The music video by Quebonafide was directed and edited by Professor Andrzej Dragan.

Link do video

“Futurama 3” is the latest music video from rapper Quebonafide. Within just two days, the clip on YouTube garnered over 2 million views. The video was directed and edited by Professor Andrzej Dragan.

“Futurama 3” is Quebonafide’s newest music video, amassing over 2 million views on YouTube in just two days.

Dragan utilized AI algorithms for the video, as he explained on the platform “X”:

I think Dragan summarized it best himself:

“All the scenic backgrounds for Quebo’s new clip (like the one in the photo below) were generated in MidJourney. Additionally, many shots and transitions were created using Runway, Luma, and Stable Diffusion, scaled, of course, with Topaz. And let’s not forget the deep fakes on specially trained GANs. This allowed all the footage to be completed in one day. Naturally, all editing and post-production were done on a laptop. Best of luck to all ‘meat interfaces’ doubting the overwhelming advantage AI brings to film production. I’d like to remind you this is just months after this technology surfaced. And as a bonus: just a day before the premiere, Yann LeCun claimed in his signature style that the current generative technology is incapable of enabling film creation. After all, it’s just matrix multiplication.”

Coca-Cola creates an AI-generated version of its classic 1995 ad.

Speaking of AI video creation, Coca-Cola’s holiday ads this year were created entirely with AI. See for yourself:

Link do video.

My comment is brief: AI can make movies. Sometimes better, sometimes worse, but it can—and it’s only a matter of time before it turns the industry upside down.

Create your avatar to attend virtual meetings in a suit while you walk your dog.

There’s not much more to add. Create your avatar in professional attire and join all your calls in pajamas. 🙂

Link do video

What does this mean in practice? It suggests that Zuckerberg’s vision of our avatars meeting in the metaverse will come true much sooner and outside the metaverse. When we consider that AI can now monitor our screens in real-time and, for instance, assist with remote recruitment, we’ll soon have to reinvent recruitment processes altogether.

AI tracks productivity… or lack thereof

You know those movie scenes where call center employees face a big board showing daily/weekly/monthly targets? AI takes this to the next level. The latest computer vision solutions can track our time to the second. These systems detect how much time we spend at our desks and even whether we’re actually working there. See for yourself:

Link do video

Robotics

A lot is happening in robotics, so let’s start with the most positive developments.

Tired of doing laundry? This robot is for you.

We’ve already grown used to robots vacuuming and even mopping our floors while we’re away. But how about robots taking over laundry and folding?

Link do video

On a serious note, this is yet another area where AI will replace humans. It’s only a matter of time before robots like the one in this video become standard in every hotel.

Another area is restaurants. During my trip to Singapore, robots in both restaurants and airports:

• Guided me to my table
• Delivered my order
• Collected dirty dishes

Link do video.

For the first time, I also saw self-driving cleaning robots similar to those used in underground garages. The difference? These in Asia operated completely autonomously. I was even more surprised when, during a recent trip to Łódź, while telling a friend about my experiences with robots, we spotted an identical robot at the hotel. This revolution is already here!

Link do video. 

A GPT-powered humanoid.

Humanoids are nothing new; they’ve been around for years. Of course, there was recent buzz about Tesla’s humanoids, primarily because, as is Elon Musk’s style, they were supposed to be autonomous but turned out to be controlled by humans during the demo (#fakeItTillYouMakeIt). 🙂

However, the real revolution is happening elsewhere. The company Figure has introduced a new robot called Figure 2, powered by GPT-4.0, the model we know from ChatGPT. From the first model’s release, I’ve maintained that this is the true revolution. Robots, or manipulators to be precise, are nothing new—Japan’s industrial dominance was built on them. However, until now, these robots were expensive, and their operation (programming) was costly.

GPT-powered robots allow anyone to “program” them by simply speaking: “Move these pallets over there, then unload the truck and sweep up.” The next level of revolution comes with flexibility—a humanoid can replace many specialized machines, significantly lowering the entry barrier to this technology. The entire human labor sector in factories is now replaceable.

Consider this: the cheapest robots start at $18,000, which, at today’s exchange rate, is about 72,000 PLN. That’s just slightly more than the annual cost of an employee earning minimum wage (around 61,748.40 PLN) in 2024. Add to that the fact that such a humanoid:

• Can work 24/7. By comparison, a full-time employee works 1,872 hours annually (considering vacation) out of 8,760 hours in a year—just 21% of what a robot could achieve with battery swaps.
• Pays for itself in just 2.5 months (assuming 100% utilization); after that, it’s pure profit.
• Doesn’t strike.
• Doesn’t form unions… at least not yet. 😉

Link do video

Pozostałe 

AI inspektorem budowlanym 

New language models have been able to process not just text and images but also video for some time. An intriguing application was demonstrated by the Claude model, which, using footage from a camera worn by someone walking through a construction site, prepared a list of issues to address.

Link do video

More LLMs gain memory

Gemini, Google’s LLM, introduces memory functionality, following in the footsteps of OpenAI’s ChatGPT.

Jak działa funkcja pamięci?

1. Explicit memory:
   Users can actively instruct ChatGPT on what to remember. For example, if someone says they’re vegetarian, the model can tailor recipes and dietary recommendations accordingly. Preferences for writing style or professional details can also be remembered.
2. Implicit memory:
   ChatGPT can independently pick up on important details during a conversation. The more interactions with a user, the more helpful it becomes. For instance, if someone mentions they run a café, ChatGPT can adjust its responses, such as when creating social media posts.
   

Writing this newsletter was challenging for one reason: I had to limit myself. The sheer volume of news is so vast that if I wanted to cover everything, this email would be as long as a newspaper. Nevertheless, I hope this condensed overview was valuable to you.

See you in the next edition!