DAI#51 – OpenAI woes, Olympic flops, and AI future crime

Welcome to our roundup of this week’s spiciest AI news.

This week, Elon Musk claimed Sam Altman tricked him.

Google flopped at the Olympics.

And Argentina’s president wants to use AI to predict crime.

Let’s dig in.

Back in court

Earlier this year Elon Musk sued OpenAI and his fellow co-founders only to drop the lawsuit a few months later. This week Musk filed a new lawsuit against Sam Altman, Greg Brockman, and OpenAI with additional claims.

The new allegations and charges offer fascinating insights into the early discussions before OpenAI became a reality. The lawsuit also shines a light on how seriously Musk views AI’s potential to destroy humanity.

The short version is that Musk says he was duped and OpenAI must pay up. Did things go down like Musks says, or could his memory be a little fuzzy?

via GIPHY

A rush for the exit

OpenAI’s week didn’t get much better as its leadership was rocked by another set of high-profile exits.

It seems like the rush for the emergency exit doors hasn’t slowed down yet. What has slowed down is the announcement of new product releases.

GPT-4 got a slight bump in performance this week but it still lags Claude Sonnet on the benchmarks. In the absence of GPT-5, voice assistants, and Sora, we’d even settle for a demo of some new feature just to keep our hopes up.

Google’s Olympic fail

The Olympics are a celebration of excellence but Google’s new Gemini ad won’t be winning any medals. The company pulled its controversial AI ad from Olympics coverage amid swift backlash.

The idea of using Gemini to replace heartfelt expressions from an ambitious little girl should probably never have made it past the marketing idea board stage.

I love using Gemini but while it still suggests that staring at the sun is a good idea, maybe we should let the kids write their own fan mail.

It’s tough at the top

NVIDIA has had a meteoric rise but being top dog comes with additional scrutiny. The company faces two antitrust probes from the U.S. Department of Justice (DOJ) over alleged anticompetitive behavior.

NVIDIA has also hit a few snags with its anticipated Blackwell chips which could give some hope to its competitors. Intel’s share price took a major dip this week but its Gaudi chips could see it claw back some of NVIDIA’s market share.

CheatGPT

It’s becoming almost impossible for educators to tell if their students are using ChatGPT to write their assignments. Accurate AI-generated text detectors just don’t seem to exist. Or do they?

It turns out OpenAI has had a tool for over a year that can detect text generated by ChatGPT. They say they’ve got good reasons why they haven’t released it but maybe there’s more to the story.

CheatGPT ?? pic.twitter.com/s1yGv28Hlc

— Renoir ❁ (@RenoirAbadie) April 14, 2023

Future crime

In the movie adaptation of Philip K Dick’s Minority Report, Tom Cruise’s character finds out the hard way that predictive policing doesn’t work as well as advertised.

Argentina’s fiery new president Javier Milei has decided to throw AI at a bunch of the country’s problems, including security. His new “Artificial Intelligence Applied to Security Unit” will use AI to ‘predict future crimes’.

Are you a criminal? Are you sure? Let’s ask the AI what it thinks you might get up to in the future.

In other news…

Here are some other clickworthy AI stories we enjoyed this week:

• OpenAI says the US government will get to see ChatGPT 5 before anyone else.
• AI music startup Suno claims training its model on copyrighted music is ‘fair use’.
• Google Gemini 1.5 Pro leaps ahead in the AI race, challenging GPT-4o.
• ByteDance researchers developed a full duplex language model that listens while speaking for more natural dialogue.
• Nvidia scraped videos from YouTube and other sources to train its AI products.
• Keir Starmer says facial recognition tech is the answer to far-right riots.
• Figure unveiled its latest humanoid robot, Figure 02 and claims it’s the world’s most advanced AI hardware.

Introducing Figure 02

The world’s most advanced AI hardware

Exclusive pics + technical writeup pic.twitter.com/2cts3pTIcN

— Brett Adcock (@adcock_brett) August 6, 2024

And that’s a wrap.

Did you read Elon Musk’s new legal filing against OpenAI? They could publish it as a book and sell it on Amazon with zero edits. It’s fascinating.

Let’s hope OpenAI gives us something more exciting to report next week. Less drama and more wow factor, please Sam.

Do you want the ChatGPT text detector to be released or do your grades depend on OpenAI holding off on that? Even if they do add watermarks, it seems like they’ll be pretty easy to remove.

Let us know what you think, follow us on X, and share any cool AI stories we may have missed.

The post DAI#51 – OpenAI woes, Olympic flops, and AI future crime appeared first on DailyAI.