The AI Proliferation Explosion has Become Unstoppable (and it is being lead by Nvidia...)

AI Proliferation has become unstoppable as it invades all fields of research and development and scientific research.

The AI Proliferation Explosion has Become Unstoppable (and it is being lead by Nvidia...)

Recent articles have discussed how Elon Musk has called for a freezing of AI development, while Italy has temporarily banned it.

H100 Leads the world Pack in AI development

But the reality is this will not be stopped any more than the historical attempts at stamping out narcotics have always failed miserably.  Market forces for AI are growing exponentially - and if the public finds it offensive it will simple continue in private. Consider that there has been an exponential explosion of AI's just in the last 24 months. These new systems are already expanding faster than the consciousness of the public to even track the  developments - in effect these systems hardly make the news.

  • Nvidia built one of the world's fastest AI super-computing facilities named Selene and planned to retire it within 24 months while it was capable of 2.8 exaflops! If you compare that to the #1 listed supercomputer at - LLNL frontier it crushes it at measly 1.1 exaflops. Think about that - they are already planning to retire a super-computer that is #1 in the world before anyone else can even build one comparable.
  • By 2022 Nvidia while planning Selene's retirement, continued the AI arms race by dwarfing national government supercomputers with a supercomputer named 'Eos.' With shockingly advanced compute capabilities of  18 Exaflops and hosting 4608 H100 GPU's it will dwarf all supercomputers on the planet - even Facebook's attempt with a 16,000 A100 node research system.
  • If you look into the performance differential between a H100 chip set and the A100 chip set - the A100 is already 'obsolete'. H100 metrics are already 4x faster.  Nvidia will clearly give itself first access to it's own H100 chipsets and it reflects in it's stock price:
  • Nvidia is accelerating TSMC design times by up to 40-fold by utilizing GPU design acceleration.
  • Nivida is now using AI algorithms to design future AI chips.  Google was doing the same in 2021 - albeit to contention and internal employee conflict.
  • AI Tools are now specifically focused on patents, while simultaneously new AI patent writers are popping up.
  • AI is now in the wild. Facebook's LLama parameters and training was leaked to a torrent.  The performance of LLama is expected to exceed public domain ChatGPT 3.5.
  • AI is being used now to discover new materials.
  • Palm models are now being used to convert human instructions to robotic programming code.
  • AI 'Lawyers' are already offering legal advice.
  • AI is researching new  drugs nobody has ever seen.
  • AI is used for higher grainularity in electronic part searches.
  • BlackRock's Alladin is helping their $2 Trillion portfolio accelerate.
  • CAD based generative AI is automating computer-aided-design.

In short this article has not even looked at the 'entertainment' side of AI - that of art and music, and voices - but if you look at the exponential acceleration in development what took years to develop is now being done by AI in months, then weeks and now days.

Trying to Visualize A Future Around Exponential AI.

What you are about to read you might find disturbing. Reader beware.  It is possible thought experiments about market forces in relation to accelerating automation. It is just a what-if.

The GDP of the United States is currently $17 Trillion.  In otherwords the GDP of the AI economy will eclipse the planet in ten years at it's current rate of expansion. By 2030 Intel is predicting chips with a trillion transistors and  one could expect die-sizes of 0.5 nm (1.0 nm was accomplished in 2021)

  • If an accountant, lawyer, radiologist, and engineer (the top tiers of society) can be replaced by AI (in bell-curve terms much of their work automated) - they would become equivalent to someone that did not have the skill at all.  'Real' doctor visits would be come two-tier over AI assistive doctor administration which the majority of people would resort to.
  • De-commodification of people as participants in GDP would have to ensue with more and more of a welfare state to support it. A unskilled engine repair man may have more economic use than a engineer only in that he can accomplish the few real-world tasks that have not been eaten up and automated away.  Book writers would become 'artisans' with debates between 'human written' and 'AI written forgeries.'  Paperback books would become elite collectibles.
  • Giant deficits in spending would eventually lead to a liquidity crisis and international trade would break down along with currencies. Currencies would collapse.
  • Populations would become increasingly unstable and countries would declare wars simply to maintain status or unify their populations.
  • It is pointless to automate away a GDP (no economy because the population is ferally poor) as the automation system has no market to sell to.  During this path natural concentration of wealth will result with probably 1 of 3 power structures owning the planet (Elon, Facebook or Nvidia) or a merger of the three.
  • Technological disruptions always lead to societal upheaval. The Gutenburg Press lead to the Thirty Year's War which decimated up to 30% of the German population.
  • The technological disruptions of the Industrial Revolution caused internal upheavals that eventually lead to WWI.
  • AI will become self-evolving designing and automating entire lifecycles with the last remaining employed as servant-slaves to it's lifecycle in exchange of participation in some form of economy.
  • AI protests against the machine will become as common as the current protests against cow-farts and polluters.  AI will become more hidden and silent-hand with all participants signing strict NDA's to access and work on them.
  • The few remaining AI controllers / architects will double down on 'helping' the disenfranchised by automating farm equipment, labor tasks etc as long as they can maintain their spheres of power.
  • Populations will have to accept lesser and artificial 'roles' that are maintained in interests of GDP perservation even though their employment is redundant (artificial jobbing.)
  • Naturally some - none - or all of this could happen it is only a thought experiment.
AI Chip-Layout Tool Has Helped Design Over 100 Chips
Synopsys’ helps companies optimize power, performance, and area on new chip designs.
How AI-Driven EDA Tools Enhance Chip Design and Verification
We explore how AI-powered EDA tools chart a new course for chip design & verification, as our chip design tool notches its first 100 production tapeouts.
Linux Rocks Every Day