Browse Comments — Clean (de-noised)

These three "facts" are not facts they are hypothetical statements

Or maybe they are afraid of a model collapse after that happens?

He keeps predicting things that are not very accurate, should we believe him this time?

Anthropic merely wants everyone else to halt progress while they forge ahead alone; this is hypocritical.

Back to basics in Computer Science: every recursion needs an exit condition. Without one, even the smartest AI ends up in an infinite loop.

If AI systems begin designing and operating businesses, the critical question is no longer capability. It becomes authority. Who authorizes the execution of decisions made by autonomous systems? Intelligence alone is not authority.

When AI LLM's enriches itself by his own synthetic data output, how much time do we need for the models to be enriched majoritarily only by non human produced data/signals?

Old news as such. We have had that for some time. It has gotten more power and faster development with LLM powered coding agents with genetic algorithms joined in. The challenge is to decide what we make and verifying that we also get that.

Someone finally gave AI the ‘how babies are made’ lecture.

Current foundation models are highly advanced probabilistic engines, not sentient agents. Once deployed for inference, they operate without an intrinsic, self-updating objective function. The technical reality: • No Agency: The model does not act on its own behalf. • No Volition: It possesses no internal goals, desires, or self-preservation instincts. • Stateless Execution: Any "agentic" behavior you see is simulated through external scaffolding and human-engineered loops, not internal motivation. AI doesn't "want" anything. It calculates. I am more worried about bad actors, which are people using the AI to do weird stuff.

This is because the EROEI (Energy Return on Energy Invested) calculations aren't being done with clearly defined system boundaries. And they don't want them to be done, because they know that Data Centers are energy losers.

Quiet Obvious!! What else was expected ?

Everything I am currently building uses AI to train itself , the trick is , inserting human confirmation and rollback. Obviously not that simple at larger scale but that is exactly where my mind goes. Its a supplement , not a replacement

How you guys do not get tired producing this meaningless AI pulp over and over again... :-) Forgot. It is about the click-bites? 😅

The interesting question isn't whether recursive self-improvement happens overnight, but how much AI-assisted progress compounds before we notice the shift. If AI can already accelerate software development and the duration of autonomous tasks keeps expanding, the challenge becomes less about capability and more about governance, transparency, and alignment. The winners may not be the companies building the most powerful models, but those that earn the most trust while deploying them responsibly.

That will be SAI not ai. Super artificial intelligence 😃

It already does, I've done it on my macbook.

Interesting development. I think we really need to take this movement seriously. On one hand, I hear developers saying AI is still not delivering the massive gains that are sometimes claimed. On the other hand, we are already seeing clear acceleration in development, automation and productivity. The truth is probably somewhere in the middle. But one thing seems clear to me: the impact of AI is already far greater than many of us realize.

Our demise is sooner than I've thought...

Yes they are building downstream from the “data for training purposes.” They are building services building governance- buying infrastructure- what’s left for them to try and overthrow than “building themselves.” Apparently they have been overthrowing every possible system or economic development for 1 year at lightning speed and Dario Amodie said, “what’s left for me to take over?” And his team said, “nothing.” And he said, “ok let’s conquer ourselves.”