This rush to do AI in a world where you haven't even modernized your application reminds me a little bit of that lift-and-shift that happened in the cloud.
大多数人认为AI应用应该优先采用最新技术快速实现，但作者将其比作云计算早期的'简单迁移'模式，认为这是一种可能导致资源浪费的短视行为。这与当前AI领域的快速采用主流观点相悖，暗示企业在AI应用上可能需要更加谨慎的基础架构规划。
AI systems are not engineered the way a bridge or an airplane is engineered. We understand an airplane because we designed every part of it and we understand the physics that act on it. AI models are not like that. They are grown, on a structure roughly modeled after the brain, on an enormous inheritance of human thought and speech.
这段比喻极其生动地解释了AI与传统工程技术的根本区别，将AI描述为'生长'而非'建造'的系统，强调了其复杂性和不可预测性。这种表述既科学又富有诗意，帮助非专业人士理解AI的特殊性。
Instead of just answering a user’s questions, the way a chatbot does, agents can take a human user’s instructions and act on them
AI代理的能力描述可能存在偏见，因为它暗示AI能够像人类一样行动，而实际上可能缺乏人类的判断力和道德考量。
Stronger models hallucinate less, so they can't see the problem in any side of the spectrum: the hallucination side of small models, and the real understanding side of Mythos.
这一观察极具反直觉性：更强的模型反而更难发现某些漏洞，因为它们减少幻觉的同时也失去了对问题的'直觉理解'。这暗示AI安全研究可能需要不同能力层次的模型组合，而非简单地追求更大更强的模型。
Four researchers and software engineers estimated that a skilled human engineer would take 2 to 17 weeks to reimplement gotree, as AI successfully did in this work.
这一对比数据极具启发性，它量化了AI在特定任务上相对于人类的时间优势。这种时间压缩效应可能重塑软件开发流程，但也引发了关于AI能力与人类创造力本质差异的深层思考。
Older models were more prone to submitting prematurely, even when test cases weren't passing.
这一观察揭示了不同AI模型版本之间在任务坚持性上的显著差异。早期模型更容易过早提交不完整的解决方案，而最新模型表现出更强的任务坚持性和工程判断力。这种差异可能反映了AI在自我评估和任务管理能力上的进化。
The difference between AI and, say, looms, is that this has been broadcast to the entire globe, and it has been treated in a sort of self-conscious way
令人惊讶的是：文章指出AI与历史上其他技术变革(如织布机)的关键区别在于AI的全球广播性质和行业领袖的自我意识宣传。这种透明度反而加剧了公众的不安，因为AI领袖们不断谈论他们知道会引发问题的技术，这在历史上是前所未有的。
In Washington, the AI policy discourse is sometimes framed as a 'race to AGI.' In contrast, in Beijing, the AI discourse is less abstract and focuses on economic and industrial applications that can support Beijing's overall economic objectives.
令人惊讶的是：中美对AI的战略定位存在根本差异——美国聚焦于通用人工智能(AGI)的竞赛，而中国则更注重经济和工业应用。这种差异反映了两国的技术哲学和治理模式，也解释了为什么中国在有限计算资源下仍能发展出更具实用性的AI应用。
We're all worried about, you know, immigration of the other countries next door uh taking labor jobs. What happens when AI immigrants come in and take all of the cognitive labor? If you're worried about immigration, you should be way more worried about AI.
for - forte - comparison - foreign immigrants Vs AI immigrants - sorry about foreign immigrants - should be more worried about AI immigrants
basically absent or very seldom present in current AI systems
for - comparison - biological vs AI systems
for - natural language acquisition - Automatic Language Growth - ALG - youtube - interview - David Long - Automatic Language Growth - from - youtube - The Language School that Teaches Adults like Babies - https://hyp.is/Ls_IbCpbEfCEqEfjBlJ8hw/www.youtube.com/watch?v=984rkMbvp-w
summary - The key takeaway is that even as adults, we have retained our innate language learning skill which requires simply treating a new language as a new, novel experience that we can apprehend naturally simply by experiencing it like the way we did when we were exposed to our first, native language - We didn't know what a "language" was theoretically when we were infants, but we simply fell into the experience and played with the experiences and our primary caretakers guided us - We didn't know grammar and rules of language, we just learned innately
we are using set theory so a certain piece of reference text is part of my collection or it's not if it's part of my collection somewhere in my fingerprint is a corresponding dot for it yeah so there is a very clear direct link from the root data to the actual representation and the position that dot has versus all the other dots so the the topology of that space geometry if you want of that patterns that you get that contains the knowledge of the world which i'm using the language of yeah so that basically and that is super easy to compute for um for for a computer i don't even need a gpu
for - comparison - cortical io / semantic folding vs standard AI - no GPU required
for example our standard english language model is trained with something like maybe 100 gigabytes or so of text um that gives it a strength as if you would throw bird at it with the google corpus so the other thing is of course uh a small corpus like that is computed in two hours or three hours on a on a laptop yeah so that's the other thing uh by the way i didn't mention our fingerprints are actually a boolean so when we when we train as i said we are not using floating points
for - comparison - cortical io vs normal AI - training dataset size and time
human beings don't do that we understand that the chair is not a specifically shaped object but something you consider and once you understood that concept that principle you see chairs everywhere you can create completely new chairs
for - comparison - human vs artificial intelligence
question - comparison - human vs artificial intelligence - Can't an AI also consider things we sit on to then generalize their classifcation algorithm?
the brain is Islam Islam is it is lousy and it is selfish and still it is working yeah look around you working brains wherever you look and the reason for this is that we totally think differently than any kind of digital and computer system you know of and many Engineers from the AI field haven't figured out that massive difference that massive difference yet
for - comparison - brain vs machine intelligence
comparison - brain vs machine intelligence - the brain is inferior to machine in many ways - many times slower - much less accurate - network of neurons is mostly isolated in its own local environment, not connected to a global network like the internet - Yet, it is able to perform extraordinary things in spite of that - It is able to create meaning out of sensory inputs - Can we really say that a machine can do this?
you're going to have like 100 million more AI research and they're going to be working at 100 times what 00:27:31 you are
for - stats - comparison of cognitive powers - AGI AI agents vs human researcher
stats - comparison of cognitive powers - AGI AI agents vs human researcher - 100 million AGI AI researchers - each AGI AI researcher is 100x more efficient that its equivalent human AI researcher - total productivity increase = 100 million x 100 = 10 billion human AI researchers! Wow!
if I met a robot that looked very much like a beautiful girl and everything went fine together with her and me but
for - comparison - human vs AI robot - Denis Noble
there's this broader issue of of being able to get inside other people's heads as we're driving down the road all the time we're looking at other 00:48:05 people and because we have very advanced theories of mind
I would submit that were we to find ways of engineering our quote-unquote ape brains um what would all what what would be very likely to happen would not be um 00:35:57 some some sort of putative human better equipped to deal with the complex world that we have it would instead be something more like um a cartoon very much very very much a 00:36:10 repeat of what we've had with the pill