Random AI/Robot Thoughts and News

[The Cat]

"Don't worry, It's totally unsinkable." - Thomas Andrews jr. (7 February 1873 – 15 April 1912)

 

[ygolo]

The sink of the Titanic didn't lead to banning doing anything involving water.

Facts About Dihydrogen Monoxide

Dihydrogen Monoxide Research Division provides an international clearinghouse for information on Dihydrogen Monoxide including current research results, governmental regulations, and exposure and contamination alerts. Dihydrogen Monoxide (DHMO) can be dangerous to humans. Dihydrogen Monoxide...

www.dhmo.org

 

[The Cat]

You're plucky and passionate. I dig that.

 

[ygolo]

You're plucky and passionate. I dig that.

So, are you calling for a ban of water?

 

[The Cat]

So, are you calling for a ban of water?

Nope. Sorry to disappoint. I'm afraid you'll have to keep searching for someone to fight you over a technology fad whether passing or not. I am passingly curious as to why you correlate any potentially perceivable skepticism or I suppose cynicism on anyone's part that they must be calling for as you say a ban on water?

 

[ygolo]

Nope. Sorry to disappoint. I'm afraid you'll have to keep searching for someone to fight you over a technology fad whether passing or not. I am passingly curious as to why you correlate any potentially perceivable skepticism or I suppose cynicism on anyone's part that they must be calling for as you say a ban on water?

I'm glad you have perspective .

Because a lot of people don't. I made a whole thread about the laws and general turn in sentiment.

Software, however, is not a passing fad. From it's inception, "AI" just meant the cutting-edge software. As a buzzword, it comes and goes.

I'm skeptical myself of a lot of it. But a lot of it has value.

 

[ygolo]

Sneaky Shiftiness On The Boundaries Between AI Versus AGI And Ultimately AI Superintelligence

A lot of sneakiness is happening when it comes to declaring or predicting AI, AGI, and ASI. Sadly, the tricksters are getting away with it. Here's the insider scoop.

www.forbes.com

 

[The Cat]

My biggest cause for skepticism comes from how this whole thing is gonna be managed. Pay to play seems the way its gonna end up, and that makes me pretty cynical. I've appreciated places who are using it in a way that lets users use it for free. I'm pretty sure it's made editing the minis in heroforge a lot more doable than it used to be. I dont mind artists using it to help get their content out. I'd like for it to be something that can be profitable for everyone, but I cant shake the idea that its just gonna be like any of the other cool tech out there, for the military and millionaires.

Tbh, I'm a little bummed. Honest question, what do you think is gonna end up happening? Once its all said and done either way in the next 5 years?

 

[ygolo]

My biggest cause for skepticism comes from how this whole thing is gonna be managed. Pay to play seems the way its gonna end up, and that makes me pretty cynical. I've appreciated places who are using it in a way that lets users use it for free. I'm pretty sure it's made editing the minis in heroforge a lot more doable than it used to be. I dont mind artists using it to help get their content out. I'd like for it to be something that can be profitable for everyone, but I cant shake the idea that its just gonna be like any of the other cool tech out there, for the military and millionaires.

Tbh, I'm a little bummed. Honest question, what do you think is gonna end up happening? Once its all said and done either way in the next 5 years?

The biggest thing to realize it's not just one thing.

Big Tech likes the pay-to-play nature. Little Tech does not.

People won't stop making software.

The question is, will those making it, including people making the training data, get a piece of the prosperity, and get credit for the productivity that it creates.

Fatalism is self-fulfilling.

 

[ygolo]

We knew this was coming, and here it is:

ChatGPT x Apple Intelligence—12 Days of OpenAI: Day 5

Sam Altman, Miqdad Jaffer, and Dave Cummings introduce and demo ChatGPT integration into iOS and macOS while wearing holiday sweaters.

www.youtube.com

Again, I don't see AGI. But it could be useful, maybe? Novelty use cases right now. I'm not sure the price.

 

[ygolo]

Introducing Gemini 2.0: our new AI model for the agentic era

Today, we’re announcing Gemini 2.0, our most capable AI model yet.

blog.google

 

[ygolo]

https://www.sfgate.com/tech/article/sf-tech-startup-scale-ai-sued-wage-theft-19976761.php

I had tried sub-contracting for one of their contractors, and the whole experience left a bad taste in my mouth too.

 

[ygolo]

Harvard Is Releasing a Massive Free AI Training Dataset Funded by OpenAI and Microsoft

The project’s leader says that allowing everyone to access the collection of public-domain books will help “level the playing field” in the AI industry.

www.wired.com

I've said before, what these LLMs do amounts to fancy compression.

[2309.10668] Language Modeling Is Compression

So the data is everything, basically. Now that we know at least one major source clearly, we can scrutinize it fully and publicly.

You could state that what the human brain does is also fancy compression, but that link is much more speculative. You could maybe see the analogy in papers like this:

Neuroimaging evidence supporting a dual-network architecture for the control of visuospatial attention in the human brain: a mini review - PMC

Neuroimaging studies conducted in the last three decades have distinguished two frontoparietal networks responsible for the control of visuospatial attention. The present review summarizes recent findings on the neurophysiological mechanisms ...

pmc.ncbi.nlm.nih.gov

But, in the machine learning case, the argument seems iron clad. In the human brain case, the argument is currently a stretch.

 

[ygolo]

How Software Development is Changing Forever, and How You'll Need to Change With It

Discover thousands of collaborative articles on 2500+ skills

Discover 100 collaborative articles on domains such as Marketing, Public Administration, and Healthcare. Our expertly curated collection combines AI-generated content with insights and advice from industry experts, providing you with unique perspectives and up-to-date information on many skills...

www.linkedin.com

As many have said, English is the hottest new programming language. Many other natural languages will be as well.

 

[ygolo]

OpenAI Unveils o3 System That Reasons Through Math, Science Problems

The artificial intelligence start-up said the new system, OpenAI o3, outperformed leading A.I. technologies on tests that rate skills in math, science, coding and logic.

www.nytimes.com

 

[ygolo]

Accelerating LLM Inference on NVIDIA GPUs with ReDrafter

Accelerating LLM inference is an important ML research problem, as auto-regressive token generation is computationally expensive and…

machinelearning.apple.com

 

[ygolo]

MarkItDown: Microsoft’s Game-Changing Tool That Turns Any Document into Clean Markdown

A deep dive into Microsoft’s new Python library that seamlessly converts PDFs, Office files, and more into clean, version-control-friendly…

levelup.gitconnected.com

 

[Julius_Van_Der_Beak]

MarkItDown: Microsoft’s Game-Changing Tool That Turns Any Document into Clean Markdown

A deep dive into Microsoft’s new Python library that seamlessly converts PDFs, Office files, and more into clean, version-control-friendly…

levelup.gitconnected.com

If this could be used to create documents derived from Word or Google Docs that appear the same regardless of their source, that could be a game changer. PDFs are unfortunately not so consistent, IIRC. It looks easy to use, too. It's also much more involved than I was expected, as it includes OCR and speech-to-text.

I think the really big thing could be audio files. What do you think of the potential to use this to get the speech off of the entire library of YouTube videos? That would generate a lot of data.

 

[ygolo]

If this could be used to create documents derived from Word or Google Docs that appear the same regardless of their source, that could be a game changer. PDFs are unfortunately not so consistent, IIRC. It looks easy to use, too. It's also much more involved than I was expected, as it includes OCR and speech-to-text.

I think the really big thing could be audio files. What do you think of the potential to use this to get the speech off of the entire library of YouTube videos? That would generate a lot of data.

YouTube often provides transcripts themselves. Gemini Flash can do a lot of transcribing for cheap. OpenAI Whisper is an inexpensive API also, and has an open source version.

For the more DYI things, you are responsible for looking at the rights of what you are transcribing.

 

[Julius_Van_Der_Beak]

YouTube often provides transcripts themselves. Gemini Flash can do a lot of transcribing for cheap. OpenAI Whisper is an inexpensive API also, and has an open source version.

For the more DYI things, you are responsible for looking at the rights of what you are transcribing.

That's true. Is it done through software with YouTube? I imagine it would have to be given the sheer volume.