šŸŖ™ FTX sues Binance for $1.8B

PLUS: The Beatles win a Grammy?; ChatGPT needs a math tutor

Todayā€™s market performance šŸ†ļø

S&P 500: +0.09% šŸ“ˆ
Nasdaq 100: -0.05% šŸ“ˆ
Dow 30: +0.69% šŸ“ˆ
Russell 2000: +1.47% šŸ“ˆ

In partnership with

FROM OUR FRIENDS @ THE RUNDOWN AI

Learn AI in 5 minutes a day.

The Rundown is the worldā€™s most trusted AI newsletter, with over 700,000+ readers staying up-to-date with the latest AI news, understanding why it matters, and learning how to apply it in their work.

Their expert research team spends all day learning whatā€™s new in AI, then distills the most important developments into one free email every morning.

TOP STORY
šŸŖ™ FTX sues Binance for $1.8B

šŸ“ø Protos

You're still around? FTXā€”the once $32 billion crypto exchange that sent its founder to jail and lost people millions, is apparently still alive.

Donā€™t call it a comebackā€¦

Earlier this week, the estate of this former industry leader filed a suit against Binance and its former CEO, Changpeng Zhao for a whopping $1.8 billion.

šŸ’¬ FTX collapsed in 2022, leading to a 25-year prison sentence for its founder, Sam Bankman-Fried, while Binance's former CEO, Changpeng Zhao, received a 36-month sentence for anti-money laundering charges in 2023.

Binance Founder and Ex-CEO, Changpeng Zhao (šŸ“ø Getty Images)

Thatā€™s one way to announce your back from the grave, ha. Whatā€™d Binance do?

Well first some background: Binance, like FTX, was a major cryptocurrency exchange just a couple of years ago; in fact, it was the largest in the world by volume at its peak.

The problem:

  • In 2021, Binance, Zhao, and others sold their shares in FTX back to the company, including some of its U.S. operations. 

  • FTX used money from its trading partner, Alameda Research, to buy these shares, even though Alameda was already struggling financially and couldnā€™t really afford it.

  • Now, FTX is arguing that this buyout was "fraudulent" because it relied on unstable funds the company didnā€™t actually have, which ultimately contributed to its collapse. 

But didnā€™t you let it happen, FTX? Couldā€™ve done some more due diligence, just saying.

Talk about throwing someone under the bus.

Binance has (obviously) denied the claims and says it "will vigorously defend ourselves."

And thatā€™s on period, said Zhao.

šŸ“ˆ Bitcoin ($BTC) is up ~100% this year.

ENTERTAINMENT
šŸ†ļø The Beatles win a Grammy?

šŸ“ø Wikimedia Commons

Howā€™s that even possible? The Beatles, you know the words most famous band and such are nominated for 2 Grammys.

Clones? Time Machine? Not exactly, very close though, AI.

The Beatlesā€™ song ā€œNow and Then,ā€ was refined using AI sometime last year and now, itā€™s up for both Record of the Year and Best Rock Performance.

Paul McCartney

How did this come to be? As the story goes, one day, Beatles singer and songwriter Paul McCartney decided to use AI to clean up a 1978 demo by John Lennon, preserving Lennonā€™s original voice.

McCartney was inspired by tools from Peter Jacksonā€™s 2021 documentary ā€œThe Beatles: Get Backā€ and decided to use similar AI tools to separate voices and instruments for a clearer recording.

šŸ’¬ This AI-based audio editing is similar to how video chat platforms like FaceTime, Google Meet, or Zoom might filter out background noise from a call.

That's awesome, and it's legal? Yes, since McCartney is a surviving member of the Beatles, he actually has the right to use any Beatles recordings and technology to create new content from archived material.

Who wouldā€™ve thought that in 2024, Billie Eilish and Kendrick Lamar would be competing with The Beatles for a Grammy?

ARTIFICIAL INTELLIGENCE
āž• ChatGPT needs a math tutor

OpenAI is done with the delays.

In hopes of jumping past the hurdles holding back AI companies building large language models, OpenAI and others are working on training methods that help AI ā€œthinkā€ a little more like humans.

What does that actually mean? It means that theyā€™re making models smarter, you idiot.

  • OpenAI is training its models to make decisions by considering multiple options before settling on an answer, similar to how most people, (not all), solve problems.

  • In principle, it means GPT will now consider multiple options before deciding; in practice, it means GPT might take a few extra seconds to give you an answer or solve your homework.

It sounds promising, but how exactly would this help with building LLMs faster?

Essentially by making the models smarter with fewer resources and using thoughtful decision-making instead of massive data and computing power, this approach speeds up training.

How are they doing it? 

Theyā€™re training models with expert-curated data, including feedback from PhDs and specialists in fields like math, coding, and language.

Anyone know a good math tutor for ChatGPT?

Reply

or to participate.