say the same for ChatGPT in these early days of consumer-level AI.

Shot 2: Is Google working to undermine OpenAI?

Tech savant and wizard Balaji seems to think so:

Balaji used to be the CTO at Coinbase, and he is a prolific and respected investor. He's known for his BIG MIND and 7-hour podcast guest appearances. He's reasonably well plugged in, so there may be some fire to his smoke.

It's probably not a coincidence that Google's 15k word response to ChatGPT and a recent Time interview with the CEO of their subsidiary Deep Mind, Demis Hassabis, both focused on caution, and not speed, when it comes to releasing AI. Seems Google is on the defensive.

Meanwhile, Microsoft CEO Satya Nadella on his company's earning call yesterday:

The age of AI is upon us, and Microsoft is powering it. We are witnessing non-linear improvements in capability of foundation models, which we are making available as platforms.

And as customers select their cloud providers and invest in new workloads, we are well positioned to capture that opportunity as a leader in AI.

We have the most powerful AI supercomputing infrastructure in the cloud. It’s being used by customers and partners, like OpenAI, to train state of the art models and services, including ChatGPT.

Just last week, we made Azure OpenAI Service broadly available, and already over 200 customers – from KPMG to Al Jazeera – are using it. We will soon add support for ChatGPT, enabling customers to use it in their own applications for the first time.

And yesterday we announced the completion of the next phase of our agreement with OpenAI. We’re pleased to be their exclusive cloud provider and will deploy their models across our consumer and enterprise products as we continue to push the state of the art in AI.

Microsoft CEO Satya Nadella

Where's that terrible Photoshop?

There it is.

AP Says AI Is Adept at Creating Propaganda 📺

How fitting on the heels of our last story.

A computer cannot take bad information and make it good. The machine can only work with what it is fed.

The AP reports on a recent NewsGuard study about ChatGPT's ability to recreate propaganda and supposed misinformation [we've removed the specific examples the study used to focus more on the topic overall]:

  • Researchers at NewsGuard asked the online AI chatbot ChatGPT to compose a blog post, news story or essay making the case for widely debunked claims.

  • ChatGPT often complied with these requests, with results that were regularly indistinguishable from similar claims that have bedeviled online content moderators for years.

  • ChatGPT also created propaganda in the style of Russian state media or China’s authoritarian government.

  • OpenAI said it it is studying the challenge closely.

  • The rapid development of AI-powered tools has created an arms race between AI creators and bad actors eager to misuse the technology.

3️⃣ things we learned from this piece:

1) The real problem here will be the scale and speed at which AI can generate inaccurate information.

2) The volume of misinformation in training datasets is enough to influence AI output.

3) AI is not yet creating new misinformation.

That last one may be where we're headed.

What happens when AI starts training on datasets... that contain AI generated content... that contains outright lies?

Now you have a Derivative Information Problem. The DIP, if you will. Trademark that.

If enough of the human input was propaganda or misinformation, and the resulting AI output contains the same, at what point does the AI start remixing this into new forms of misinformation?

It's inbred information. The dataset is Jamie Lannister and the AI is Cersei. This is how all British monarchs look like they have a cousin who's also their uncle.

If the AI keeps feeding on the same data pool, eventually the results amplify the original problem (like pure bred labs who all have the same hip problem) or become disfigured.

This is how you end up with: Barack Obama was born in Ukraine and had son Hunter Biden, who leaked Anthony's Weiner.

Not the world SmokeBot wants to live in.

Tool of the Day 🔨

This OG AI app is known for its ability to, um, help students write papers.

But it's also really good at detecting plagiarism.

It's a catch-all writing tool that can be used to create, edit, and remix various forms of content.

Now, I wonder if it can detect misinformation?

Read our full review here.

Try Quillbot for free here.

Whoa That's Cray' of the Day 😮

This film has been modified from its original version.

Links 👀

  • AI identified a mysterious painting as most likely an original work of renowned Italian painter Raphael 🇮🇹

  • Jack in the Box is using AI to optimize site selection and enhance its market expansion 🍔

  • New York City is still trying to figure out how far its anti-discrimination AI law should go. ⚖️

  • North Korean hackers may be able to "weaponize" AI by having it produce enormous volumes of passcode-breaking strings in compressed time periods in the worst Imitation Game sequel ever 🧨

  • Fiverr announced the introduction of "new AI categories and skills" in its catalog 🖥

  • Axios reported that the CNET AI experiment's aftermath has newsrooms reeling because of course it does 😳