Thread of Videos

OpenAI seemingly forgot how AI testers / the Internet worked, when they gave out platform access to some testers, and unsurprisingly, footage of those tests leaked out. OpenAI is now scrambling to revoke access to those artists after the leakers claimed they leaked it for unpaid labor. These artists said they felt exploited for providing research and development and not really supporting artistic development.

PR puppets, the group behind the leak, is aiming to try to get a more equitable relationship between AI tech and artists. OpenAI responded to this leak by temporarily halting access, claiming that the test was voluntary without any obligation to provide feedback.

The “good” news, is it shows what Sora can do, but the footage that was leaked didn’t seem better than currently available GenAI Video models. It doesn't seem like OpenAI went with malicious intent, and instead, PR puppets wanted to be paid for providing testing as is expected in other indusstries.

If you are specifically seeking out experts, you should pay them for their time. You pay them for their time in most industries, why would you not do this here? OpenAI wants to hide behind the notion of “voluntary testing”, which doesn't really hold up.

Andrew Robinson of Runningbird and Mindcorp recently showed off how AI aids in creating pitch materials like synopses, decks, storyboards, and sizzle reels. They hope AI's agility and speed will help producers understand and visualize stories with minimal effort and budget.

The problem?

This is almost like developing a new way to develop film for everyday users after the release of a digital camera. This technology benefits the large studio model, but if the film / tv of the future is produced with smaller teams or even one-person this mass review / production of material is likely less important.

Many of you believe that your phone is still listening to you, despite stories and podcasts convincing that it's not true. Now there appears to be more fear about AI developing a digital profile of you and knowing more about you than Google or Meta.

In response, the advertising industry is considering something called contextual targeting, which aims to use less personal data and be more focused on recognizing patterns across content, in order to show ads that are relevant to you. This should be a happy medium between people who want to see relevant ads but don't want to have their personal data just sold everywhere on the Internet.

"Putin," directed by Patryck Vega, examines Vladimir Putin's life and ascent to power, using AI technology to create a realistic portrayal by actor Slawomir Sobala.

The film explores Putin's manipulation of Russian bureaucracies and his control over media, depicted through scenes like smuggling toilet paper into Moscow.

With a budget of $14 million, the movie was filmed across multiple countries, including risky locations like Chernobyl, reflecting the director's commitment to authenticity.

Elon Musk has decided that he wants to individually “Make Gaming Great Again”. His move comes in response to the perception online that games have become too “woke”

There's definitely an argument to make that games are more woke than they've been in the past, but people still seem to be playing those games unlike most woke movie. Naturally, he's going to use AI, which could be helpful in finally showing that AI can make innovative games.

if you hate typing but don't mind talking, Humai may have the app for you. They've developed a system where users can control their computer by voice, representing the next step forward in human technology interaction. I expect that we're going to start seeing more moves in this direction, with fewer hard interfaces and more voice commands.

How this comes to play in an office setting, I'm still a bit curious about. You likely need a future office with white noise, that you don't easily hear people talking to their computer. You may also have a solution that more people work from home, allowing you to talk with an app to talk through ideas and then generate new things.

I'm really excited to see what this partnership with Humai and Anthropic does because they will also give a good competitor to what OpenAI is already working on with their own desktop app & voice app.

xAI is also trying to explore creating its own standalone app, to make the app more easily accessible to most people. I don't know who the audience is for this though.

I don't view Grok and what xAI has developed so far as being that good. For me personally, the thing I use it most for is getting Flux images generated, but I still use probably Claude or to some lesser extent ChatGPT or even more for regular workflows.

What changes that perception for me, is if they can better integrate real-time data from X. Despite being a Perplexity power user, I find it struggles with pulling “News of the Day”.

Post Office Scandal | Drama Version (Worth Watching)

The UK government is falling behind on their use of AI and transparency. Despite saying that there's going to be an AI registry of AI apps used by individual UK government departments, there's been nothing published since February 24 when that regulation was supposed to come into effect.

Based on contract awards it's known it is being used in various government sectors, but the extent is unclear. This is particularly important to the UK after the Horizon Post Office Scandal of the early 2000s.

If you, like me, had never heard of this scandal before, I highly recommend watching the tv drama about it. The TLDR is the UK government had an accounting system called Horizon installed in Post Offices that falsely accused people from stealing from the Post Office leading to Arrests, Suicide, and Financial ruin. As we approach a more AI-enabled future we must take care we have safeguards in place to prevent this or a Minority Report style scenario.

The OpenAI vs. New York Times and Daily News lawsuit continues to provide some interesting stories. Most recently, OpenAI “accidentally” deleted data that would've been crucial to the lawsuit. While they managed to recover much of it, the data is unusable for legal analysis.

It's a bit funny because you would think OpenAI would know how to avoid deleting data, but "oh we lost it" or "the AI ate my homework" is more interesting. OpenAI claimsthere was just a system misconfiguration that caused that which seems a bit odd.

This case is going to be interesting because it will likely determine how copyright creators get compensated for the work being used to train AI models. How that system will work though is unclear.

Pathway, a startup focusing on 'Live AI,' has secured a $10M seed round to enhance AI systems with real-time learning capabilities, aiming to make AI more dynamic and useful for enterprises like NATO and La Poste.