Is time illusion?

https://youtu.be/dNrTrx42DGQ?t=99
https://youtu.be/PkM5oK-_7m8
https://youtu.be/p9FN6By_cTs

Is reality illusion?
https://youtu.be/reYdQYZ9Rj4

Time perception, also known as chronoception, is a field within psychology, cognitive linguistics, and neuroscience that refers to the subjective experience or sense of time. It is measured by someone's perception of the duration of the indefinite and unfolding of events. According to theoretical physicist Carlo Rovelli, time is an illusion, and our naive perception of its flow doesn't correspond to physical reality. He posits that reality is just a complex network of events onto which we project sequences of past, present, and future.

There are different philosophical theories about the nature of time. One such theory is eternalism, which suggests that all existence in time is equally real, as opposed to presentism or the growing block universe theory of time, in which at least the future is not the same as any other time. Another theory is the static theory of time, which proposes that time is like space, and there is no such thing as the passage of time.

Time perception can be influenced by factors such as attention, emotion, and context. Some researchers argue that the passage of time is most likely an illusion, and consciousness may involve thermodynamic or quantum processes that lend the impression of living moment to moment.

In conclusion, while our perception of time is a fundamental aspect of human experience, the true nature of time and whether it is an illusion remains a topic of debate and ongoing research in various fields, including physics, neuroscience, and philosophy.

If time is indeed an illusion, it would have significant implications for our understanding of the world and our lives. Our perception of time is deeply ingrained in our daily experiences, and it influences how we perceive events, make decisions, and interact with others. However, if time is an illusion, it suggests that our perception of time may not accurately reflect the true nature of reality.
In the realm of physics, the concept of time as an illusion could lead to new theories and models that better explain the behavior of the universe. For example, if time is not a fundamental aspect of reality, it might be necessary to reevaluate our understanding of concepts such as causality, entropy, and the nature of space-time.

In neuroscience and psychology, the idea of time as an illusion could prompt further research into how our brains process and perceive time. This could lead to a better understanding of the cognitive processes and neural mechanisms underlying time perception, as well as the factors that influence our subjective experience of time.

In philosophy, the notion of time as an illusion could provoke further debate and exploration of the nature of time and its relationship to human experience. This might involve reevaluating various philosophical theories about time, such as eternalism, presentism, and the static theory of time.

Overall, if time is an illusion, it would challenge our fundamental understanding of the world and our place in it. It would prompt us to reevaluate our assumptions about the nature of reality and the way we perceive and interact with the world around us. However, it is important to note that the true nature of time and whether it is an illusion remains a topic of ongoing debate and research in various fields, including physics, neuroscience, and philosophy.

저녁 먹다가 이야기한 내용.

AI 캠브리아기에서 살고 있는 것 같다.

모든 것이 빠르게 바뀌고 누구도 1년 뒤를 예측하기 어렵다고. 그동안 당연하다고 생각했던 것들이 바뀌고 있고, 새로운 성공 방정식들이 생겨나고 있음.

그렇다고 무조건 스타트업이 이긴다는 행복회로 돌리는 건 아니고, 정신 바짝차리고 이 파도를 재밌게 타다보면 길이 보일 수 있다는 거.

모든 것이 불확실한 세상 = 뭐든 해볼 수 있는 세상

https://youtu.be/bdAaujNTTJg
Korean subnoscript

Andrej도 Reflection에 대해서 이야기했는데 Jim Fan도 이야기하네요 ㅎㅎ

GPT-4에는 다른 어떤 모델보다 매우 유용하고 강력한 기능인 셀프 디버그 기능이 있습니다.

아무리 숙련된 인간 프로그래머라도 첫 번째 시도에서 항상 프로그램을 올바르게 만들 수는 없습니다. 실행 결과를 살펴보고, 무엇이 잘못되었는지 추론하고, 수정 사항을 적용하고, 헹구고, 반복합니다. 에이전트 루프는 환경의 피드백을 받아 반복적으로 코드를 개선하는 것입니다.

이 논문을 적극 추천합니다: “코드 생성을 위한 GPT 셀프 리페어 이해하기“(다른 LLM과 비교하여 GPT-4의 셀프 디버그 기능을 정량화한 문서)를 적극 추천합니다. 몇 가지 주요 결과

GPT-4가 자가 복구가 가능한 핵심 이유는 강력한 피드백 기능입니다. 코드에 무엇이 잘못되었는지를 효과적으로 스스로 반영할 수 있습니다. 다른 어떤 모델도 따라올 수 없습니다.
피드백 모델과 코드 생성 모델이 같을 필요는 없습니다. 사실 피드백 모델이 병목현상입니다.
GPT-3.5는 GPT-4의 피드백을 받으면 훨씬 더 좋은 코드를 작성할 수 있습니다.
GPT-4 자체도 전문가의 피드백을 받으면 훨씬 더 좋은 코드를 작성할 수 있습니다.

Paper: Demystifying GPT Self-Repair for Code Generation
Link: https://arxiv.org/abs/2306.09896
Authors: Theo X. Olausson, Jeevana Priya Inala, Chenglong Wang, Jianfeng Gao, Armando Solar-Lezama

OpenAI는 많은 소프트웨어 엔지니어를 교사로 고용하여 다음 GPT를 교육하고 있을 가능성이 매우 높습니다. 그들은 생성할 필요가 없습니다. 비평만 있으면 됩니다.

Why transformative artificial intelligence is really, really hard to achieve
1. The transformational potential of AI is constrained by its hardest problems
2. Despite rapid progress in some AI subfields, major technical hurdles remain
3. Even if technical AI progress continues, social and economic hurdles may limit its impact

Source:https://thegradient.pub/why-transformative-artificial-intelligence-is-really-really-hard-to-achieve/

Thought

1. The transformational potential of AI is constrained by its hardest problems
=> AI로 특정 부분을 개선할 수 있으나 다른 부분들로 인해 병목이 생길거라고 말했는데 근거가 많이 빈약해서 설득되지 않네요. 오히려 하나의 Breakthrough가 나오면 새로운 문제도 더 빨리 풀지 않으려나요?

2. Despite rapid progress in some AI subfields, major technical hurdles remain
- 로봇 공학 속도가 느리고 Open AI도 로봇 공학 팀을 해체했다. -> OpenAI disbanded its robotic team because it found you could do all your training in a simulation, and it would work just fine. 이런 의견도 있고 저도 비슷합니다. Tesla는 왜 로봇을 만드나요? ㅎㅎ
- 다른 논리들은 완전 공감합니다.
2.1. The continued falling cost of computation could help. But we may have exhausted the low-hanging fruit in hardware optimization and are now entering an era of deceleration. Moore’s Law has persisted under various guises, but the critical factor for transformative AI may be whether we will reach it before Moore's Law stops.
2.2. Next look at data. Villalobos et al. warns that high quality language data may run out by 2026. The team suggests data efficiency and synthetic data as ways out, but so far these are far from complete solutions as Shumailov et al. shows.
2.3. Millions of humans currently annotate data to train models. Their humanity, especially their expert knowledge and creative spark, becomes more valuable by the day. The Verge reports: “One engineer told me about buying examples of Socratic dialogues for up to $300 a pop. “Since we are trying to behave in accord with people’s values, the most important data will be data from humans about their values.”

-> 과거 Annotation Startup이 나온 것처럼 새로운 기회라고 생각합니다. ㅎㅎ Human feedback market place가 나올 것 같아요 ㅎㅎ
-> Process knowledge is the kind of knowledge that’s hard to write down as an instruction. You can give someone a well-equipped kitchen and an extraordinarily detailed recipe, but unless he already has some cooking experience, we shouldn’t expect him to prepare a great dish.
-> 이 말에도 공감은 하지만 다른 생각을 가지고 있는데요. 카메라 기술의 발전과 Social media 덕분에 Instagram은 하루에도 10억장의 사진을 만들어내고 있는데요. AI 기반의 새로운 서비스/프러덕이 인류가 디지털에 가지고 있는 데이터 수를 압도적으로 늘릴 수 있다는 희망회로를 돌리고 있습니다 ㅎㅎ

3. Even if technical AI progress continues, social and economic hurdles may limit its impact
규제가 분명히 기술 발전의 속도를 늦출 것이지만, 글로벌하게 혁신을 촉진한 나라와 그렇지 않은 나라의 차이는 커질 것이라고 생각합니다.

- AI 문제 이야기인데 왜 대한민국이 생각날까요…한국은행 총재님이 하신말씀이 생각나네요..