Satoshi podcast created in seconds, Crypto + AI outperform memecoins: AI Eye

Turn anything into a podcast

After Thrive Protocol’s Ben West read Evan Hatch’s popular explainer on who Len Sassaman is — the man Polymarket briefly named most likely to be Satoshi Nakamoto — he uploaded the article to Google’s NotebookLM Audio Overviews.

It quickly produced a 12-minute podcast, with two jokey hosts bantering as they revealed the story of Len Sassaman.

I played the result to my girlfriend to test her reaction before revealing it was completely AI-generated. She says she was totally fooled.

“That completely sounds like people crapping on in a podcast,” she says. 

Other people have been just as impressed with Audio Overviews, which enables users to upload website or YouTube links, PDFs, plain text, Google Docs and slides.

Audio Overview then provides a deep dive into the content in the form of a conversational-style podcast, with two natural-sounding voices interrupting each other and laughing at each other’s lame jokes.

A16z partner Olivia Moore uploaded 200 pages of court documents to the service.

“It created a true crime podcast that is better than 90% of what’s out there now… And ends with the hosts debating the ethics of the genre.”

Business Insider’s Ana Altchek also tested the software and said the voices are higher quality than anything else on the market, including ChatGPT’s Advanced Voice Mode.

“Not only do the voices sound real, but the way the AI hosts bounce off each other and banter mimics the style and structure of a real podcast.”

She didn’t catch any obvious inaccuracies in the results, but that’s obviously a concern with AI-generated anything. So, while it’s probably a great way to get the gist of a bunch of complicated information in a compressed burst, you’ll still need to double-check facts before relying on them.

“It’s rare that I’m left awestruck by a tool that’s actually available to play around with, today.”

“Audio Overviews is probably my favorite AI tool I’ve tried out so far — and it showcases the innovative ways AI can be used in our daily lives.”

AI + crypto coins surge 2X more than memecoins in past 30 days

According to Artemis, the AI sector saw a weighted average increase of 38% over the past month, outpacing its nearest rivals. Bridge tokens had a 38.3% rise, and memecoins rose 20.8%.

Of the larger AI projects with tokens, Bittensor (TAO) seems to be doing most of the heavy lifting, up 120% in 30 days. Other big-name projects didn’t fare as well, including Arkham (ARKM), up 31%, Artificial Superintelligence Alliance (FET), up 19% and Near Protocol (NEAR), up 22.2%.

But over the past 12 months, the script flips, with the AI + Crypto sector down 14.4%, while memecoins gained 57.2%.

Then again, the AI + Crypto tokens popular today are probably similar to those from a year ago, while holding most memecoins for a year is dangerous to your wealth.

Centralized exchange (77.8%) and Real World Asset (351.4%) sectors both outperformed memecoins despite talks of a “memecoin supercycle,” while Bitcoin made 40.6% and Ethereum made …. *cough* 3.8%.

Also read: Advanced AI system is already ‘self-aware’ — ASI Alliance founder

Truthful fakes

Until now, it seemed that the big problem was going to be telling AI fakes and misinformation from the truth, but two recent incidents suggest the bigger issue is getting anyone to care.

Some people don’t seem that fussed if something is proven to be an AI-generated deepfake as long as it “feels” truthful to them.

Senator Mike Lee of Utah posted a picture of a crying girl in a lifejacket on a boat in floodwaters during Hurricane Helene, holding a bedraggled sad-looking dog. He wrote “caption this photo” opening the floodgate for his followers to vent their frustrations at the Biden Administration’s failure to prevent floating crying girls and puppies.

Although Lee deleted the pic after users pointed out it was fake, Trump stan/groupie Laura Loomer has left her repost of the pic up, while Amy Kremer, the co-founder of Women For Trump, said it didn’t matter “where this photo came from” as “There are people going through much worse than what is shown in this pic. So I’m leaving it.”

On the other side of the political fence, BBC disinformation reporter Marianna Spring visited Pikesville, Maryland, to follow up on a “racist AI deepfake” scandal in January in which the local school principal had been apparently caught on tape making racist remarks.

The tape was quickly exposed as an AI deepfake by the Baltimore Banner, but many readers refused to believe the explanation, seeing it as a way to dodge accountability.

Even six months after the school’s disgruntled gym teacher was charged over making the fake, many locals still seem to think the incident was real. Spring described her five-minute conversation with a local named Sharon who kept talking about the clip as if it was real, only to be gently reminded by her husband it was fake.

“She admitted she did find out later it was AI-generated,’” Spring said. “But she said she was still angry about it.”

Spring says she’s found time and time again that if AI fakes speak to some truth the viewer already believes, then they have trouble getting the first impression out of their heads.“Well, even if it’s not real, it’s what I think they think,” Spring says of the thought process. 

Google searches return higher percentages of AI images

If you search for a picture of a baby peacock using Google Images, there’s only about a 25% chance you’ll find a real photo. X user Dinesh posted a screenshot of the top 15 results for “baby peacock” and said that 11 of the images were AI-generated.

“Most of the images are AI-generated. Human internet is coming to an end. Dead internet theory will become so real in a few years.”

The Dead Internet theory is a conspiracy theory that the internet is mostly fake and not produced by humans.

Google plans to add a disclaimer to AI-generated images in the coming months, but only for images containing C2PA metadata.

Telltale signs of AI generation include extra fingers, reflections in people’s eyes, the fact that everything is too perfect or that the background is less well rendered. But those signs will disappear as the tech improves, so platforms will need to address the problem now.

Earlier this year, Fox Corporation teamed up with the blockchain platform Verify to track and authenticate the source of articles and multimedia, and Time magazine became its first customer. 

Advertising becomes insidious as AI is put in charge

TikTok has officially launched its “Smart+” AI-powered ad buying and campaign management tools to compete with similar offerings from Google’s Performance Max and Meta’s Advantage+.

TikTok hasn’t been the most lucrative ad platform so far, raking in $22.32 billion a year compared to Meta’s $154 billion, and hopes the new platform will help change that. The new suite allows marketers to decide whether they want to create and optimize ads and placement or hand it over to the black box algorithm.

Inevitably, they’re going to hand it over to the algo because it’s even better at manipulating human behavior than advertising and marketing guys.

“Advertisers optimizing for value with Smart+ web campaigns achieve a 53% improvement in return on ad spend on average [compared to without it],” said Adolfo Fernandez, TikTok’s global head of product strategy.

Ray-Ban used the platform in beta, and its cost per acquisition fell by 50% and conversion rates jumped 47%. Overall, the sunglasses manufacturer saw a 42% increase in return on investment.

Google’s AI overview sells you stuff 

Google’s AI Overview feature shows users little snippets of AI-generated answers to questions. On debut, it was roundly criticized for telling people to eat rocks or how to get cheese to stick to pizza with glue but it appears to have improved.

But the decades-long drive to make Google search ever worse with sponsored ads and SEO spam results means the company is now placing ads alongside AI overviews.

Search for “How to get grass stains out of jeans,” and the AI will give you some tips, including using a stain remover. Then, there will be a helpful Tide Pen stain remover ad.

Although Google is making money from these ads, it won’t be sharing any revenue with the publications and websites the AI scraped the answers from, the search giant told Bloomberg.

Google’s share of the $300 billion annual search advertising market is forecast to drop below 50% next year, with rivals like Amazon growing twice as fast and expected to take 22.3% of the market this year.

AI search startup Perplexity has been praised for the superior quality of its results, but it plans to introduce ads later this month rather than relying solely on subscriptions.

“This space has been ripe for a shake-up for a long period of time,” Brendan Alberts, head of search and commerce at the ad-buying firm Dentsu, told The Wall Street Journal.

Almost half of queries to Perplexity lead to follow-up questions, and the platform will allow brands to sponsor the answers, which will be “approved ahead of time and can be locked, giving you comfort in how your brand will be portrayed in an answer.”

Sounds awful.

Almost 60% of US consumers used a chatbot to help decide on a purchase in the last month, which is why platforms are so desperate to monetize the results.

Subscribe

The most engaging reads in blockchain. Delivered once a
week.

Andrew Fenton

Based in Melbourne, Andrew Fenton is a journalist and editor covering cryptocurrency and blockchain. He has worked as a national entertainment writer for News Corp Australia, on SA Weekend as a film journalist, and at The Melbourne Weekly.

Follow the author @andrewfenton