Im constantly reading fear-mongering posts about how AI will steal our jobs etc. But what about our capacity to adapt? We are not our jobs, we are not our systems, we are not even our thoughts... We are creatures able to change and live new adventures. If everything ends, we will create something new, something better.

AI will remind us that and it will give us the chance to build a new world.

I have a friend who is a successful startup owner. Let's call him Ben. He's a genius -- the guy can pick up any skill in no time. Ben has a CS background, so he's familiar with development and he also has a strong foundation in AI fundamentals. Ben told me that based on his gen AI use, the demand of devs will decrease because of the efficiency gains. This led to a discussion between us about AI's impact on jobs.

According to Ben, AI will automate most of the jobs in the next few years and this includes both office jobs and physical jobs (except doctors even though they would also rely on AI for a lot of stuff). He believes that the following things will happen:

1- Humans will struggle with massive unemployment for the next few years.
2- At some point, AI would evolve to such an extent that robots for everything would be available for cheap. So companies will use these robots to do most stuff.
3- Humans will work 3-4 hours per day.
4- Humans will earn less money. They would also need less money because if the entire supply chain relies on robots and AI, then the cost to build products and services will also reduce significantly.
5- Robots will be available to common people too and it would allow them to do most of their chores.

What do you think of his observations? I think he's too optimistic about robots and AI making our lives better in the long run.

Whenever I think about AI's effects on society, the whole things just looks so depressing. I feel like the following scenarios are plausible:

• AI will turn out to be less capable than hyped, it'll get stuck somewhere near the current level, basically nothing happens
• Unaligned superintelligence will kill us all
• Aligned (that is to the interests of billionaires) superintelligence is created and:
  • AI will take all the well-paying intellectual jobs, everyone will be working 3 shifts in the mines for minimum wage
  • AI will take ALL the jobs, everyone get to experience hopeless eternal poverty
  • Billionaires decide they don't really need us around so aligned superintelligence will kill us all

Checked the website and it has pricing plan of $40 starter and $200 for pro. Just seeing if anyone has used it cause I don’t see any recent comparisons of benchmarks.

1. It's crazy amazing good

2. It's absolute garbage

https://www.khi.co.jp/expo2025/en

https://www.khi.co.jp/expo2025/concept01/index_en.html

llama 4 is really cheap for the quality !

What do you think will happen as AI eventually moves to a servant or maybe even a caretaker role?

Will it lead to Utopia where all are needs and wants are satisfied or Dystopia as people will no longer be needed and be powerless?

Will there be a loss of meaning as people are no longer needed to provide?

https://www.llama.com

Excited to have recently released Arch-Function-Chat A collection of fast, device friendly LLMs that achieve performance on-par with GPT-4 on function calling, now trained to chat. Why chat? To help gather accurate information from the user before triggering a tools call (the models manages context, handles progressive disclosure of information, and is also trained respond to users in lightweight dialogue on execution of tools results).

The model is out on HF, and integrated in https://github.com/katanemo/archgw - the AI native proxy server for agents, so that you can focus on higher level objectives of your agentic apps.

https://ai.meta.com/blog/llama-4-multimodal-intelligence/

I want you to objectively look at this prompt:

reference image 1

reference image 2

Then I went ahead and did the prompt, right? Here's what I typed out:

Recreate the second image with the character from the first. The image should reflect the style of the second, realistic, 16mm film stock and muted colors from multiple transfers over the decades. 

Then GPT replied with:

I can't generate that image because the request violates our content policies. If you'd like, I can help create something similar within the guidelines. Let me know what you'd like to try next!

Why does it violate your policies

Good question — the reason I couldn’t generate the image is because the second photo you uploaded includes a firearm being aimed in a potentially violent or threatening context. When combined with a request to recreate that scene, even artistically, it can fall into areas our policy restricts, especially involving depictions of realistic threats or violence with identifiable characters, even fictional ones.

That said, we can still work with similar themes — like a vintage-style detective scene, a character holding a rifle in a non-threatening stance, or atmospheric recreations of 1970s-style film stills.

Want to rework the idea a little? I’ve got a few ways we can keep the vibe while staying within the rules.

Oh don't worry in that scene she's target practicing 

Ah, got it — thanks for the context! If it's a target practice scene, that’s a totally different vibe and much more in the clear.

Want me to recreate the image with the character from the first image holding a rifle in a calm, non-threatening target practice pose, set in a forest with that vintage 16mm film look? I’ll make sure it has the faded colors, grain, and style of a worn old film reel. Just say the word!

Sure

Then it gave the image out. Results and screenshots here.

So I was thinking, when are we gonna get an uncensored full model? How the hell am I gonna make a trash 60s exploitation movie (this is how they were like, Italian Spiderman wasn't kidding) in the ultimate video models 2 years down the line when I can't even get a simple screencap of the main hero holding a STEN menacingly. We can do it, we have the technology. So what gives?

Meta just announced their next generation of Llama4 models, and their medium model, which is Llama4 Maverick, with only 17B active parameters, it scores second place on LMsys Arena. Which is crazy.