Aaaaaah finally! It has been a long wait unitl another superb clip from AI Explained     See Less

I am not understanding how they can posibly automate the process of telling if a given reasoning and or answer is correct or not. I would think you need an AGI for that (a system that is abl     See More the correct answer by itself, to know wheather a given answer is correct or not)    See Less

I dont think your explanation is accurate at all. You seem to be mixing in RAG for more accurate answers with the CoT reasoning that the new model is utilising.     See Less

my best friend is fleeing to the netherlands and it sucks     See Less

I really wanted to like this channel, but I think while you are well-read and have a strong foundational vocabulary as well as a passion for AI, your videos lack depth. Without that depth, i     See More icult to genuinely provide value to your audience. However, you do present information in an unbiased and thorough manner, which is commendable.

I'm not here to tell you how to run your channel—this is simply feedback that I believe could benefit you. I usually don’t comment on videos unless I see potential in the creator's chosen genre or subject matter. You could create content that debunks new terminology from businesses and marketing teams, contrasting it with foundational terminology in AI research. Your ability to read between the lines in research papers positions you to become a much-needed resource in the scientific community, particularly in the role of peer reviewer.

Peer reviewers often have one of the least flashy and least respected jobs, and it's widely misunderstood regarding their importance. Many researchers are uncomfortable with having their work scrutinized, but since you clearly demonstrate aptitude in YouTube, this role ties into the integrity and transparency of science. Peer reviewers serve as a crucial component of the scientific community, much like the media is an essential part of democracy, yet this role is rarely acknowledged.

Best of luck, and just in case anyone reaches the end of this message, here’s a brief overview regarding "reasoning" in AI models:

- *Traditional AI Reasoning (Symbolic AI):* This approach relies on explicit rules, logic, and knowledge bases. It is generally more explainable and can be formally understood.

- *Modern AI Reasoning (Connectionist/Statistical AI):* This method uses patterns learned from large datasets to predict likely outcomes. It is based on statistical inference and may not always be as transparent or logical in the traditional sense.    See Less

Why don't they use the reasoning technique from o1 on multimodal world models? They could be immensely powerful.     See Less

If you take request’s: Max Tegmarks new paper pls.     See Less

11:54 it's weird to hear this, I had this exact problem yesterday. I was working with Llama 70B and the function f(x)     See More 1) was part of what we were working with. At one point, the model calculated f(100) terribly wrong, so I asked it, "wait, did you multiply 1.09 by itself 99 times to get that answer? And it said "no let me try that" and then typed out the words "I am multiplying 1.09 x 1.09 x 1.09... 99 times" and then got the correct answer despite not actually even typing (thinking) through all 99 steps. Was a wild moment for me.    See Less

gotta watch this video again and again to understand the impact.... what future will we have man. Can someone pinch me? Is this real haha     See Less

Either most of the comments are Ai generated, or the viewers of this channel writes like chatbots     See Less