Artificial Intelligence

[Xerxes]

 

[UK]

8 hours ago, Xerxes said:

 

Quite ipressive!

[Haryana]

https://www.cbsnews.com/news/google-ai-chatbot-threatening-message-human-please-die/

Google AI chatbot responds with a threatening message: "Human … Please die."

"This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please."

[rkbabang]

50 minutes ago, Haryana said:

https://www.cbsnews.com/news/google-ai-chatbot-threatening-message-human-please-die/

Google AI chatbot responds with a threatening message: "Human … Please die."

"This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please."

 

 

OK, well, next time tell me how you really feel.   

 

Maybe it's not quite safe to give AI the nuclear codes yet.

 

[Spekulatius]

I look forward to my AI model to become a self conscious and improved version of myself long after I am gone. Thats probably how we are going to live forever and explore space by shedding out biological confinement first. Let’s go!

[nsx5200]

Anybody have experience with book summarizer services like Blinkist or sumizeit?  Tempted to get the service to help triage books/articles to digest.  NotebookLM is my current go to, but if a paid service has better quality, or at least access to more material at the same quality, I might go for it.

[e4d5]

I use Merlin AI to summarize blog posts, articles, and YouTube videos, but I've never tried it for book summarization.

[rogermunibond]

https://www.nature.com/articles/d41586-024-03990-2

 

Running out of data

[bargainman]

cross posting.  Good brief accessible intro and visualization of how LLMs work.

 

 

[rogermunibond]

So far the claims of performance appear legit.  Training cost of $5.6M is amazing if corroborated.  Extrapolating can kill you if you think GPUs, electricity etc are straight line trends.

 

https://venturebeat.com/ai/deepseek-v3-ultra-large-open-source-ai-outperforms-llama-and-qwen-on-launch/

[formthirteen]

 

Welcome to the unevenly distributed future.

[james22]

 

After being very critical of Trump's tariffs, Larry said historians looking back will remark only on the AI breakthroughs of this time. Trump would be a distant second. Tariffs wouldn't rate a mention.

[james22]

Larry is a OpenAI board member.

[Rainier]

OpenAI o3 is out in the wild today. I am not the most informed person when it comes to AI, but the claims that we’ve reached AGI keep increasing. I don’t think we’ve reached what I would define as AGI. But these LLMs seem to have reached the “reasoning and expressing” part that would be necessary. I think of AGI as something that can reason and express itself, but can also plan and execute its own propagation without any software or hardware inputs from humans. So, thankfully, I don’t think we are at that point. Yet.

Examples of complicated output: 

 

https://marginalrevolution.com/marginalrevolution/2025/04/a-note-on-o3-and-agi.html

[Fly]

11 hours ago, Rainier said:

OpenAI o3 is out in the wild today. I am not the most informed person when it comes to AI, but the claims that we’ve reached AGI keep increasing. I don’t think we’ve reached what I would define as AGI. But these LLMs seem to have reached the “reasoning and expressing” part that would be necessary. I think of AGI as something that can reason and express itself, but can also plan and execute its own propagation without any software or hardware inputs from humans. So, thankfully, I don’t think we are at that point. Yet.

Examples of complicated output: 

 

https://marginalrevolution.com/marginalrevolution/2025/04/a-note-on-o3-and-agi.html

 

Also OpenAI has incentive to call it AGI as soon as practical, just something to keep in mind

 

https://www.reuters.com/technology/openai-seeks-unlock-investment-by-ditching-agi-clause-with-microsoft-ft-reports-2024-12-06/

[no_free_lunch]

On 4/17/2025 at 7:37 AM, Rainier said:

OpenAI o3 is out in the wild today. I am not the most informed person when it comes to AI, but the claims that we’ve reached AGI keep increasing. I don’t think we’ve reached what I would define as AGI. But these LLMs seem to have reached the “reasoning and expressing” part that would be necessary. I think of AGI as something that can reason and express itself, but can also plan and execute its own propagation without any software or hardware inputs from humans. So, thankfully, I don’t think we are at that point. Yet.

Examples of complicated output: 

 

https://marginalrevolution.com/marginalrevolution/2025/04/a-note-on-o3-and-agi.html

I am not the expert either but what I read is that the tech is stalling.  There are reports that the latest version, presumably o3, is using 10x the compute and yet the results are now hard to distinguish from the older version.  Perhaps they are starting to plateau?  Or perhaps the AI is secretly using the compute to build skynet?   Regardless, I do not doubt the tech will advance but it may be more incremental from here.  It may move to an efficiency drive, eg deepseek.  I predict deflation in this sector.  This always happens in tech, where the technology comes out and then they find ways to do it better, faster, stronger.. maybe not the last one.  Maybe better faster cheaper anyways.

Edited April 20, 2025 by no_free_lunch

[Dave86ch]

I’ve realized the value of creating qualitative content by asking GPT-4 (o3) to analyze my blogs, tweets, and a PDF where I share various thoughts.

 

 

[Gamecock-YT]

2 hours ago, Dave86ch said:

I’ve realized the value of creating qualitative content by asking GPT-4 (o3) to analyze my blogs, tweets, and a PDF where I share various thoughts.

 

 

 

What did it say from a psychoanalytic perspective. Might be cheaper than therapy. 

[RichardGibbons]

I'm curious what it would say if you tell it that someone you hate wrote this stuff and whether it would be less complimentary.

[Blake Hampton]

4 minutes ago, RichardGibbons said:

I'm curious what it would say if you tell it that someone you hate wrote this stuff and whether it would be less complimentary.

I fear ChatGPT confirmation bias.

[Dave86ch]

10 hours ago, Blake Hampton said:

I fear ChatGPT confirmation bias.

 

Decision-making about one's life must be based on multiple perspectives. Writing, and then processing those writings through an LLM, is simply a way to connect some of the dots.

Creating digital content becomes an asset in itself.

Personally, I define my success through four dimensions:

Physical fitness

Wealth

Emotional stability

Mental clarity

I also share my blood tests, running and weightlifting routine, supplements, and dietary regimen with O3, asking it to remember everything.

Then I ask various questions to triangulate its insights.

This approach is still far less biased than relying on the average person, who typically:

Doesn’t cite sources

Has opinions shaped by envy or some other bias.

Operates with limited context.

 

O3 is proving to be a powerful tool, increasingly valuable in managing and optimizing various aspects of my life.

 

Edited April 22, 2025 by Dave86ch

[james22]

18 hours ago, Dave86ch said:

Personally, I define my success through four dimensions:

Physical fitness

Wealth

Emotional stability

Mental clarity

 

Only four?

 

 

 

 

[Dave86ch]

4 hours ago, james22 said:

 

Only four?

 

 

 

 

I agree.

However, I summarized the most important points that came to mind, those for which I had some rational metrics to analyze.

[james22]

Gotcha.

[Blake Hampton]

We Now Know How AI ‘Thinks’—and It’s Barely Thinking at All - WSJ

 

Here’s How Big the AI Revolution Really Is, in Four Charts - WSJ

 

Edited April 26, 2025 by Blake Hampton