good source on how LLMs can’t really reason

I disagree. It is able to come up with something that “sounds right” based on what it’s been trained on, and it can be transferred to specific domains. LLMs are mimicking “fast” human thinking. It’s like the world’s smartest BS-ing alien trying to say things at a cocktail party to convince the guests that it is human despite not knowing anything and just repeating sounds that seem to please the audience. Humans have similar “off-the-cuff” automatic behaviors. However, LLMs seem to have no capability to mimick human “slow” critical thinking, and there is no real internal representation of the world to speak of. There is nothing I know about on the AI research roadmap to overcome these limitations because right now we’re taking neural networks and throwing them into GPUs until investor money comes raining from the sky.

Even if these technical limitations were miraculously overcome within 10 years, the biggest problem by far is that the energy consumption is basically uneconomical. A full-fledged human AGI takes tens of watts, whereas a silicon AGI takes orders and orders of magnitude more.

Yes, it is well-known in tech land that OpenAI’s entire modus operandi is to just throw everything at the wall to try to build AGI. The company has a cult-like reputation filled with AI zealots. It’s an incredibly dangerous organization. The only good news is that they almost certainly won’t succeed. The current tech seems to be basically half of an AGI, and an incredibly expensive one, at that.

here’s the paper in the ‘70s that set off this quest for the holy grail. many have tried but we’ll see if this one sticks

memristors are a pretty old idea. they mathematically complete electronics and solve some fundamental computing problems. it will be a long road to take advantage of this though because memristors require a different architecture than the Von Neumann one that basically every current computing device has.

Great read - I can’t believe how quickly AlphaFold is enabling science like this

Play stupid games…

This will be an incredible development when they pull it off. It seems murky to me how semiconductor shrinks will be pulled of after the next few, but this could really change things.

I’m sure you must be right that a civilization with 15% of Earth’s population and thousands of years of history has no “long term potential,” whereas the pieces of it imperialists have tried (and failed) to break off, like Hong Kong and Tibet, somehow do have “long term potential.” The only thing inhibiting “long term potential” here is your hubris and myopic worldview.

This is an even bigger L for the US because China is blocking gallium exports and they have 80% of the world production lmao

I saw some of the news that you mentioned on graphene as well. that would be incredible!

This is exciting to see. The semiconductor industry is only possible through a truly global supply chain that is of unfathomable complexity. China is probably the only country on the planet capable of something like semiconductor autarky. With China being locked out from the high-end stuff by the west, it’s going to be a matter of when, not if, China winds up achieving semiconductor superiority, and whether that will be before or after reunification with Taiwan. After all, Taiwan represents just 2% of China’s power, based on population.

I’m familiar with the industry and the ASML sanctions are probably going to hurt the most based on the CPC bragging about SMEE’s ability to produce 28nm DUV machines domestically before such a machine has actually been proven to exist. China has been stockpiling the ASML DUV machines for some time now in anticipation of this and making enough improvements to them to make 7 nm production feasible (the limit of DUV), which will buy them some time for SMEE to play catchup. Catching up though, is going to be an absolutely herculean task for China to have domestically-produced EUV. China will not be able to compete on the cutting-edge of technology until this is achieved. With 7nm DUV, they should be able to produce sophisticated enough pieces to eventually build EUV light sources, and the export controls on rare-earth metals from China will make it much more expensive to produce the high-tech necessary for ASML to ship more EUVs.

Still though, EUV took decades of basic R&D in the west with armies of PhDs at the national labs tackling the problem, and another couple decades to commercialize the technology. It was essentially one of the last basic research efforts that the west actually pursued. Until now, the Chinese semiconductor industry has been focused on scaling production instead of pioneering high-end process nodes. I’m honestly expecting some movie-worthy espionage schemes will be hatched at TSMC to accelerate domestic development of this technology. It’s gonna be a banger to watch this technology war play out over the next 10 years, or maybe it will be quicker, we’ll see.

The west’s brain-dead investment strategies have basically left technological progression of semiconductors after basic research ran out of steam to Nvidia, Apple and other western firms working with ARM, TSMC, and ASML to advance SOTA with all state support coming from Taiwan. I’m sure China and her people are aware of the situation, and the sheer speed of technological development we will see come out of China as a result of this national focus will be mind-blowing.