erik8lrl

I think although it's definitely important to keep AI accessible for people and not be owned by major corporations, companies like OpenAI, StabilityAI, or Tesla are operating not only for profit, tho profit is important. Their goals for advancing tech and helping humanity are real. I think Sam Altman said that his biggest fear is to release something with good intentions but people end up using it the wrong way which leads to disasters. Hence why they are releasing things slowly to give society time to respond.

Yes, this will be the case, anyone will be able to make a film soon. Of course, it still takes craft to make anything good, and there are still limitations to AI. However, the release of these models will be a huge deal for independent and documentary filmmakers. Hollywood will likely take full advantage of it as well. It will be implemented into all the workflows to make things faster and cheaper. For some AI is scary because it might very well take their jobs. But for others, it's a major game-changing development that will allow them to do what was impossible. I think people should all learn AI, and keep up with the development. We are already at a time where if you don't incorporate it into your workflow, you will be far less productive than others who do, and eventually be replaced. I think smart business owners will not seek to just replace people with AI, but train people to learn to use AI to multiply their productivity. You can keep the same amount of people but instead produce 5-10 times more results. This will likely be the case for film.

4 is definitely better than 3.5 at many benchmarks. The best thing about 4 is getting access to vision, web search, and customGPTs. Vision allows it to understand images. Web search allows it to stay up to date with information. (3.5 only has information up to 2023 I think). The best thing is customGPTs, both the ability to train/finetune your own GPTs, and access GPTs other users made from their own proprietary data. There are customGPTs for any task possible. You can also release your own GPTs to earn profit based on how many users use them.

You can also run local open source LLMs instead of ChatGPT if you want.

Close to 800. Tho I’m not doing as much linking as possible, mainly very large notes with tags. If I linked every concept there would be 10 times more notes. But it’s really time consuming to do that.

You can import OneNote into Obsidian pretty easily with this plugin: https://help.obsidian.md/import/onenote

Try Copilot maybe, it should be similar but for OneNote. Obsidian is really simple if you want it to be, it's open-source so people have developed endless plugins for it to do whatever you want. There are also plugins for simplifying everything to be more minimalistic. Its power comes from the level of customizability you get. From a note-taking perspective, all you really need to know is linking, and tagging. This simple function allows for really good interconnection between notes over long periods of note-taking. I used to use OneNote too, but with Obsidian my note-taking process is fundamentally different. Obsidian incentivize making connections between notes, so even when you are writing new notes, you are thinking from the perspective of your whole knowledge vault to make connections with what you already have. It trains your brain for non-linear thinking and insights. For ChatGPT, if your notes are well linked and clear, then it can totally understand everything. It is for my case.

Not the exact same, but Copilot is on OneNote I think. The power of Obsidian is that you can connect different notes with tags and links, which allows ChatGPT to read and understand your note structure.

Yes, tho many open-source models are being developed and released too. https://x.com/thibaudz/status/1761506136455340470?s=20

I see your perspective and I agree. No one knows what will happen in the future, but let's hope there will be countermeasures to these problems. I work in film production, so from my experience, if you have mastered the ins and outs of AI art generation, you can absolutely produce work that is completely unique and self-expressive. Of course, most people will just prompt and be done so I see your point of diminishing the value of art. However, it goes the other way too, I go through a very complex workflow in order to get the image I want down to every detail, which has totally unique style and wholeness developed through a long process of iteration. There are quite a lot of tools and techs that give you total control over the image generation process, and new tools are being released daily. We use AI mostly as a tool for quick iteration and the starting point of our work, you still need to edit and paint and work on these generations to fix things and polish them, it's really a mixture of AI and human traditional work that allows you to make something good. AI simply speeds up the production process by 4-5 times. Most people only tried generation models like Midjourney or Dalle and don't know how much depth and freedom you can have with image generation. For example, we could have our artist develop an art style, and paint a series of concept art in that style, then we train our own Lora model and then be able to generate images from that style. We could use ControlNet and other tools to control every aspect of the image, and this allows us to iterate very quickly in our production process. The artist is still very much present in these works, they are the ones prompting with both text and image and often painting fast compositions or color palettes to guide the AI, they pick out what's good based on their artistic taste and vision. They then edit and improve the work to make sure every part of the work is whole and meaningful. It helps save a lot of time and the work they produce is better overall because they can try more things within a shorter amount of time frame. It all depends on how you use it.

Yes, I think your point of view is valid but different from the perspective most people in this post are coming from. Yes, art or intelligence without a conscious agent loses its meaning from the perspective of other conscious agents. The value/meaning of art is through the exploration and expression of one's self, which AI doesn't have yet. However, while AI can't create meaningful art, humans have used AI to create art that is self-expressing. The AI doesn't have agency, but humans can use it to create meaning. I think most of us are speaking less from an art philosophy perspective and more from a scientific perspective. Just simply pointing to the fact that if these AI kept advancing at the speed they are, will impact society greatly. Not because the AI can or can't do something, but because of how humans will use it to create or destroy things. It's as you said, humans use neural networks too, but if we have AGI, it means that everyone will have a neural network the size of the entire knowledge of humanity. "Intelligence" in this instance is less about conscious behaviors but more so a democratization of knowledge and understanding through AI. Which when used by humans, could either lead to greatness or disaster.

This is what I've been asking. Neural networks are not just "knowing a lot of facts". It also connects them to form associations and meanings, this is the source of why they seem to be "intelligent" because they can make connections and generalizations similar to humans, even tho their ability is not perfect.

https://www.vice.com/en/article/y3w4am/scientists-claim-ai-breakthrough-to-generate-boundless-clean-fusion-energy

https://x.com/cb_doge/status/1759847548578333021?s=20 Elon is soon to implement AI for sense-making on X. "What I'm really trying to achieve with 𝕏 is to improve the signal to noise of the collective consciousness." https://x.com/cb_doge/status/1759837783877947487?s=20

Yes, exactly.

AI Chip breakthroughs also happening, hardware is accelerating as well.

This shows how AI expert's prediction of the AGI timeline keeps getting shortened as the exponential curve progresses.

Yeah, I think current models definitely have this problem, especially for complex logical reasoning. It might be solved with more parameters or better architecture. We do see that the scaling of "intelligence" and reasoning ability increase as we scale. I think the new Gemini 1.5's reasoning ability scored 80-90% on benchmarks. We'll see if newer models like GPT5 improve this area, it might or might not be a problem related to scale since we still don't fully know what emergent property can appear as we scale larger and larger. I think for me, a lot of the times when GPT4 doesn't do what I ask, it's mostly because of my prompt not being specific or detailed enough, most of the time I find improving my prompting helps, but not always tho. I think for coding, context window size does help, for me, most of the code GPT4 gives me is out of context, so they are often the most common implementation of something. If Gemini can generate code in the context of the whole system, then it should produce better results.

Yeah, I think this is mainly caused by the size of the context window and the architecture. Right now, even with GPT4, the context window is small, meaning it can't remember long-term conditions naturally. You could get around this sometimes with prompting, something like "Don't do (this) from now on when answering my questions". You can also use the new memory feature that's released recently, it basically allows you to add a system prompt to the GPT so that it remembers it and applies it globally. I think if you are using GPT4/Dalle3 to generate images, you won't have a good time with customization. Dalle-3 is really good at text interpretation, but it can't really do inpainting or negative prompts. How it changes images is by reverse prompting text description from your images, and then using that prompt to generate similar results, but it's not editing the images, so you won't have consistency. GPT4/Dalle-3 is really only good for generating initial images that require complex text interpretation. For image generation, I think it's best to use Midjourney, SD, or if you can, learn comfyUI. These give you far more customization ability and negative prompting works great for them. With comfyUI you can even customize the lower-level processes of the models to fit your needs, there are also endless workflows shared by other users for specific image generation tasks. GPT4/Dalle-3 is really not designed for image generation, GPT5, however, will be a different story.

Totally. It's possible they already have AGI or getting really close.

Yes, if you want to word it this way. It's true, that we don't inform intelligence through data alone, we inform them through connection and pattern recognition, which is what neural networks do also aside from data collection. My point is we don't know that it is fundamentally different for sure, it's still too early to tell. We don't know enough about neural networks to definitively say that it might not be the starting point that could lead to Qualia.

I sent these articles to GPT4 to ask its opinions lol, and this is what it gave me: ChatGPT The argument that GPT-4, and by extension, other large language models (LLMs), cannot truly reason is valid within the context provided. It highlights the distinction between the appearance of understanding or reasoning and the actual cognitive processes involved. LLMs are trained on vast datasets to predict the likelihood of sequences of words, which allows them to generate plausible-sounding responses. However, this process does not equate to genuine reasoning or understanding, as it lacks the ability to consider context, apply logic consistently, or understand the underlying principles in the way humans do. The critique is rooted in the current limitations of AI in mimicking the depth and flexibility of human thought processes. User Do you think this will be improved as the size of data and parameters increase? ChatGPT Improvements in AI's reasoning capabilities with larger datasets and more parameters are expected to some extent, as these enhancements can lead to better pattern recognition, more nuanced language understanding, and more sophisticated response generation. However, true reasoning involves more than just processing power or data scale; it requires the ability to understand context, apply logical principles, and make inferences based on a deep understanding of the world. While scaling up might address some limitations, addressing the fundamental gap between simulating reasoning and actual understanding will likely require advances in AI approaches and architectures beyond just increasing size. User What do you think such advances could be? Search the internet for the latest research if you need to. ChatGPT Advances in AI reasoning capabilities are indeed on the horizon, driven by innovative architectures and training paradigms that could potentially bridge the gap between current limitations and more human-like reasoning. Two notable developments are Google DeepMind's Gemini 1.5 and Google's Pathways Language Model (PaLM), as well as insights from Yann LeCun on building AI systems that learn and reason like humans and animals. Google DeepMind's Gemini 1.5 introduces enhanced performance through a more efficient architecture, capable of sophisticated reasoning tasks across different modalities, including text and coding. This model represents a significant step forward, surpassing previous state-of-the-art models in understanding and problem-solving capabilities across a range of benchmarks. It shows that as models become more capable and architectures more efficient, AI's ability to reason and understand complex scenarios improves significantly. Google's PaLM, with its breakthrough in scaling to 540 billion parameters, demonstrates impressive capabilities in language understanding, generation, and reasoning tasks. Its use of chain-of-thought prompting for multi-step arithmetic and common-sense reasoning tasks showcases how large language models can be pushed towards more sophisticated reasoning abilities. The model's ability to generate explicit explanations for complex scenarios indicates a move towards AI systems that can not only solve problems but also articulate their thought processes in a way that mimics human reasoning. Yann LeCun proposes an architecture for autonomous intelligence that could enable machines to learn world models in a self-supervised fashion. This approach involves constructing predictive models of the world that can simulate possible future states and use these simulations to predict, reason, and plan. Such an architecture, which includes modules for perception, world modeling, and action planning, represents a comprehensive framework for developing AI systems capable of complex reasoning and decision-making. These advancements suggest that while scaling up data and parameters can lead to improvements in AI's reasoning capabilities, significant leaps also require innovative approaches to model architecture and training. By developing models that can better understand and simulate the world, and by employing strategies like self-supervised learning and chain-of-thought reasoning, we move closer to AI systems that can reason in ways that are more aligned with human cognitive processes. For more details, you can explore the developments in Gemini 1.5 on Google DeepMind's website and PaLM on Google's Research Blog, as well as Yann LeCun's vision for AI on Meta AI's page. --- The world simulation model is exactly what OpenAI, Google, and even Tesla are working on right now. We can already see it with Sora, GPT5 will likely include Sora in its model understanding. Similar with Google Gemini going multi-modal.