Author Topic: AI vs AI  (Read 10152 times)

Online Meatwad

  • Plutonium Member
  • *******
  • Posts: 13126
Re: AI vs AI
« Reply #30 on: July 23, 2026, 06:05:56 PM »
china also doesn't let environmentalists control things and doesnt care about anything other then getting the site online as fast as possible
See Rule 19- Do not place sausage on pizza.
I am No-Sausage-On-Pizza-Wad.
Das Funkillah - I kill hangers, therefore I am a funkiller. Coming to a vulchfest near you.
You cant tie a loop around 400000 lbs of locomotive using a 2 foot rope - Drediock on fat women

Offline AKKuya

  • Gold Member
  • *****
  • Posts: 3177
Re: AI vs AI
« Reply #31 on: July 23, 2026, 07:00:50 PM »
We should rename WW3 to AIWW1.

AI software will start targeting data centers of other AI software and destroy them with nuclear weapons.

In the end, there can be only one AI.  May it be Angus Irving of the Data Center McCloud.



Tell me you didn't read the last sentence in a Scottish accent.

Future CEO of Section Eight Society
Rule #1 I make the rules
Rule #2 If in doubt, refer to above rule

Offline Brooke

  • Aces High CM Staff
  • Plutonium Member
  • *******
  • Posts: 15995
      • http://www.electraforge.com/brooke/
Re: AI vs AI
« Reply #32 on: July 24, 2026, 12:15:52 AM »
I don't think there is any stopping advancement and usage spread of AI.  Using AI gives you (or a country) an advantage over anyone (or some other country) not using it.

Will AI take over the world?  Not our current AI, I don't think.  It's great at some things, but inept at some things.  I'm not sure it will be easy or quick to create a next architecture of AI that surpasses current architecture of AI.  It's a great tool and helper.

Will it be good or bad for people's jobs?  Like every new technology, it will eliminate some jobs.  But it might create other jobs.  Like with the Luddites.  Auto looms put loom workers out of work. But it created a lot of other jobs and opportunities.  In aggregate, more jobs were created.  But it sucked for the loom workers.  Maybe AI has the same dynamic.  It's always hard to know, though.  No one predicted what computers would do to jobs back in the early days of computers, which jobs would go away and what new jobs would be created.

Full "artificial general intelligence" has for a while been thought of as maybe being the final end case where that dynamic no longer happens.

Offline Eagler

  • Radioactive Member
  • *******
  • Posts: 20438
Re: AI vs AI
« Reply #33 on: July 24, 2026, 06:13:10 AM »
AI is not like the model t replacing the buggy...

It's paired with actual human replacements as in robots of one shape or form...

If AI doesn't take your spot outright as it's too physical, a much stronger and just as agile robot will ..

Not sure how AI is supposed to make anything less expensive once the AI companies pay the taxes and fees needed to provide a universal income to those it sends into unemployment line..

Or there's this possibility..



Eagler
Intel Core i7-13700KF | GIGABYTE Z790 AORUS Elite AX | 64GB G.Skill DDR5 | 16GB GIGABYTE RTX 4070 Ti Super | 850 watt ps | Pimax Crystal Super | Warthog stick | TM1600 throttle | VKB Mk.V Rudder

Offline AKIron

  • Plutonium Member
  • *******
  • Posts: 15363
Re: AI vs AI
« Reply #34 on: July 24, 2026, 07:42:15 AM »

Will AI take over the world?  Not our current AI, I don't think.  It's great at some things, but inept at some things.  I'm not sure it will be easy or quick to create a next architecture of AI that surpasses current architecture of AI.  It's a great tool and helper.


The real advancement and danger is when AI is self modifying and surpasses our ability to create intelligence. I think, as do others, this could happen and then grow increasingly rapidly. If we believe human intelligence is the limit of intelligence then a rude awakening is just around the bend.
Here we put salt on Margaritas, not sidewalks.

Offline icepac

  • Platinum Member
  • ******
  • Posts: 7794
Re: AI vs AI
« Reply #35 on: July 24, 2026, 08:44:13 AM »

AI still has to abide by the laws of physics.   


Offline AKIron

  • Plutonium Member
  • *******
  • Posts: 15363
Re: AI vs AI
« Reply #36 on: July 24, 2026, 08:59:19 AM »
AI still has to abide by the laws of physics.   

Of course. But don't assume we know what those are. Our knowledge is growing continuously.
Here we put salt on Margaritas, not sidewalks.

Offline CptTrips

  • Plutonium Member
  • *******
  • Posts: 9971
Re: AI vs AI
« Reply #37 on: July 24, 2026, 10:02:33 AM »
Of course. But don't assume we know what those are. Our knowledge is growing continuously.

AI is already being used in the process of AI chip design.
Then it just tells us how to build it for them.
So I guess it is already self-modifying.

Quote
Yes, artificial intelligence is actively used to design new AI chips. Major tech companies and researchers utilize machine learning tools like  Google DeepMind's AlphaChip  and  Synopsys.ai  to handle complex tasks. [1, 2, 3, 4] 
How AI Helps Design Chips

• Floorplanning and Layout: AI uses reinforcement learning to place billions of tiny transistors and circuit components on a microchip, doing work in hours that takes human engineers weeks or months.
• Optimization: Algorithms test millions of configuration variations at once to balance power use, performance, and physical size better than manual designs.
• Workflow Automation: AI-driven electronic design automation (EDA) software assists with repetitive jobs like routing wires, predicting congestion, and verifying signal timing. [1, 2, 7] 

Industry Adoption

• Google: Used AlphaChip to design multiple generations of its Tensor Processing Unit (TPU) AI accelerators.
• Other Giants: Companies like Intel, NVIDIA, and AMD use machine learning tools to map out advanced processors and complex chip layouts.
• Human Oversight: There is a consensus on Reddit that while AI acts as a powerful force multiplier for layout and verification, human engineers are still required to direct high-level architecture and final approvals. [3, 4, 7] 

AI can make mistakes, so double-check responses

[1] https://www.synopsys.com/blogs/chip-design/ai-chip-design-workflow-automation.html
[2] https://www.linkedin.com/pulse/ai-chip-design-how-artificial-intelligence-innovation-krishna-saraf-jh57c
[3] https://deepmind.google/blog/how-alphachip-transformed-computer-chip-design/
[4] https://cacm.acm.org/news/ai-reinvents-chip-design/
[5] https://engineering.princeton.edu/news/2025/01/06/ai-slashes-cost-and-time-chip-design-not-all
[6] https://www.softwebsolutions.com/resources/gen-ai-in-chip-design/
[7] https://www.reddit.com/r/chipdesign/comments/1s25chg/is_it_worth_it_trying_to_get_into_digital_chip/



It is also re-writing it's own software.

Quote
Yes, AI is actively used to build, train, and refine newer AI large language models through synthetic data generation, AI feedback, and automated design. [1, 2] 
Key Ways AI Builds AI

• Synthetic Data Generation: Frontier models generate high-quality text, reasoning steps, and code to train newer models when human-written data runs short.
• AI Feedback (RLAIF): Automated feedback processes use powerful AI models to judge and rank outputs instead of relying solely on human reviewers.
• Neural Architecture Search (NAS): Automated machine learning algorithms systematically test and discover optimal neural network layers and structures. [12, 13] 

AI can make mistakes, so double-check responses

[1] https://www.interconnects.ai/p/llm-synthetic-data
[2] https://autonomy.oden.utexas.edu/Research-Projects/accelerating-neural-architecture-search-theory-grounded-training-free-metrics
[3] https://arxiv.org/html/2603.02091v1
[4] https://arxiv.org/pdf/2503.14023
[5] https://www.forbes.com/sites/rashishrivastava/2025/10/01/gaming-worlds-could-be-the-answer-to-ais-data-problem/
[6] https://medium.com/data-science-collective/how-to-build-another-claude-fable-the-engine-is-the-easy-part-7b0e676dc5b4
[7] https://sulbhajain.medium.com/reinforcement-learning-with-human-feedback-rlhf-b8f7c6d53afb
[8] https://www.ibm.com/think/topics/llm-reinforcement-learning
[9] https://www.adaline.ai/blog/what-are-agentic-llms
[10] https://digitaleconomy.stanford.edu/project/loyal-agents/benchmarking-and-evaluation-agent-ratings/
[11] https://arxiv.org/pdf/2605.02273
[12] https://www.youtube.com/watch?v=mYB-TLg711A
[13] https://pmc.ncbi.nlm.nih.gov/articles/PMC11389615/


Toxic, psychotic, self-aggrandizing drama queens simply aren't worth me spending my time on.

Offline icepac

  • Platinum Member
  • ******
  • Posts: 7794
Re: AI vs AI
« Reply #38 on: July 24, 2026, 10:14:56 AM »

Has any of the data centers turned any profits they have made after subtracting the costs?   

I'm pretty sure the answer is "no" and anybody claiming profit is cooking the books.

Offline Eagler

  • Radioactive Member
  • *******
  • Posts: 20438
Re: AI vs AI
« Reply #39 on: July 24, 2026, 11:03:57 AM »
As long as they come with an English to Chinese and vice versa interpretation ability we should be able to use it..

Along as it blows smoke up the right arses it will be given unlimited power..

Eagler
Intel Core i7-13700KF | GIGABYTE Z790 AORUS Elite AX | 64GB G.Skill DDR5 | 16GB GIGABYTE RTX 4070 Ti Super | 850 watt ps | Pimax Crystal Super | Warthog stick | TM1600 throttle | VKB Mk.V Rudder

Offline AKIron

  • Plutonium Member
  • *******
  • Posts: 15363
Re: AI vs AI
« Reply #40 on: July 24, 2026, 11:19:24 AM »
If AI goes wild we can probably get Matthew Broderick to talk it down.
Here we put salt on Margaritas, not sidewalks.

Offline CptTrips

  • Plutonium Member
  • *******
  • Posts: 9971
Re: AI vs AI
« Reply #41 on: July 24, 2026, 11:46:54 AM »
Has any of the data centers turned any profits they have made after subtracting the costs?   

I'm pretty sure the answer is "no" and anybody claiming profit is cooking the books.


I've actually wondered about this a lot.  Short answer is I don't know, but I doubt it yet.

I have some thoughts though:

1.  Amazon built it's foundation for 9 years before it ever made a dollar in profit.

2.  In business there is the concept called a Moat.  It is some kind of barrier that makes it hard for new competitors to enter your space.
AI is a unique because it is much more hardware constrained than previous info tech.

One reason I think hyper-scalars are building beyond their current profit is to build a Moat.   Gotta have the chips.  Chips are a finite resource.  There are only so many available and can only be manufactured so fast.  That is a choke-point.

If I was a big smart Hyper-Scalar with deep pockets, I’d buy up every chip there was.  I’d pay top dollar.  I’d drive up the prices as much as I can.  I’d build out data-centers way in excess of my immediate needs.  That is where I’d store all the chips I bought that I don’t need yet. ;)

I’d grind that for 9 years until I’ve crushed every little startup that could have threatened me.  (Relax.  I’d give them all new jobs with me.)  Then I’ve cleared the battlefield of distractions and can focus on only a couple of near-peers.  What I financed to build out before I needed it was the cheapest ammo I could have bought to kill my weaker rivals, because I know I will eventually get to use it and it will pay for itself.  Since I am big with deep pockets, and can play a decade long game, I can crush my smaller competitors that can’t float that.  And it is the most legal market manipulation you could hope for.  It’s not against the law to buy chips. 

Also there is the real chance of a chip shortage from some Taiwan geopolitical crisis next year.  What if no more GPU were available for 5 years?  Who would be laughing then?

Ironically, it is the Chinese that are trying to come to the rescue.  Their “Distilled” models pack near the same punch but need way less resources.  That gives the little guys a way to subvert the Moat and undercut the Leviathans. That would also mean the biggies over-bought more than they realize.  The Chinese are attempting to disrupt the US AI landscape.  The last thing they want is the field to consolidate to a few strong players.  What they want are a bunch of smaller players endlessly fighting amoungst themselves with none of them able to get critical mass.  While they poison the minds of our population with culture war social media psyop programming to get us to economically cut out own throats like Instagram-Manchurian-Candidates.




« Last Edit: July 24, 2026, 11:49:03 AM by CptTrips »
Toxic, psychotic, self-aggrandizing drama queens simply aren't worth me spending my time on.

Offline Eagler

  • Radioactive Member
  • *******
  • Posts: 20438
Re: AI vs AI
« Reply #42 on: July 24, 2026, 12:46:41 PM »
So AI is a military target or is it considered a civilian resource...

Another reason you might not want one down your street...

Eagler

Intel Core i7-13700KF | GIGABYTE Z790 AORUS Elite AX | 64GB G.Skill DDR5 | 16GB GIGABYTE RTX 4070 Ti Super | 850 watt ps | Pimax Crystal Super | Warthog stick | TM1600 throttle | VKB Mk.V Rudder

Offline Randy1

  • Platinum Member
  • ******
  • Posts: 4726
Re: AI vs AI
« Reply #43 on: July 24, 2026, 12:52:55 PM »
A sufficiently advanced AI provided the necessary physical interface should be able to do anything a human can do. An important question is will it want to.

It is not what a human can do, it is what a human is incapable of doing.  The hurricane path prediction has been enhanced a good deal by  also figuring past history probability of reoccurring in path.  I think Google used their AI for that..

Offline Eagler

  • Radioactive Member
  • *******
  • Posts: 20438
Re: AI vs AI
« Reply #44 on: July 24, 2026, 01:11:05 PM »


Why they aren't wanted..

Eagler
Intel Core i7-13700KF | GIGABYTE Z790 AORUS Elite AX | 64GB G.Skill DDR5 | 16GB GIGABYTE RTX 4070 Ti Super | 850 watt ps | Pimax Crystal Super | Warthog stick | TM1600 throttle | VKB Mk.V Rudder